Basic level questions and tutorials on our education resources pages are aimed at undergraduate students.
BSH education resources
Our educational resources can support your professional development. Whether you are an undergraduate, a junior doctor, a specialist trainee, a nurse, allied health professional or a healthcare scientist – there will be something here to suit you.
These resources are open to non-members but you will need to create a profile on our website.
Careers in Haematology
Our Careers in Haematology booklet (PDF), aimed at undergraduate students, discusses the attractions of haematology, the range of conditions that haematologists treat and what makes a good haematologist.
This curriculum outline acts as a guide both to students and medical schools regarding the extent and depth of knowledge of haematology that would be expected of newly-qualified doctors commencing their first posts.
It is expected that this knowledge will be acquired during both the pre-clinical and the clinical years
Essential Haematology Day
Essay Prize 2017
The British Society for Haematology invited undergraduate medical and BMS students from UK schools in any year of training to write an essay on: 'Discuss the possible roles of Artificial Intelligence in the future practice of haematology'
Applications for this award have now closed.
In 2017 the essay title was 'Discuss the possible roles of Artificial Intelligence in the future practice of haematology'. The winner was Marco Narajos, and the runner-up Karan Dahele.
It is said that there are as many definitions of Artificial Intelligence (AI) as there are AI researchers in the field. Many understand the term AI to mean a computing system that behaves as a human mind would. However, many AI systems do not, in fact, aim to mimic the vast complexity of the human mind, let alone that of a haematologist!
In simplest terms, all that computers do is take an Input through a Process to obtain an Output. Machine learning algorithms, for example, can enable computers to learn for themselves, such as by presenting the computer with Inputs and Outputs and prompting it to figure out the Process all by itself, without a human programming it in the first place. We can achieve this so-called ‘deep learning’ in different ways, such as with an artificial neural network (ANN) – a computer system inspired by the rules that govern how neurons in the brain send impulses.
Deep learning with ANNs is only one type of machine learning and machine learning is only one facet of AI. However, we humans, and especially haematologists, pride ourselves on our ability to learn and solve problems. Thus, over the last few years, machine learning has become synonymous with AI. This has not always been the case, as many authors have previously used the term AI to refer to decision support systems that make clinical judgements from complex information like a human clinician would.1
No matter what definition or aspect of AI is in vogue, it is clear that AI will play significant roles in the future practice of medicine. This paper seeks to identify and predict how AI, focusing on machine learning, will interact with diagnostics, prognostics, therapeutics, research, and even management in haematology.
Haematology’s strength as a specialty lies in its diagnostic capability. Not only is haematology its own valid specialty, it also acts as a clinical support service for primary care and inpatient settings. For this reason, a fast turnover of laboratory results and accurate diagnoses are important, and AI is on its way to doing just that.
In 2009, an Iranian group developed an ANN that they trained to extract features from a dataset of 360 patients and validated it on a sample of 90 others.2 Because of the way ANNs mimic how neurons take on multiple inputs from other neurons, this form of AI has the ability to handle multiple input parameters of different weights or levels of importance. In this case, 11 values from a full blood count (FBC) alone could simultaneously churn out the absence or presence of five diseases – megaloblastic anaemia, thalassaemia, idiopathic thrombocytopaenic purpura, chronic myelogenous leukaemia, and lymphoproliferative disorder – with supposedly near 100% accuracy. There are many flaws in the paper but it is proof of concept that an AI can make a FBC alone be of diagnostic significance. This is especially important in developing countries where there may not be access to other investigations or indeed consultant haematologists.
The holy grail would be the ability to train a computer to report on and even make diagnoses from blood films and bone marrow samples. A Nature Communications paper sparked a media frenzy last year over AIs supposedly outperforming human physicians.3 In fact, the published work was for programming an AI to learn to differentiate between lung adenocarcinoma and squamous cell carcinoma from H&E-stained slide images alone and to extract features from the images that would predict mortality. The media frenzy, no doubt, made sure that AI was the cynosure of all investors’ eyes, but in truth, the AI was trained and validated on images that had unequivocal adenocarcinoma and squamous cell carcinoma features. Pathologists, and haematologists likewise, will see cases on the borderline with artefactual distortions, which means that AI trained on model images would unlikely outperform a clinician. Also, even though we consider the human eye as the gold standard diagnostic test, intraobserver concordance between consultants may not even approach 100%. For this reason, AI still has a long way to go before reporting on blood films and bone marrow samples on behalf of human doctors.
Despite that diagnosis is often an easier task than prognostication, more research exists for using AI in prognostics in haematology, perhaps reflecting haematologists’ eagerness to peer closer into the crystal ball. One 2010 paper examined how well an ANN machine learning algorithm could predict acute graft-versus-host disease (aGVHD) after unrelated donor haematopoietic stem cell transplantation (UD-HSCT) for β-thalassemia major.4 The Italian team compared the ANN with logistic regression. This is a type of classical statistical analysis that can examine how variables determine an outcome, such as the absence or presence of aGVHD. They used 24 variables to train the ANN to analyse the most important factors that associate with aGVHD. Although the ANN did not calculate which factors were most important, it performed significantly better at predicting who would not develop aGVHD, with a specificity of 83% vs 22% for logistic regression. There was no significant difference between the ANN’s sensitivity at picking up future aGVHD, which was high at 90% compared to 81% for logistic regression. This is proof of concept that machine learning can be more effective than traditional statistical methods at predicting complications of haematological procedures.
Many of our haematological therapies, especially when used in combination, have narrow therapeutic windows: too small a dose and we risk not treating the disease, too large a dose and we risk toxic side effects. Much like any field of clinical medicine, haematologists titrate drug doses based on clinical wisdom and target values.
Decision support systems (DSSs) make titration judgements easier and simpler, and have been around for a relatively long time. An example would be HiruMed’s anticoagulant dosing system (RAID), which does not use machine learning AI. However, the problem with DSSs is that their software has to be reprogrammed with new parameters every time there is new data that could change clinical decisions. Machine learning algorithms solve this issue because each new clinical data point added to the system allows the AI to learn for itself what factors contribute towards better or worse maintenance of a patient’s target value – in this case, the INR. It would be a tool not only for therapeutics but also for research, as so-called ‘supervised learning’ can allow the AI to identify patients who are at high or low risk of adverse events, and therefore, affect clinical management by ensuring the patient is seen or assessed more or less frequently.
A Swedish group tasked themselves with developing a DSS that would do just this, using Bayesian networks (BNs).5 A BN is a type of machine learning that uses Bayesian statistics, which is a more realistic or ‘more human’ way of inferring information from probabilities than traditional ‘frequentist’ statistics commonly taught in schools. Bayesian statistics looks at a priori information, such as a patient’s history of (recent) bleeding, strokes, alcohol consumption, and other variables, to determine a probability. Intuitively, this is how all clinicians think. The beauty in this algorithm is that it can provide a recommendation (such as a 10% dose increase in warfarin) and also a second probability that this recommendation would work to improve a patient’s INR value. The more information we provide about the patient, the more we can increase the algorithm’s confidence in its recommendation because BNs rely on a priori knowledge to adjust their outputs. The team trained the DSS with data from over 3,000 patients on nearly 50,000 hospital appointments. This impressive feat of Scandinavian recordkeeping has allowed this DSS to learn ‘rules of dose titration’, which agreed with hospital clinicians over 75% of the time. Whether the dose titrations that the DSS recommended performed just as well as clinicians is another matter. Interestingly, the DSS’s recommendations on the next follow-up date only agreed with clinicians around half of the time, with the DSS recommending shorter follow-up times.
It is important to emphasise that machine learning is not at all needed to personalise dosing to improve treatment response and safety from side effects. One group in Los Angeles developed a complex statistical model that they call phenotypic personalised medicine (PPM).6 PPM does not involve machine learning, and yet, science news media, including Nature Medicine, have bestowed it with the title of AI.7 PPM is a way of using a patient’s readouts of treatment efficacy and safety to determine the appropriate medication dose using the patient’s own record of how well they have done on a given dose of the drug. It consists of an elegant series of quadratic equations that use mathematical constants that clinicians can calculate individually for each patient. Instead of learning from thousands of patients to give a recommendation of what dose to provide (as in machine learning), PPM uses the patient’s own data (termed ‘n-of-1 medicine’). This makes personalised dosing possible without using machine learning. However, whether PPM is ‘true’ AI is debatable.
The Californian group developed this model for the maintenance therapy of paediatric patients with acute lymphoblastic leukaemia, and in their paper, showed a comparison between a patient whose drugs were dosed clinically compared to one whose drugs were dosed by PPM.8 According to clinical protocol, the clinician could only modulate the dose of 6-mercaptopurine and methotrexate, whereas the PPM was also able to modulate the doses of vincristine and dexamethasone. Clinical modulation led to significant clinical deviations away from the absolute neutrophil count (ANC) target, whereas the PPM kept the patient within the ANC and platelet count targets. Of course, the study is plagued with issues, as it was not a randomised controlled trial, clinician blinding was impossible, and the outcome measures were not patient-centred (e.g. mortality, quality of life, symptomatology). However, it is evidence that something AI-like can aid haematologists in their approach to therapy.
Machine learning AI has the capacity to find patterns in data. In supervised learning, the AI has specific patterns it is taught to look for, whereas unsupervised learning allows the AI to search for features in the data that can be grouped together, even if the developers had not asked the AI to investigate specific variables. Both approaches are incredibly powerful tools in hypothesis-testing and hypothesis-generating research, respectively.
A group from Harvard used a supervised learning approach to direct a piece of machine learning AI to find a gene expression profile in diffuse large B cell lymphoma (DLBCL) at the time of diagnosis that would be predictive of a prognosis, such as cured disease versus fatal or refractory disease.9 From the 58 patients that the researchers separated into cured and fatal/refractory, they found 13 genes that were predictive of the patients’ prognosis.
In contrast, we can take an unsupervised learning approach to classify patients with DLBCL into two groups based on similar gene expression profiles, without directing the AI to specific genes.10 This led to the discovery of two putative types of DLBCL, associated with gene expression profiles and prognosis. This discovery was replicated by another group who used a supervised learning method with an ANN, which improved both diagnosing and prognosing DLBCLs based on the gene expression profile only.11
AI clearly has a major role to play in the future of haematology research, such as in finding markers of subgroups of disease that would aid our diagnostic classifications and prognostic understanding of disease.12 AI could potentially also support research logistics, such as identifying the patients who would be suitable for and amenable to entering a clinical trial.
Management of a clinical haematology department and its various services is an important part of a consultant haematologist’s role. AI’s ability to learn from patterns in data, without the need for programming of rules, could help to streamline management by learning from datapoints such as patients’ addresses, staffing levels, (non-)attendances, numbers of inpatient and GP requests for haematology blood tests and transfusions, and costs of procedures and equipment. The AI could perhaps assist with budgeting, allocation of clinic time, audits, clinical governance, and supply and procurement of new equipment.
No research yet exists on how AI will affect the managerial role of haematologists, but it is a unique opportunity given the hybrid specialty that haematology is – part-laboratory and part-clinical. It is also feasible, given that software is already available to manage a haematology department, such as the DAWN management software package, which even includes its own, tried-and-tested dosing support system.13
Aside from management, there are also various opportunities for AI in haematology such as the development of chatbots that can answer queries from patients (or indeed medical students) about their care, which could improve compliance with medication, clinic attendance, and even medical students’ education.14
Many haematologists will muse about whether they are at risk of losing their jobs; most will assure themselves, rightly, that they are not at risk. All jobs are simply made up of a series of tasks and decisions. At the present moment, AI has such a limited repertoire of tasks that it can perform and decisions that it can make in haematology. For this logistical reason, there is certainly no imminent worry that haematologists will be replaced by AI.15 However, there are other sociocultural reasons that doctors are likely to keep their jobs. As clinicians, we take on a social and legal responsibility by being tasked to diagnose, prognose, and treat a patient. Furthermore, we as clinicians provide much emotional labour; AI will never be able to deliver or replicate empathy, companionship, and the human touch.16
Nevertheless, it is impossible to ignore the relentless rise of machine learning which has come in recent years, powered by the nascent prominence of the trifecta of Big Data,17 greater processing power,18 and state- and industry-led investment.19 In the not-too-distant future, it is likely we will see integration of AI in haematology clinics, laboratories, and even in mobile phone apps. We will probably see AI helping nurses to run clinics, in lieu of haematologists and GPs to reduce workload and address the shortage of doctors.20 AI could also take the form of an app prompting the haematologist with a list of differential diagnoses with the statistical probabilities of those conditions, cementing the role of the haematologist in providing his or her final diagnostic conclusion. Additionally, AI could assist haematologists with continuing to personalise therapies to the patient’s disease as well as ideas, concerns, and expectations of treatment. AI-powered research will also become more prominent in the coming years with the growing number of machine learning developers and their collaboration with haematologists could lead to important discoveries of causes and treatments of haematological disease. Moreover, AI could have a small but not insignificant role to play in managing growing, busy haematology departments to use data to improve organisation, decrease patient and doctor time wasted in clinics, and drive productivity.
All haematologists will need to be aware of AI – its types, mechanisms, benefits, and limitations for the practice of haematology – and how to utilise it to improve patient care, optimise services, and drive efficiency savings in an ever resource-limited society.
- Barbieri, B., Gamberoni, G., Lamma, E., Mello, P., Pavesi, P., & Storari, S. (2005). An expert system for the oral anticoagulation treatment. Innovations in Applied Artificial Intelligence, 7-18.
- Payandeh, M., Aeinfar, M., Aeinfar, V., & Hayati, M. (2009). A New Method for Diagnosis and Predicting Blood Disorder and Cancer Using Artificial Intelligence (Artificial Neural Networks). International Journal of Hematology-Oncology and Stem Cell Research, 3(4), 25-33.
- Yu, K. H., Zhang, C., Berry, G. J., Altman, R. B., Ré, C., Rubin, D. L., & Snyder, M. (2016). Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features. Nature communications, 7.
- Caocci, G., Baccoli, R., Vacca, A., Mastronuzzi, A., Bertaina, A., Piras, E., ... & La Nasa, G. (2010). Comparison between an artificial neural network and logistic regression in predicting acute graft-vs-host disease after unrelated donor hematopoietic stem cell transplantation in thalassemia patients. Experimental hematology, 38(5), 426-433.
- Yet, B., Bastani, K., Raharjo, H., Lifvergren, S., Marsh, W., & Bergman, B. (2013). Decision support system for Warfarin therapy management using Bayesian networks. Decision Support Systems, 55(2), 488-498.
- Zarrinpar, A., Lee, D. K., Silva, A., Datta, N., Kee, T., Eriksen, C., ... & Wang, S. E. (2016). Individualizing liver transplant immunosuppression using a phenotypic personalized medicine platform. Science translational medicine, 8(333), 333ra49-333ra49.
- Chakradhar, S. (2017). Predictable response: Finding optimal drugs and doses using artificial intelligence. Nature Medicine 23, 1244–1247.
- Lee, D. K., Chang, V. Y., Kee, T., Ho, C. M., & Ho, D. (2017). Optimizing Combination Therapy for Acute Lymphoblastic Leukemia Using a Phenotypic Personalized Medicine Digital Health Platform: Retrospective Optimization Individualizes Patient Regimens to Maximize Efficacy and Safety. SLAS TECHNOLOGY: Translating Life Sciences Innovation, 22(3), 276-288.
- Shipp, M. A., Ross, K. N., Tamayo, P., Weng, A. P., Kutok, J. L., Aguiar, R. C., ... & Ray, T. S. (2002). Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nature medicine, 8(1), 68-74.
- Alizadeh, A. A., Eisen, M. B., Davis, R. E., Ma, C., Lossos, I. S., Rosenwald, A., ... & Powell, J. I. (2000). Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature, 403(6769), 503-511.
- O'Neill, M. C., & Song, L. (2003). Neural network analysis of lymphoma microarray data: prognosis and diagnosis near-perfect. BMC bioinformatics, 4(1), 13.
- Zini, G., & d'Onofrio, G. (2003). Neural network in hematopoietic malignancies. Clinica chimica acta, 333(2), 195-201.
- Poller, L., Keown, M., Ibrahim, S., Lowe, G., Moia, M., Turpie, A. G., ... & Palareti, G. (2009). A multicentre randomised assessment of the DAWN AC computer-assisted oral anticoagulant dosage program. Thromb Haemost, 101(3), 487-494.
- Kerly, A., Hall, P., & Bull, S. (2007). Bringing chatbots into education: Towards natural language negotiation of open learner models. Knowledge-Based Systems, 20(2), 177-185.
- Manyika, J., Chui, M., Miremadi, M., Bughin, J., George, K., Willmott, P., & Dewhurst, M. (2017). A future that works: Automation, employment, and productivity. McKinsey Global Institute, New York, NY.
- Luneski, A., Konstantinidis, E., & Bamidis, P. (2010). Affective medicine: a review of affective computing efforts in medical informatics. Methods of information in medicine, 49(3), 207-218.
- Lohr, S. (2012). The age of big data. New York Times, 11(2012).
- Steinkraus, D., Buck, I., & Simard, P. Y. (2005, August). Using GPUs for machine learning algorithms. In Document Analysis and Recognition, 2005. Proceedings. Eighth International Conference on (pp. 1115-1120). IEEE.
- Chui, M. (2017). Artificial intelligence the next digital frontier?. McKinsey and Company Global Institute, 47.
- Fitzmaurice, D. A., Hobbs, F. R., Murray, E. T., Holder, R. L., Allan, T. F., & Rose, P. E. (2000). Oral anticoagulation management in primary care with the use of computerized decision support and near-patient testing: a randomized, controlled trial. Archives of Internal Medicine, 160(15), 2343-2348.
Artificial Intelligence: the future of haematology?
Artificial intelligence (AI) is broad, often misunderstood term. It refers to the science of simulating the brain’s cognitive ability within a computational system (1). In popular culture, the term is most often associated with robots in a dystopian future. In the media, AI often features in sensationalised headlines about job losses. This representation has fuelled fears of the impact its development will have on society.
The rapid development of AI has facilitated its use to solve a diverse range of problems in recent years, from playing high level strategy games, to understanding natural human speech. AI has huge application in medicine; there is exciting commercial and academic activity in cellular biology, genetics, drug discovery, diagnostics, treatment, and outcome prediction. There is no doubt that AI will play a huge role in healthcare in the future. This essay aims to introduce some of the key principles of AI, present some of the most exciting developments in haematology, and discuss how AI might shape the practice of haematology in the future.
Principles of AI
Machine learning refers to the process by which software takes data, “learns” from it, and produces functions to interpret new data. This is opposed to a traditional “hard-coded” predictive model, where the person designing the software would specify the function producing an output for each input. For example, in designing software to determine the probability of anaemic symptoms based on iron intake, a hard-coded solution might be programmed to increase risk by X%, below a certain value of iron intake. On the other hand, a machine learning solution would tackle the problem by analysing a large dataset of relevant patients, and learning itself the nature of the relationship between the two factors. This approach quickly comes into its own when working with large, multi-factorial patient data, where manually finding complex patterns would be difficult.
There are two broad types of machine learning: supervised and unsupervised.
In supervised machine learning, the system is provided with a set of inputs, and a set of outputs. Algorithms aim to produce a function which will map the input to an output. Training refers to the process of feeding the system with data, so that it can learn trends and produce this function. This algorithm can then be applied to real cases, taking inputs and providing predictions. The system has been given data explicitly labelled with inputs and outputs (2).
The crucial difference in unsupervised learning, is that the system is not given an explicit outcome variable. The dataset used to train the system only has inputs. The model is tasked with finding trends and associations, without a specific outcome to link to; algorithms have free rein to elucidate any interesting patterns (2).
In medicine, the most popular machine learning algorithms are currently support vector machines (SVM) and artificial neural networks (ANN) (2).
SVMs are used to classify a subject into one of two groups. Different traits (inputs) can impact into which group the subject is classified. Each trait is weighted, depending on how much it correlates to each of the two classification groups. Training an SVM involves optimising the values of the weights, so that the model’s predicted classification most correctly matches the actual classification given in the data set.
An ANN is modelled on the organisation of the brain, with artificial neurones connected in layers. Each artificial neurone receives a set of inputs, which are weighted (the “synapse” of the neurone), and produces an output. Each input’s weighting can be optimised through training (akin to an SVM). The layering of neurones allows the relationship between input and output to be broken down into smaller steps, allowing more complex patterns to be identified. These patterns might not be straightforward correlations between variables. A simple neural network might have 3 layers of neurones: an input, hidden, and output layer. Conversely, a deep neural network has many layers of neurones in between the input and output neurones (shown in Figure 1).
Figure 1: Simple and deep neural networks (3)
AI has potential uses in every stage of biomedical research and clinical practice.
Clinicians, including haematologists, are involved in both benchside and clinical research. AI will play a crucial role in biomedical and pharmaceutical research. Artificial neural networks can analyse sparse or noisy data, such as genomic sequences, which would pose a problem for standard statistical methods. They are being used to aid prediction of protein folding based on genetic data (4), which forms the basis of diseases such as sickle cell. AI has also shown potential in designing new diagnostic pathways for clinicians to use. For example, a data mining approach was used to simplify the diagnostic process for Polycythemia Vera. The model produced an accurate diagnostic algorithm for clinicians to use, with four parameters (Hct, Platelet count, Spleen and WBC), in contrast to the existing pathway requiring eight parameters (5).
Drug discovery is becoming increasingly expensive and difficult, with drugs frequently failing late clinical trials. Researchers are seeking new higher yield approaches to discover therapeutics. AI is being used to investigate potential drug targets in the treatment of Diamond Blackfan anaemia (6), a rare genetic disorder. Researchers are crossing many different drug compounds with cells of many genotypes (“high-content high-throughput” – around 1000 cell features per cell, with 10 million cells per week), and measuring the outcomes using immunofluorescent imaging. This is then analysed by machine learning algorithms, which suggest the most promising combinations for further investigation. AI is helping make sense of the “data deluge” produced by high-throughput approaches in biomedicine, but still requires highly knowledgeable clinician-scientists to oversee the training of models and to interpret results.
As well as research, AI can play an important role in diagnostics. Automated systems for carrying out full blood counts are common, and have made this investigation relatively rapid and inexpensive. A peripheral blood smear is another frequently used investigation, important for diagnosis of disorders of erythrocytes, leukocytes, and haemoglobin, as well as bloodborne infections. Currently, interpreting blood films to find diagnostic markers is done manually, requiring expert knowledge. There has been considerable progress in using deep neural networks to interpret blood films. Systems have been developed that can diagnose sickle cell disease with high accuracy, and even categorise sickled cells, with the potential to make predictions on severity of the disease (7). Researchers have also trained convolutional neural networks to identify markers of lymphoma in H&E stained histopathology slides of human lymph node tissue. They reported a 38% decrease in residual error over current comparable techniques (8). As well as interpreting images, AI has been successful in incorporating genomic sequence data into diagnosis. IBM Watson diagnosed a patient with a rare form of leukaemia, initially misdiagnosed by clinicians, by cross-referencing the patient’s genome with millions of known mutations learnt from oncology papers. Over the past year, Watson has been used for around 100 patients with haematological diseases at the same hospital (9). However, Watson has still not proved effective at diagnosing other types of cancer, largely due to a lack of accurately labelled, comprehensive patient data to train with.
As well as diagnosis, AI is shaping how haematologists treat patients. One example is the management of anaemia associated with chronic kidney disease (CKD). Damage to the kidneys can cause them to fail to produce enough erythropoietin, which leads to a hypoproliferative anaemia. Anaemia becomes more prevalent as CKD progresses, with almost all end stage CKD patients developing anaemia. Erythropoiesis stimulating agents (ESAs) are commonly used to treat the condition, despite associations with off effects, including cardiovascular events and malignancy. It is important to use the lowest effective dose, which still maintains haemoglobin concentration, whilst minimising dangerous side effects. However, it is difficult to calculate this optimal dose. In 2016, researchers used an artificial neural network to provide better individualised ESA dose recommendations (10). This approach resulted in higher rates of on-target haemoglobin values, less haemoglobin level fluctuation, lower average doses of ESA, and lower transfusion requirement. This study illustrates the role AI can play in personalising treatment.
Humans have inherent biases, which often impact clinical judgement. Doctors are considerably over-optimistic in estimating life expectancies, whilst AI models are much more accurate in estimating prognoses. For example, a Random Forest machine learning model has been used to predict likelihood of relapse of childhood acute lymphoblastic leukaemia with high accuracy (11). Other machine learning models have been used to predict mortality following allogenic hematopoietic stem transplantation, used to treat some patients with haematological malignancies (12). If used clinically, these models, and others like it, would be useful to haematologists to help stratify patients and personalise treatment plans.
Models like these could be considered to diminish clinicians’ autonomy. However, approaches taken to historical advancements in technology, such as imaging, suggest that these models may be integrated into a clinician’s decision-making process, rather than replacing it. It would form one more point of consideration among many other patient factors.
The future of haematology
A haematologist is responsible for the investigation, accurate diagnosis and therapeutic management of blood disorders. Haematologist have a range of roles: working in laboratories to interpret investigations, as well as caring for patients on wards and in outpatient clinics. To what extent will AI replace or aid in these roles?
Figure 2 Supply and demand for clinicians (13)
As the population ages, and the prevalence of chronic diseases continue to increase, demand on healthcare providers will grow. As shown in Figure 2, even within the next 10 years, demand for clinicians is set to outstrip supply. As well as improving patient outcomes, AI could provide a way of increasing efficiency. It may take over “low-hanging”, repetitive chores, freeing up doctors to spend more time on more challenging tasks. For example, machine vision and deep learning have been used in radiology to analyse images, passing on the most pressing images to the radiologist, while writing reports for unremarkable images itself (14). An example from the past is the implementation of automated blood pressure measurement and automated blood cell counts, which have not lead to deskilling or redundancy, rather clinicians now work on more interesting and complex tasks, with more time to interact with patients.
Haematologists will have more time to train for and carry out more complex procedures, such as bone marrow aspiration and intrathecal drug administration, which are unlikely to be automated. In addition, haematologists may be able to spend more time working on complementary interests, such as teaching or research. AI may aid clinicians in carrying out administrative tasks, currently loathed by many doctors. Systems using speech recognition and natural language processing may be able to listen in to consultations and write automated reports, leading to more comprehensive patient records (15). Developments like these are in their infancy, but show promise in reducing the clerical workload for all healthcare professionals.
In developing countries, healthcare providers with limited resources face the challenge of providing good quality healthcare to a growing population. AI could help meet this demand. For example, researchers have developed a mobile application that uses a smartphone camera to screen for anaemia. The app captures images non-invasively, through the patient’s skin, and then analyses them using a support vector regression model (16). Another example is the recently launched “robo-hematology” service, which uses a natural dialogue system that can explain the results of simple investigations to patients and can answer their questions (17). This type of service may not provide the same level of patient care as a healthcare professional, but in a resource-poor setting, it may prove a valuable addition to the multidisciplinary team.
There are several challenges that need to be addressed before machine learning decision support systems can be widely implemented. Machine learning models are only as accurate as the data they are trained on. There is always a risk of gaps and inaccuracies in data, especially in historic data derived before widespread adoption of electronic records. Furthermore, an estimated 80% of all medical data is unstructured (18).These records present a challenge to interpret, and conversion into machine useable format introduces another potential source of error. Despite attempts to standardise record keeping, such as ICD-10, medical records are still prone to ambiguity. This challenge is exemplified by IBM Watson, which has suffered from failures in oncology diagnosis, chiefly due to a lack of quality patient records to train on. However, this challenge will be overcome in time, as healthcare data is now being produced at a rapid rate; by 2020, estimates suggest that 94% of all medical data available will have been created after 2013 (18).
Another challenge is the difficulty of understanding how deep learning algorithms come to their decision. The multi-layered neural network is often viewed as a “black box”; there is often little understanding of how the algorithm was derived. The EU’s new General Data Protection Regulation (which will apply from May 2018) states that “Organizations that use ML to make user-impacting decisions must be able to fully explain the data and algorithms that resulted in a particular decision” (19). This is currently not possible in many machine learning systems. As well as the regulatory challenge, this poses an issue of trust; clinicians and patients may not have faith in recommendations from an opaque system which they cannot scrutinise. Furthermore, there is an issue of accountability. In the future, AI may play a part in a poor clinical decision, leading to harm to a patient. In such a circumstance, who is held accountable? The provider producing the model, the provider of the training data set, and the clinician checking the recommendations could all be considered responsible.
Patients with genetic diseases, such as sickle cell disease, haemophilia, and thalassaemia, often require lifelong care, and so can form a strong doctor-patient relationship with their haematologist. This relationship and support is hugely important for many patients, often just as much as the actual medical attention they receive. Some would argue that care dominated by automated systems would cause patients to feel like they are receiving cold, impersonal care. While using AI powered decision support systems, there is a risk of the patient being reduced to their medical history, instead of a unique human being with different preferences, thoughts, and psychosocial background. It is essential that even while using these tools, doctors continue to practise in a way which views patients holistically, treating them as individuals. This consideration means that doctors are unlikely to be entirely replaced, regardless of how medical artificial intelligence advances.
Doctors have a duty to provide the best care to patients. As such, the profession has a responsibility to ensure advancements like AI are used to their full potential, regardless of concerns of employment and clinical autonomy. History provides a positive outlook for haematologists and doctors in general; professions are not often made redundant by technology, instead they undergo redefinition to more human facing roles (20). I am excited to practise a new kind of medicine in the future, focussing on prevention and personalisation, leading to better care and healthier patients.
Word count: 2499
- Zini G. Artificial intelligence in Hematology. Hematology. 2005;10(5):393-400.
- Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S et al. Artificial intelligence in healthcare: past, present and future. Stroke and Vascular Neurology. 2017;2(4):230-243
- Deep Neural Network [Internet]. 2017 [cited 30 December 2017]. Available from: https://www.xenonstack.com/blog/log-analytics-with-deep-learning-and-machine-learning
- Casadio R, Compiani M, Fariselli P, Jacoboni I, Martelli P. Neural Networks Predict Protein Folding and Structure: Artificial Intelligence Faces Biomolecular Complexity. SAR and QSAR in Environmental Research. 2000;11(2):149-182
- Kantardzic M, Djulbegovic B, Hamdan H. A data-mining approach to improving Polycythemia Vera diagnosis. Computers & Industrial Engineering. 2002;43(4):765-773.
- HLS13-04 High-Content High-Throughput Screen for Diamond Blackfan Anemia Treatments | SBIR.gov [Internet]. Sbir.gov. 2017 [cited 28 December 2017]. Available from: https://www.sbir.gov/sbirsearch/detail/1030847
- Xu M, Papageorgiou D, Abidi S, Dao M, Zhao H, Karniadakis G. A deep convolutional neural network for classification of red blood cells in sickle cell anemia. PLOS Computational Biology. 2017;13(10):e1005746.
- Codella N, Moradi M, Matasar M, Sveda-Mahmood T, Smith J. Lymphoma diagnosis in histopathology using a multi-stage visual learning approach. Medical Imaging 2016: Digital Pathology. 2016;.
- Otake T. IBM big data used for rapid diagnosis of rare leukemia case in Japan | The Japan Times [Internet]. The Japan Times. 2017 [cited 29 December 2017]. Available from: https://www.japantimes.co.jp/news/2016/08/11/national/science-health/ibm-big-data-used-for-rapid-diagnosis-of-rare-leukemia-case-in-japan/
- Barbieri C, Molina M, Ponce P, Tothova M, Cattinelli I, Ion Titapiccolo J et al. An international observational study suggests that artificial intelligence for clinical decision support optimizes anemia management in hemodialysis patients. Kidney International. 2016;90(2):422-429.
- Pan L, Liu G, Lin F, Zhong S, Xia H, Sun X et al. Machine learning applications for prediction of relapse in childhood acute lymphoblastic leukemia. Scientific Reports. 2017;7(1).
- Shouval R, Labopin M, Unger R, Giebel S, Ciceri F, Schmid C et al. Prediction of Hematopoietic Stem Cell Transplantation Related Mortality- Lessons Learned from the In-Silico Approach: A European Society for Blood and Marrow Transplantation Acute Leukemia Working Party Data Mining Study. PLOS ONE. 2016;11(3):e0150637.
- Collier M, Fu R, Yin L. Artificial Intelligence in Healthcare [Internet]. Accenture.com. 2017 [cited 29 December 2017]. Available from: https://www.accenture.com/us-en/insight-artificial-intelligence-healthcare
- Aidoc Receives CE Mark for First Commercial Deep Learning Solution Streamlining Head and Neck Imaging for Radiologists [Internet]. Prnewswire.com. 2017 [cited 29 December 2017]. Available from: https://www.prnewswire.com/news-releases/aidoc-receives-ce-mark-for-first-commercial-deep-learning-solution-streamlining-head-and-neck-imaging-for-radiologists-660261973.html
- Halpern Y, Horng S, Choi Y, Sontag D. Electronic medical record phenotyping using the anchor and learn framework. Journal of the American Medical Informatics Association. 2016;23(4):731-740.
- Wang E, Li W, Hawkins D, Gernsheimer T, Norby-Slycord C, Patel S. HemaApp. GetMobile: Mobile Computing and Communications. 2017;21(2):26-30.
- ai [Internet]. Doc.ai. 2017 [cited 30 December 2017]. Available from: https://doc.ai/#products
- Patel A. Macroeconomic trends in the healthcare sector [Internet]. Merciatech.co.uk. 2017 [cited 30 December 2017]. Available from: https://www.merciatech.co.uk/news-media/insights/2017/dec/20/macroeconomic-trends-healthcare-sector/
- The European Parliament and the Council of the European Union. General Data Protection Regulation. Official Journal of the European Union; 2016.
- Autor D. Why Are There Still So Many Jobs? The History and Future of Workplace Automation. Journal of Economic Perspectives. 2015;29(3):3-30.
In 2016 the essay title was 'Discuss the Impact of Obesity on the Incidence and Management of Haematological Disorders'. The winner was Prateek Yadav from UCL Medical School, and the runner-up was Maria Fala from the University of Cambridge.
The obesity epidemic is one of the biggest healthcare challenges of our lifetimes. The WHO reports that obesity has more than doubled worldwide since 1980, By 2014 1.9 billion adults were overweight with 600 million of these classified as obese. This accounts for 39% of adults over 18 years of age being overweight worldwide. When looking at child and adolescent populations, it is estimated that 41 million children under 5 years of age are overweight worldwide. Most notably, developing countries and recently developed countries account for a large proportion of these; in 2014 almost half of overweight children under 5 lived in Asia. In addition, the number of overweight children under 5 years of age doubled in the African continent from 1980 to 20141. Both the rate of rapid development across the world and the increasing proportion of children who are obese indicate that this is not a problem that is going to go away; if anything, the pressures of obesity on our healthcare systems will only increase in the coming years. Furthermore, haematologists must be highly aware of how obesity and the associated metabolic syndromes interact with haematological conditions to provide high quality personalized care for patients. This essay looks at the interplay between obesity and haematological conditions and the impact on patients and our healthcare systems.
What drives the pro-coagulable state in obese patients?
Obesity is associated with hypertension, hyperlipidaemia and atherosclerosis, which are all risk factors for clotting events and obese patients are more likely to have difficulties with mobility, causing venous stasis. There are also a multitude of co-existing factors that create the increased clotting risk in the obese, and a plethora of hormonal disturbances and inflammatory reactions putting patients at a higher risk. Adipose tissue is an active endocrine organ, releasing cytokines such as leptin, adiponectin, interleukin-6 and tumour necrosis factor-α (TNF-α) that have subsequent effects on insulin resistance, haemostasis and inflammation. In terms of clotting components, obese patients tend to have higher circulating levels of fibrinogen, von Willebrand factor and plasminogen activator inhibition as well as factors VII and VIII2. These increased levels are likely to be produced by alterations in hepatic synthesis via pro-inflammatory cytokines. A single standard deviation increase in fibrinogen, in both men and women, was associated with a 20% increased incidence of primary cardiovascular events when adjusted for age and other risk factors2.Atherosclerotic disease is also promoted by increased expression of intercellular adhesion molecules such as ICAM -1 which are at higher levels in obese patients3.
Higher than normal levels of plasminogen activator inhibitor-1 (PAI-1) is regarded as part of the metabolic syndrome of insulin resistance and obesity4. High levels of PAI-1 is a risk factor for atherosclerosis and thrombosis, as it is the principle inhibitor of tissue plasminogen activator which in turn has a key role in fibrinolysis. It has been shown that adipocytes can produce PAI-1, which is a potential mechanism for increased coagulability in obesity and indeed waist circumference and PAI-1 levels correlate strongly. Weight loss in turn reduces PAI-1 levels but does not necessarily reduce fibrinogen levels, providing further evidence that PAI-1 is a key player in the higher clotting risk in the obese .
The inflammatory markers that are raised in obesity and insulin resistance, including TNF-α and transforming growth factor-β, have been shown to promote PAI-1 production5. This can contribute to the already significant production of PAI-1 by adipose tissue. These effects may be further compounded by disturbances in glucocorticoid and insulin levels in these patients, although the causal direction is unclear; there is some evidence to suggest that PAI-1 drives the development of insulin resistance rather than the other way around.
The leptin produced by adipose tissue increases thrombogenicity via a variety of potential mechanisms, including increasing platelet aggregation and von Willebrand factor levels3. Leptin also enhances calcification of the vasculature, providing targets for thrombogenic processes. The concentration of pro-inflammatory cytokines increases with adiposity; obesity is accompanied by a chronic inflammatory process. IL- 6 can be three times higher in obese patients than slim patients and it is thought to contribute to insulin resistance syndrome. The commonly measured inflammatory marker C reactive protein (CRP) has a strong association with obesity even accounting for age, race and smoking status differences6. CRP is independently associated with atherothrombotic risk and is an important direct link between obesity, inflammation and thrombotic risk. Raised CRP induces platelet adhesion to endothelial cells and interestingly, activated platelets convert pentameric CRP to its monomeric form, where it is more likely to capture neutrophils. This evidently makes up part of the atherosclerotic chain of events. Raised CRP levels also inhibit release of tissue-type plasminogen activator, which is important to produce the fibrinolytic substance plasmin, and upregulates release of PAI-1 from endothelial cells. Furthermore, in vitro experiments have shown CRP stimulates blood monocytes to produce tissue factor, which is a key part of the clotting cascade7.
Management of Haemostasis in Obese Patients
Evidence suggests that normal doses of thromboprophylaxis are suboptimal in morbidly obese patients and a trend towards higher doses of thromboprophylaxis may be seen. This is because, despite clinicians currently prescribing weight-adjusted doses for VTE treatment, the doses for thromboprophylaxis are often fixed dosage regimens. This regardless of evidence of a dose-response relationship for VTE prevention and is likely to be due to fears of bleeding risk. It has been shown that high dose thromboprophylaxis with heparin (7500 units TDS) or enoxaparin (40mg BD) nearly halved the incidence of VTE in the morbidly obese when compared to standard doses without increasing bleeding risk8. It is important to note that while this study has a very large sample size of 9241 and a good mix of patient settings, it has borderline statistical significance (p=0.050).
The tendency towards higher doses is compounded by the fact that stockings and pneumatic compression has reduced efficacy in the severely obese. More insight and clarity is needed into the specifics of dosage regimens in the obese for prevention of VTE as there are a large variety of dosages tested in studies. There are other practical considerations to consider, for example enoxaparin is often manufactured in prefilled 40mg syringes which might make it difficult for early adopters of a 0.5mg/kg regimen or similar. It remains to be seen whether dosage regimens for the new direct oral anticoagulant drugs (DOACs) will need to be optimized for patient weight.
Knowledge of the pathways and mechanisms influencing the high risk of clotting events may open more obscure avenues to optimize patient care for prevention. For example, it is useful to note that decreases in PAI-1 activity are noted with ACE-inhibitors and angiotensin receptor blockers, along with diabetes drugs such as metformin and thiaizolidinediones9. This could potentially allow doctors to use synergistic drug combinations in comorbid patients in the future. Many patients at thrombotic risk are taking statins which have been shown to reduce CRP levels7 and in the future anti-inflammatory medications could potentially mitigate the inflammation mediated thrombotic risk in obese patients.
Obesity and Transfusion Medicine
A subject of much debate is whether being obese leads to increased numbers of surgical complications and therefore increased requirement for transfusions, or whether higher BMI and circulating volume protects against blood loss. The ‘obesity paradox’ of outcomes is the idea that overweight or obese patients may have better outcomes in certain situations than those who are of normal weight. The relationship between obesity and the need for perioperative transfusions is not clearly identified and much of the evidence is from low powered studies with no uniform definition of obesity across studies. In one recent study of elective procedures, namely total hip and total knee arthroplasties, patients who had an elevated BMI (>30) had decreased rates of blood transfusion (p=0.001). Patients with elevated BMI also lost significantly smaller percentages of blood volume perioperatively10. While this example is apt, as arthroplasties are more common in the obese population due to increased risk of osteoarthritis, this is not a generalizable result.
The relationship between obesity and transfusion needs in trauma has not been fully established. It is possible to speculate that an obese patient might lose a lower percentage of their blood volume than a non-obese patient in comparable situations. However, patients who are obese are at higher initial risk of mortality after trauma mostly due to haemorrhagic and hypovolaemic shock11. Risk in obese patients is compounded by the difficulty doing other procedures during resuscitation such as intubation and insertion of central venous catheters. A study looking at the frequency of massive transfusion in obese patients, with massive transfusion (MT) defined as 10 units of packed red cells in the first 24 hours, found that obese patients were more likely to need massive transfusion than non-obese patients with an odds ratio of 1.68 (95%CI 0.97 – 2.72), p = 0.07. The p value is relatively large, and the 95% confidence intervals overlap; obese patients had 15%(95%CI: 9-23%) receiving MT, non-obese 10% (95%CI: 8-12%) . Despite the drawbacks to this study it might point to a new direction in transfusion protocol and while scores such as the Trauma Associated Severe Haemorrhage score (TASH) score do not mention BMI or weight, this might be a useful parameter in the future to assess transfusion demands in trauma patients.
Obesity and Benign Haematology
Despite associations with thin children with poor growth, conditions like sickle cell disease and thalassaemia now have drastically improved lifespans and prognoses and to see an obese patient with these conditions is not uncommon. Obesity complicates care as systemic changes in obesity add to the risk already present in these patients. The spectrum of complications that sickle cell anaemia can cause overlap significantly with risk factors obesity brings. Obesity and sickle cell both contribute to risk of stroke, cardiovascular events and vascular disease. In addition, obesity related hypertension combined with sickle cell anaemia can contribute to retinopathy and end-organ damage. It follows that these patients need close monitoring and careful management, and sickle cell patients of normal BMI do have lower admission rates than those with very high or low BMI12. However, it was found that BMI has an overall inverse relationship with number of admissions. More research is needed in this area to see how metabolic changes in obesity will affect patients with benign haematological conditions.
Obesity and haematological malignancies
Obesity increases the risk of the haematological malignancies, including leukaemias, lymphomas and myeloma13. If the obesity epidemic is not tackled successfully, demands on haemato-oncology services are going to increase in coming years. The distribution of risk is a dose-response relationship; increasing weight leads to increasing risk and vice versa. The meta-analysis of cohort studies found significant association between obesity and risk of developing non-Hodgkin’s lymphoma and myeloma. The evidence for increased risk of the leukaemia generally was significant in 11 out of 15 studies, however stratification into different subtypes was not carried out in most of the studies so specific information on these was not reported13. However, a strong association between obesity and promyelocytic leukaemia was found.
The mechanism by which obesity increases cancer risk is not fully clear but two major theories are called the ‘inductive’ and ‘selective’ hypotheses. The former relies on inflammatory and metabolic changes in obesity encouraging neoplastic changes, while the latter hypothesizes that the environment obesity produces selects in favour of already present abnormal cells that are dormant. Leptin produced by adipocytes induces cancer progression via the MAPK, PI3K and STAT pathways. In a similar way to the disturbances in haemostasis in obesity, proinflammatory cytokines including TNF-α, IL-2, IL-8 and IL-10 mediate increased cancer risk. Insulin and IGF-1 levels in obese patients is also implicated in cancer risk and mortality, thought to be acting via the AKT/PI3K/mTOR cascade. The fact that metformin reduces cancer incidence and mortality, but drugs that stimulate insulin secretion increase incidence and mortality, emphasizes the importance of insulin and IGF-1 in cancer biology14.
The presence of a malignancy drastically increases risk of clotting events and may produce a perfect storm in obese patients who are already high risk and much of the mortality and morbidity is accounted for by hypercoagulation15. These patients need close attention to prevent VTE, cardiopulmonary events and stroke. There are further complications in that conditions such as type 2 diabetes, with high prevalence in the obese population, can modulate toxicity of cancer therapy regimens. The presence of diabetes significantly increases the risk of cardiotoxicity, from anthracycline treatment in patients with lymphoma, breast cancer and gastric cancer16. Chemotherapy regimens may have to be adjusted to prevent adverse effects that are exacerbated by comorbid conditions, especially as cancer prognoses improve.
Obesity and Healthcare Funding
It is important to look at the way funding structures across healthcare systems change as the burden of obesity-related problems increases. In the UK, the direct cost to the NHS was estimated at £4.2 billion in 2007, with indirect costs estimates ranging to up to £15.8 billion17. When compared to the estimate of cost in 1998, at 469.9 million, it puts into perspective the speed and magnitude of the growth in spending. Future projections of total indirect costs reach as high as £15.8 billion. This will put pressure on all branches of medicine, but haematology is a speciality that could be particularly affected due to the high costs of chemotherapy drugs for haematological malignancies. An instructive example of this is the drug ibrutinib for chronic lymphocytic leukaemia, which was not approved by NICE due to its high cost (£55.000 per patient) until a recent deal where prices were lowered for the NHS18. As the obesity burden increases, healthcare providers may be forced to make difficult decisions about funding treatments and more of the financial burden of these expensive treatments may have to be shouldered by organisations such as the Cancer Drugs Fund. This will have a major impact on the range treatments available to haematologists and will affect how funding decisions are made in the future.
Obesity brings with it increased risk of VTE, cardiovascular and cerebrovascular events, and increasing obesity rates may change the landscape of thromboprophylaxis of the future. It will also be a strong driving force for innovations in this area, as seen recently with the development of the DOAC drugs. More investigation is needed into the interaction between obesity and benign haematological conditions, and patient management may have to change to account for greater risk. Obesity also increases the incidence of almost all the haematological malignancies and the associated metabolic syndromes may complicate management of cancer. Mechanistically, hormonal imbalances and inflammatory processes influence many of these risk increases and are still fertile areas for more research. The severe financial pressures obesity puts on healthcare services will force providers to make difficult decisions about resource allocation in the future, and may reduce the range of treatments available in the NHS to haematologists in the future.
Word count : 2489
 World Health Organization, “Media centre Obesity and overweight,” January, pp. 1–5, 2015.
 I. Mertens and L. F. Van Gaal, “Obesity, haemostasis and the fibrinolytic system,” Obes. Rev., vol. 3, no. 2, pp. 85–101, 2002.
 L. F. Van Gaal, “Mechanisms linking obesity with cardiovascular disease,” Diabetes, Obes. Metab., vol. 12, no. December, p. 21, 2010.
 I. Juhan-Vague, M. C. Alessi, and P. E. Morange, “Hypofibrinolysis and increased PAI-1 are linked to atherothrombosis via insulin resistance and obesity.,” Ann. Med., vol. 32 Suppl 1, pp. 78–84, Dec. 2000.
 S. J. Appel, J. S. Harrell, and M. L. Davenport, “Central obesity, the metabolic syndrome, and plasminogen activator inhibitor-1 in young adults,” J Am Acad Nurse Pr., vol. 17, no. 12, pp. 535–541, 2005.
 M. Visser, L. M. Bouter, G. M. McQuillan, M. H. Wener, and T. B. Harris, “Elevated C-reactive protein levels in overweight and obese adults.,” JAMA, vol. 282, no. 22, pp. 2131–5, Dec. 1999.
 W. P. Fay, “Linking inflammation and thrombosis: Role of C-reactive protein.,” World J. Cardiol., vol. 2, no. 11, pp. 365–9, Nov. 2010.
 T.-F. Wang, P. E. Milligan, C. A. Wong, E. N. Deal, M. S. Thoelke, and B. F. Gage, “Efficacy and safety of high-dose thromboprophylaxis in morbidly obese inpatients.,” Thromb. Haemost., vol. 111, no. 1, pp. 88–93, Jan. 2014.
 M. Gnacińska, S. Małgorzewicz, M. Stojek, W. Lysiak-Szydłowska, and K. Sworczak, “Role of adipokines in complications related to obesity. A review.,” Adv. Med. Sci., vol. 54, no. 2, pp. 1–8, 2009.
 N. Frisch, N. M. Wessell, M. Charters, E. Peterson, B. Cann, A. Greenstein, and C. D. Silverton, “Effect of Body Mass Index on Blood Transfusion in Total Hip and Knee Arthroplasty.,” Orthopedics, vol. 39, no. 5, pp. e844-9, 2016.
 A. De Jong, P. Deras, O. Martinez, P. Latry, S. Jaber, X. Capdevila, and J. Charbit, “Relationship between obesity and massive transfusion needs in trauma patients, and validation of TASH score in obese population: A retrospective study on 910 trauma patients,” PLoS One, vol. 11, no. 3, pp. 1–15, 2016.
 M. W. Farooqui, N. Hussain, J. Malik, Y. Rashid, M. Ghouse, and J. Hamdan, “Prevalence of Obesity in Sickle Cell Patients,” Blood, vol. 124, no. 21, 2014.
 M. A. Lichtman, “Obesity and the Risk for a Hematological Malignancy: Leukemia, Lymphoma, or Myeloma,” Oncologist, vol. 15, no. 10, pp. 1083–1101, 2010.
 I. Vucenik and J. P. Stains, “Obesity and cancer risk: Evidence, mechanisms, and recommendations,” Ann. N. Y. Acad. Sci., vol. 1271, no. 1, pp. 37–43, 2012.
 G. J. Caine, P. S. Stonelake, G. Y. H. Lip, and S. T. Kehoe, “The hypercoagulable state of malignancy: pathogenesis and current debate.,” Neoplasia, vol. 4, no. 6, pp. 465–73, 2002.
 A. Gomes, L. Lopes, A. Ferreira, M. Correia, H. Mansinho, and H. Pereira, “The effect of cardiovascular risk factors and cancer type in anthracycline’s cardiotoxicity,” Eur. Hear. J. ( 2016 ) 37 ( Abstr. Suppl. ), 572, vol. 17 (Supple, p. ii269, 2016.
 L. Morgan and D. Monica, “The economic burden of obesity,” Natl. Obes. Obs., no. October, pp. 1–13, 2010.
 “NICE reverses decision on CDF leukaemia drug after price drop | News and features | News | NICE.”
Obesity is currently one of the main challenges experienced by medicine. The obesity epidemic is constantly spreading, with the worldwide rates more than doubling in the last thirty-five years2. Its consequences are numerous and diverse. Obesity has an impact on all systems of the body, including haematology. It is stratified in terms of the Body Mass Index (BMI), where a value of 20-25 kg/m2 is generally considered to be normal weight, 25-30 kg/m2 is overweight, 30-35 kg/m2 is obese and anything above 35 kg/m2 is superobese2. Haematology is affected by obesity-driven systemic and local changes in the Bone Marrow, many of which are still unknown. This essay will explore these changes, their effect on haematological disorders and their management.
Haematological cancers form a large part of the haematological disorders in terms of their prevalence in the human population as well as their morbidity and mortality. Obesity has been found to increase the incidence and worsen the prognosis of many different types of cancers. It is, therefore, not surprising that haematological cancers are also affected by obesity.
Adipocytes constitute the most abundant cell type in bone marrow6. They play an important role in the tissue microenvironment within the bone marrow. In obese patients, the number of adipocytes in the bone marrow is increased and their cytokine and lipid profile changes, affecting the neighbouring cells. For example, studies on obese mice have demonstrated that adipocytes release more leptin and inflammatory markers and less adiponectin and anti-inflammatory proteins6. These chemicals activate various signalling cascades and lead to increased genomic instability, impaired DNA repair, tumour progression, local immunosuppression and epigenetic changes, all of which are detrimental for the patient6. For this reason, adipocytes now constitute a potential therapeutic target for several cancers.
A cancer that has been found to be affected by obesity is Multiple Myeloma(MM). This is a plasma cell neoplastic tumour which constitutes 10% of haematological cancers6. It is characterised by a clonal expansion of abnormal plasma cells in the bone marrow, osteolytic bone disease, anaemia and renal failure5. It is currently incurable. Risk factors include older age, African ethnicity, family history and Monoclonal Gammopathy of Undetermined Significance (MGUS)6. Obesity has recently been identified as another important risk factor for multiple myeloma, causing an increase in incidence and mortality6. A recent paper suggests that this increased risk is an indirect effect because obesity is, in fact, a risk factor for MGUS, which was found to be twice as common among obese patients compared to non-obese patients3. MGUS is the stage which commonly precedes MM, in which an increase in the secretion of monoclonal immunoglobulin is observed without any accompanying myeloma features such as osteolytic changes5.
One of the mechanisms by which obesity is linked to increased risk of cancer is the communication between adipocytes and cancer cells, aiding tumour initiation, growth and metastasis6. In a murine study, the role of adipocytes and the bone marrow microenvironment in tumour growth and progression has been demonstrated. 5T myeloma cells were inoculated in mice whose diet was varied. In mice on a high-fat diet, an increased amount of myeloma-specific IgG2bk para-protein was detected, before myeloma inoculation, suggesting that obese mice were in an obesity-driven MGUS state before tumour inoculation. They developed myeloma after the tumour was inoculated and the tumour burden was reduced once the high fat diet was removed. Collectively, these results suggested that the obese host provided a myeloma-permissive microenvironment for the myeloma cells to grow. Also, obesity did not have a direct effect on tumour growth and survival, but had an important role in enabling the initiation of the tumour5. A possible driver for this change in the microenvironment is Insulin-like Growth Factor 1 (IGF-1) which was raised in obese mice, even before inoculation of the myeloma cells. Therefore, IGF-1 acts as a potent myeloma growth factor. Interleukin-6, on the other hand, commonly raised in cancer, only appeared to be raised in myeloma-bearing mice, suggesting that it is released by myeloma cells6.
Conversely, some study results suggest a role of the altered microenvironment in tumour growth and progression. One of the causal mechanisms for this relates to the increased amount of Interleukin-6 (IL-6) in the bone marrow microenvironment. IL-6 leads to activation of the STAT-3 signalling pathway, leading to increased proliferation and reduced apoptosis in monoclonal plasma cells6. Additionally, tumour growth and progression is aided by the obesity-driven changes enabling better cell adhesion and angiogenesis. Better adhesion is correlated with the alpha-4 integrin, whose expression has been shown to increase in obesity6. This offers increased proliferation, survival, migration and drug resistance to the tumour cells, which is why alpha-4 integrin blockers can restore sensitivity of the myeloma cells to bortezomib6. Angiogenesis is enhanced in the obese state via two- to three-fold overexpression of Matrix Metalloproteinase 2 (MMP-2) in obese compared to normal6.
Another haematological cancer affected by obesity is Acute Myeloid Leukaemia (AML). This is the most common form of acute leukaemia in adults and is frequently fatal14. Obesity is known to increase both incidence and mortality rates but research is still undergoing to investigate the pathogenesis mechanism. In a study undertaken using murine models, the leukaemia burden was shown to be higher in diet-induced obese mice14. This was associated with higher levels of Fatty Acid Binding Protein 4 (FABP-4) and Interleukin-6 (IL-6) in serum. The Fatty Acid Binding Proteins are cytosolic intracellular receptors that can bind hydrophobic ligands and mediate lipid trafficking in the cell. FABP-4 is a good marker of obesity as it is highly expressed in adipocytes and macrophages of the obese. Upregulation of FABP-4 has been shown to promote tumour growth in AML and this study proposes a mechanism through which this occurs. They have shown that FABP-4 upregulation is correlated with increased amounts of IL-6, which leads to activation of the STAT-3 transcription factor. This transcription factor promotes DNA Methyl Trasferase-1 (DNMT-1) expression, overexpression of which leads to DNA hyper-methylation and silencing of Tumour Suppressor Genes such as p53. Aberrant DNA methylation has already been shown to be involved in AML, which makes the proposed mechanism highly likely14.
Non-Hodgkin’s lymphoma is a group of malignant diseases which originate from lymphocytes. The effect of obesity on the risk of non-Hodgkin’s lymphoma was investigated in a meta-analysis study which used the random effects model4. This showed that obesity is associated with a statistically significant increase in the risk of non-Hodgkin’s lymphoma and in particular in the risk of diffuse large B-cell lymphoma. Overweight people were found to have a 7% higher risk and obese patients were found to have a 20% higher risk of developing the disease4.The mechanism through which obesity increases the risk of this cancer is still unclear but several hypotheses exist. Altered immune function and chronic inflammation are risk factors for non-Hodgkin’s lymphoma, therefore this might be the mechanism via which obesity increases the risk of lymphoma. For example, the obesity-driven reduction in adiponectin can lead to such immune changes since adiponectin normally has anti-inflammatory effects and reduces lymphocyte proliferation. In contrast, leptin, whose levels are increased in obesity, enhances monocyte proliferation and hence the release of inflammatory cytokines such as Tumour Necrosis Factor-alpha (TNF-α). Another possible mechanism is related to the resistance to insulin which develops in obesity. As a result, the body produces more insulin which leads to a compensatory hyperinsulinaemia. This leads to elevated levels of IGF-1 which leads to increased cell proliferation and reduced apoptosis. This could also explain why Type II diabetes is also associated with an elevated risk for non-Hodgkin’s lymphoma4.
The coagulation and fibrinolytic cascades are very important cascades in haematology, as they determine a major property of the circulating blood, which is its tendency to clot. Obesity has been associated with higher levels of coagulation factors VII, VIII, IX, XII as well as the Von Willebrand Factor (VWF), tissue factor, fibrinogen and plasminogen activator inhibitor-1 and reduced levels of tissue plasminogen activator activity and activated protein C (APC) ratio1,7. As a result of this imbalance, obesity leads to a hypercoagulable state.
These coagulation and fibrinolytic factors are affected by obesity via different mechanisms. A study investigating the effects of obesity on the expression of tissue factor, found that compensatory hyperinsulinaemia in obesity might play a role in inducing tissue factor overexpression7. High insulin levels also lead to increased expression of Plasminogen Activator Inhibitor-1 (PAI-1), which also contributes to the hypercoagulable state. Thirdly, hyperinsulinaemia could also be the cause of the reduced Tissue Factor Pathway Inhibitor (TFPI) levels, leading to reduced inhibition of tissue factor activity in the clotting cascade7.
Based on the hypercoagulable state, a study has been undertaken to investigate the risk of bleeding in obese patients compared to normal weight patients, hypothesising that the hypercoagulable state would lead to a protection against bleeding in obese. However, even though the number of bleeding events was reduced in normal weight participants compared to underweight, the risk of bleeding in obese was not significantly different to the risk in normal weight participants1.
The hypercoagulable state in obesity is very important because it is a risk factor for thromboembolism. A study investigating the combined effect of obesity with other risk factors on the risk of thromboembolism has identified a number of interactions between risk factors15. People with genetic mutations that predispose them to thromboembolic events, have a higher combined risk in obesity compared to the sum of the two individual risks. Also, obesity in combination with oral steroids increases the risk of venous thromboembolism. This is because obesity produces a pro-thrombotic state while oestrogen increases the resistance to activated protein C and further increases the concentration of factors II, VII, VIII and X, thus enhancing the effects of obesity.
Haemophilia A is a hereditary X chromosomal recessive haematological disorder which is related to the coagulation cascade10. It is caused by a deficiency or a functional defect in clotting factor VIII, leading to impaired coagulation of blood and bleeding episodes, causing various problems such as limitations in joint range of motion due to haemorrhage in the joint. These episodes are prevented and controlled by intravenous infusion of Clotting Factor Concentrate (CFC). Several studies have been conducted to investigate the effect of the obesity-driven hypercoagulable state on the presentation of Haemophilia A. An increased net CFC usage was observed in obese patients compared to normal BMI patients, but this difference disappeared when adjusting the CFC usage for the patients’ weight10. No difference was seen in FVIII activity between obese and non-obese patients, which is not surprising as even though obesity leads to overexpression of factor VIII, in Haemophilia patients this would not be possible, or the protein would be non-functional. Plasminogen Activator Inhibitor–1 (PAI-1) has been shown to be increased in in obese patients but led to no reduction in the need for coagulation10. Additionally, for reasons that continue to be unclear, obesity has been found to be associated with further limitations in joint range of motion and further joint mobility loss compared to non-obese patients. More research needs to be undertaken in the effects of obesity on haemophilia, for correct dosing of Clotting Factor Concentrate, according to the patient’s weight, to be determined.
The coagulation imbalance is also important to consider when treating patients with anticoagulants. These are needed when thrombus formation is likely and needs to be prevented or treated. However, clinical trials for most of these drugs excluded obese participants. There are many uncertainties regarding dosing in obese patients due to lack of evidence. This is a serious problem, because obesity increases the risk of venous thromboembolism and the duration of inpatient stay. Studies have been undertaken to determine whether the current drug dosing recommendations are appropriate for obese patients6. For venous thromboembolism prophylaxis, fixed doses of enoxaparin, dalteparin or tinzaparin are currently prescribed in the UK. However, this study provides evidence that enoxaparin and tinzaparin dose adjusted for BMI is more effective than the fixed dose6. The pharmacokinetics of each drug depend on its volume of distribution and its clearance, hence the effective dose and its dependence on the patient’s BMI is drug-specific. Therefore, mathematical models need to be applied to predict each drug’s behaviour.
Another systemic effect of obesity discovered, is the premature thymic involution that it causes. As a result, T cell generation is compromised and T cell apoptosis is increased leading to reduced thymocyte counts16. Consequently, obesity leads to reduced peripheral naïve T cells, causing impaired immunity and increased susceptibility to persistent infections.
Lastly, obesity can affect other haematological disorders such as amyloidosis, a disease involving extracellular deposition of protein fibrils. In amyloidosis AA, the protein fibrils are derived from SAA, an acute phase protein produced by hepatocytes under inflammatory conditions. A study on the effect of obesity on amyloidosis revealed a positive correlation between body weight and degree of amyloidosis, suggesting that obesity can cause enhanced chronic secretion of SAA11.
Surgical interventions to treat morbid obesity can also affect haematology. Case reports have suggested that gastric bypass leads to microcytic anaemia due to reduced iron absorption. This is because iron absorption normally occurs in the bypassed duodenum and because iron chelation, that normally occurs when iron moves from the acidic environment to the alkaline duodenum and precipitates as a hydroxide, is impaired13. Gastric bypass surgery also leads to reduced vitamin B12 levels, due to the loss of parietal cells which normally release intrinsic factor enabling vitamin B12 absorption in the distal ileum. Another case report presents the effects of total gastrectomy on haematology8. In this case, the vitamin B12 deficiency led to high homocysteine levels, a risk factor for thrombophilia. Therefore, it might be useful to provide prophylaxis against thrombosis in patients undergoing such interventions.
Unfortunately, haematological conditions themselves are also related with obesity. Many debilitating diseases can lead to a sedentary lifestyle which can cause weight gain and obesity. It has been shown that patients suffering from childhood acute lymphoblastic leukaemia, gain excessive weight during and after two years of therapy12. Previously, hypothalamic dysfunction due to cranial irradiation in treatment of the leukaemia was suspected to cause the increased rates of obesity in the survivors. However, a study on the causes of weight gain, has suggested that weight gain occurs even in the absence of cranial irradiation12. Reduced total energy expenditure, mainly due to reduced energy expanded on activity is the main cause of obesity, meaning that measures can be taken to reduce obesity in these patients.
To conclude, medicine must face the serious problem of obesity. This is a challenge due to the many diverse effects that obesity has on health but also because of its increasing prevalence. In haematology, obesity constitutes a risk factor and affects progression of many diseases, including cancers, coagulopathies and amyloidosis. Research in this field needs to be continued to find out how to deal with these emerging problems and how current treatments need to be modified to accommodate obese patients. Importantly, preventative measures need to be taken in order to minimise the cases of obesity to avoid its destructive consequences.
- Braekkan, S.K., Graaf, Y., Visseren, F.L.J. and Algra, A., 2016. Obesity and risk of bleeding: the SMART study.Journal of Thrombosis and Haemostasis, 14(1), pp.65-72.
- Bullwinkle, E.M., Parker, M.D., Bonan, N.F., Falkenberg, L.G., Davison, S.P. and DeCicco-Skinner, K.L., 2016. Adipocytes contribute to the growth and progression of multiple myeloma: Unraveling obesity related differences in adipocyte signaling.Cancer Letters, 380(1), pp.114-121.
- Landgren, O., Rajkumar, S.V., Pfeiffer, R.M., Kyle, R.A., Katzmann, J.A., Dispenzieri, A., Cai, Q., Goldin, L.R., Caporaso, N.E., Fraumeni, J.F. and Blot, W.J., 2010. Obesity is associated with an increased risk of monoclonal gammopathy of undetermined significance among black and white women.Blood, 116(7), pp.1056-1059.
- Larsson, S.C. and Wolk, A., 2007. Obesity and risk of non‐Hodgkin's lymphoma: A meta‐International Journal of Cancer, 121(7), pp.1564-1570.
- Lwin, S.T., Olechnowicz, S.W., Fowler, J.A. and Edwards, C.M., 2015. Diet-induced obesity promotes a myeloma-like condition in vivo.Leukemia, 29(2), p.507.
- Patel, J.P., Roberts, L.N. and Arya, R., 2011. Anticoagulating obese patients in the modern era.British journal of haematology, 155(2), pp.137-149.
- Samad, F., Pandey, M. and Loskutoff, D.J., 2001. Regulation of tissue factor gene expression in obesity.Blood, 98(12), pp.3353-3358.
- Shin, S.B., Jang, Y.N., Lee, H.J., Yi, Y.M., Lee, J.W., Min, W.S. and Eom, K.S., 2016. Thrombophilia after total gastrectomy for morbid obesity.The Korean journal of internal medicine.
- Silva, Simone Vargas, et al. "Obesity modifies bone marrow microenvironment and directs bone marrow mesenchymal cells to adipogenesis."Obesity 12 (2016): 2522-2532.
- Tuinenburg, A., Biere‐Rafi, S., Peters, M., Verhamme, P., Peerlinck, K., Kruip, M.J.H.A., Gorkom, B.A.P., Roest, M., Meijers, J.C.M., Kamphuisen, P.W. and Schutgens, R.E.G., 2013. Obesity in haemophilia patients: effect on bleeding frequency, clotting factor concentrate usage, and haemostatic and fibrinolytic parameters.Haemophilia, 19(5), pp.744-752.
- Van der Heijden, R.A., Bijzet, J., Meijers, W.C., Yakala, G.K., Kleemann, R., Nguyen, T.Q., de Boer, R.A., Schalkwijk, C.G., Hazenberg, B.P., Tietge, U.J. and Heeringa, P., 2015. Obesity-induced chronic inflammation in high fat diet challenged C57BL/6J mice is associated with acceleration of age-dependent renal amyloidosis.Scientific reports,
- Ventham, J.C. and Reilly, J.J., 1999. Childhood leukaemia: a model of pre-obesity.Proceedings of the Nutrition Society, 58(02), pp.277-281.
- Yale, C.E., Gohdes, P.N. and Schilling, R.F., 1993. Cobalamin absorption and hematologic status after two types of gastric surgery for obesity.American journal of hematology, 42(1), pp.63-66.
- Yan, F., Shen, N., Pang, J.X., Zhang, Y.W., Rao, E.Y., Bode, A.M., Al-Kali, A., Zhang, D.E., Litzow, M.R., Li, B. and Liu, S.J., 2016. Fatty acid-binding protein FABP4 mechanistically links obesity with aggressive AML by enhancing aberrant DNA methylation in AML cells.
- Yang, G. and Christine De Staercke, W., 2012. The effects of obesity on venous thromboembolism: A review.Open journal of preventive medicine, 2(4), p.499.
- Yang, H., Youm, Y.H., Vandanmagsar, B., Rood, J., Kumar, K.G., Butler, A.A. and Dixit, V.D., 2009. Obesity accelerates thymic aging.Blood, 114(18), pp.3803-3812.
In 2015 the essay title was 'Haematology and the Ageing Population – Implications for the Practising Haematologist'. The winner was Jess Dunphy, and the runner-up was Jemma Proudfoot-Jones, both of the University of Birmingham.
The United Kingdom, like the world as a whole, is home to an ageing population. Currently, there are over 11 million people in Britain aged 65 or over1. The number of people aged 60 or over in Britain is now greater than the population aged 18 and under1. Furthermore, it is predicted that by 2040 almost a quarter of the UK population will be over 65.2 The ageing population is likely to have a large impact on health services, both in terms of increased access to services and of financial cost to the National Health Service (NHS).3 This essay will examine the implications of an ageing population for haematologists. It will briefly discuss the ageing process, before exploring various aspects of haematologists’ work that may be impacted upon. These aspects include a greater need for shared care, considerations about treatment options, increased workload and ethical considerations.
What Happens as People Age?
The ageing process is characterised by “declining functional capacity and increasing vulnerability to disease, disability and death”.4 As people age, they are more likely to have multiple and complex conditions, including dementia and frailty.3 On average, an elderly person has 3 to 4 chronic illnesses.5 In terms of haematological conditions, longevity increases the risk of developing both haematological malignancies (such as myeloma and lymphoma) and non-malignant conditions (such as anaemia and susceptibility to coagulation disorders).6 As a consequence of multiple comorbidities, the elderly are also more likely to be on multiple medications than the younger population.3
The ageing process is driven by a gradual accumulation of both molecular and cellular defects that occur throughout our lifetime.4 One important aspect of ageing is “immunosenence”, the ageing of the immune system. This encompasses two main areas: decreased function of memory T cells and inflammaging.7
1. Memory T cells
As people age, the number of T-cells in the peripheral blood remains constant but there is a shift towards a greater production of memory T cells (rather than naïve T cells). This means that the elderly have a decreased capacity to respond to novel antigens. Furthermore, the memory T cells in the elderly function less well and have a less diverse repertoire than those in young adults, and so are able to respond less well to known antigens.7
Inflammaging refers to clinical syndromes that are associated with ageing and include some form of inflammation (for example, atherosclerosis). It is thought that these age-related syndromes may be related to a loss of tolerance, either to self-antigens or to foreign antigens.7
Immunosenescene is important clinically because it means that elderly individuals are more predisposed to develop certain infections, or to develop more severe symptoms as a result of infections, than younger individuals. In addition, it is thought that the increased malignancy rate in the elderly, including haematological malignancies, may be due to a failure to provide immune surveillance to tumours.7
Another important facet of ageing for haematologists is the decline in function of haematopoietic stem cells (HSCs). Both the ability of HSCs to self-renew and to differentiate into mature blood lineages declines with age. This results from both intrinsic cell damage (such as DNA damage) and from changes to the micro-environment of the stem cell, including the aforementioned inflammatory changes that occur with age.8 Decline in HSC function may contribute to both myelodysplasia and, possibly, anaemia of the elderly.8
A Greater Need for Collaboration with Other Specialities
The first implication of an ageing population for practising haematologists is a greater need for collaborating with other specialities. The key speciality that haematologists need to liaise with to ensure optimum management of their elderly patients is geriatrics. One example of the importance of communicating with geriatricians is during the treatment of haematological cancer. A comprehensive geriatric assessment (CGA) is a process used to determine the health status of older patients, encompassing somatic, functional and psychosocial domains.9 A CGA can aid clinical decision making about the health status of an elderly patient with haematological cancer.9 Furthermore, it can result in the implementation of interventions which enhance a patient’s quality of life, such as measures to reduce social dysfunction and bodily pain.10 While a CGA does not have to be carried out by a geriatrician, the implementation of non-oncological interventions decreases considerably when left to the oncology specialist’s discretion rather than involving elderly care specialists.9,11 Thus, working with geriatricians can improve care of an elderly patient. As the size of elderly population increases, this collaboration between haematologists and geriatricians will occur more frequently.
Furthermore, the aforementioned increase in multiple comorbidities with age is likely to result in an increase in collaboration between haematologists and many other medical specialities, as well professions allied to medicine such as physiotherapy. There has been a drive within the UK to shift towards integrating care services, for example, merging health and social care services.3 Integrated care is co-ordinated, patient-centred care (rather than care focused on specific diseases), which aims to optimise patient health outcomes and satisfaction.3,12 There is no specific model of integrated care; the best approach is dependent upon local context. Certain models of integrated care, however, have been shown to improve both patient experience and care outcomes.3 Moreover, integrated care may be more cost-effective that delivering care based on single diseases. This may occur by reducing complications or reducing healthcare utilisation (which in turn frees up resources for other needs).3,11 This is important as average NHS spending for retired households is nearly double that for non-retired households.13 In addition, integrated care may reduce both gaps and duplication in service delivery, preventing repetition or delay in provision of services.3 This both reduces cost and improves patient satisfaction. For the haematologist, greater integration of care will therefore likely result in an increased collaboration with various medical professions. It should also reduce duplication or delay in seeing patients, resulting a more efficient service delivery.
Additionally, age may impact on a haematologist’s management plans for patients. The prognosis of certain haematological conditions (for example, both acute and chronic myeloid leukaemias) decreases with age.14 The worsening prognosis is a result of both comorbidities (including an increased risk of other haematological disorders) and the biological ageing process. The increase in comorbidity with age results in an increase in polypharmacy.3 Furthermore, the pharmacodynamics and pharmacokinetics of drugs alter with age. Consequently, first-line medications for haematological diseases may be contra-indicated, less effective or cause more side effects in elderly patients.14,15 For example, vincristine, steroids and L-asparaginase – all potentially used in the treatment of acute lymphoblastic leukaemia – are generally more toxic in older patients.14 Problems with using the best available treatment do not just apply to medications. Elderly patients are also less likely than younger patients to be considered fit for certain procedures, such as stem cell transplantation, due to an increased likelihood of multiple and/or serious comorbidities.16 In short, multiple comorbidities and resultant polypharmacy, combined with a biologically poor prognosis, mean that haematologists may not always be able to offer first-line treatments to elderly patients. In addition, the treatments they do offer, even if first-line, may be less effective in elderly patients.
The consequence of an ageing population, therefore, is that more patients with haematological diseases may be receiving less effective therapy to treat or manage their disease. One potential result of this is an increase in research specifically targeted at the management of haematological conditions in elderly patients, which is currently lacking.14 Examples of potential research include new drug treatments or different treatment regimes for elderly patients. Any research of this nature could impact on all haematologists. Evidently, haematologists are likely to be involved in carrying out any such clinical trials. More generally, however, research of this nature may impact on treatment guidelines used by all haematologists.
Also, there is a possibility that normal laboratory reference values may differ between older and younger patients. Take, for example, the case of anaemia. Anaemia is common in older adults. It affects around 10% of the population over the age of 65, and the prevalence rises with age.8 There has been debate amongst experts about whether the World Health Organisation’s definition of anaemia (haemoglobin concentration <120g/L in women and < 130g/L in men)17 should be revised for people aged over 65. The increasing prevalence of anaemia with age suggests that perhaps a lower threshold for anaemia could be considered, however research has demonstrated that low haemoglobin concentrations in adults are significantly correlated with increased mortality and the development of physical, functional or cognitive decline.8 Should consensus on this issue be reached, it is likely that new guidelines on the diagnosis and management of anaemia will be released. Changes to this laboratory value (or many others) amongst the elderly would surely impact on a haematologist’s practice, for example, through alterations to patient management.
In terms of the impact of using age to determine patient management, it should be noted that the elderly are a very diverse group in terms of functional capacity. On the one hand, there are fit and healthy elderly people who do not have functional limitations or disabilities, despite the overall population increase in the prevalence of chronic disease (a process known as compression of morbidity).18 On the other hand, there are elderly people who age faster and die younger than others, partly due to differences in social environments.18 This marked heterogeneity in function means that a person’s physiological state, rather than their age per se, may be more important in making decisions about patient management.14,18 This is already reflected in management guidelines in the UK. For example, the British guidelines on treatment of acute myeloid leukaemia state that standard treatment should be offered to patients up to the age of 60 years. Above 60 years, however, the patient’s ability to tolerate intensive chemotherapy (for example, due to other comorbidities) must be considered.16 Consequently, as the population ages, the need for haematologists to make such decisions about capacity to tolerate treatment are likely to become more frequent. Overall, the variability in functional capacity amongst the elderly, including the need to consider comorbidities, is likely to result in haematologists delivering more individualised patient care.
An Increasing Workload for Haematologists
A further effect of an ageing population may be an increasing workload for haematologists, despite increasing numbers of both haematology trainees and consultants with the UK.19 The incidence of both haematological malignancies and non-malignant haematological conditions increases with age.6 Consequently, as the population ages, more people are developing haematological disease and need to be seen by a haematologist. Furthermore, the management of haematological diseases has become more complicated and/or more intensive and thus requires more of a haematologist’s time.20 In addition, the need to collaborate with other specialities and consider the (potentially complex) management of the patient’s other conditions further increases clinical workload.20 Elderly patients are especially likely to have complicated management plans as a result of comorbidities and reduced physiological reserve. Laboratory work is also increasing each year, which generates more new patient referrals and increases the demand for haematologists to give advice on how to interpret test results.20 Combined, all of these factors contribute to an increased clinical workload for the practicing haematologist. Perhaps in the future, integrating care will streamlines services and thus contribute to a reduction in the haematologist’s workload. At least in the short-term, however, haematologists are likely to have an increased workload. This is mainly the result of an increase in the number of patients needing to be seen by a haematologist and the increased complexity of patient management, especially in the elderly.
A Possible Increase in Ethical Issues
Also, treating more elderly patients may mean that haematologists are faced with more ethical issues. While ethical concerns may occur with patients of any age, they more frequently occur with older people. This is because conditions that impair understanding and decision making, such as dementia, are more prevalent within the elderly population.3,5 This may make it more difficult for doctors, including haematologists, to ensure that elderly patients have understood their management options and given informed consent for treatment.5 Furthermore, it may be difficult for haematologists to determine if some elderly patients have the capacity to make certain decisions, especially important decisions such as those surrounding end-of-life care.5 When the patient lacks capacity and has no advanced directive, the haematologist may be involved in deciding whether to withdraw life-sustaining interventions.5 As the population ages, more patients will potentially have conditions that could impair understanding and decision making. As a result, ethical issues surrounding consent and capacity will become more frequent for haematologists.
In conclusion, as the population ages, more people will develop haematological diseases. In addition, these people are more likely to have multiple medical conditions requiring treatment. Consequently, haematologists will need to liaise with other medical specialities, especially geriatrics, as well as professions allied to medicine more frequently than in the past. Furthermore, haematologists are more likely to need to create personalised management plans for patients, based on their age, functional status and comorbidities. This, along with the increase in haematological diseases, will probably result in an increased workload for haematologists. Additionally, haematologists may be faced with more ethical issues, such as those regarding consent, capacity and decisions about withdrawing life-sustaining interventions, as they treat more elderly patients. Overall, perhaps the biggest short-term impact of the ageing population for haematologists is not a change in the way that they manage elderly patients but rather an increase in the number of elderly patients that they encounter, and thus a greater need to consider the patient’s age as part of his or her management.
Word count: 2251
1. Office for National Statistics. Annual mid-year population estimates, 2014 [online]. 2015 [cited 2015 Nov 14]. Available at: http://www.ons.gov.uk/ons/rel/pop-estimate/population-estimates-for-uk--england-and-wales--scotland-and-northern-ireland/mid-2014/stb---mid-2014-uk-population-estimates.html#tab-Main-Points
2. Office for National Statistics. National population projections: 2014-based projections [online]. 2015 [cited 2015 Nov 14]. Available at: http://www.ons.gov.uk/ons/rel/npp/national-population-projections/index.html
3. Oliver D, Foot C, Humphries R. Making our health and care systems fit for an ageing population. London, UK: The Kings Fund; 2014.
4. Mitnitski A, Collerton J, Martin-Ruiz C, Jagger C, von Zglinicki T, Rockwood K et al. Age-related frailty and its association with the biological markers of ageing. BMC Medicine 2015; 13: 161-9.
5. Mueller PS, Hook CC, Fleming KC. Ethical issues in geriatrics: a guide for clinicians. May Clin Proc 2004; 79: 554-562.
6. Chomienne C, McCann S, Green T, Hagenbeek A, Lacombe C, Guenova M et al. Age and ageing in blood disorders: EHA theme of the year 2013-2014. Haematologica 2013; 98: 831-832.
7. Suchard M. Immunosenescence: ageing of the immune system. S Afr Pharm J 2015; 82 (8): 28-31.
8. Merchant AA, Roy CN. Not so benign haematology: anaemia of the elderly. Br J Haematol 2011; 156: 173-185.
9. Hamaker ME, Prins MC, Stauder R. The relevance of a geriatric assessment for elderly patients with a haematological malignancy – a systematic review. Leuk Res 2014; 38 (3): 275-283.
10. Rao AV, Hsieh F, Feussner JR, Cohen HJ. Geriatric evaluation and management units in the care of frail elderly cancer patients. J Gerontol A Biol Sci Med Sci 2005; 60 (6): 798-803.
11. Kenis C, Bron D, Libert Y, Decoster L, Van P, Scalliet P et al. Relevance of a systematic geriatric screening and assessment in older patients with cancer: results of a prospective multicentre study. Ann Oncol 2013; 24 (5): 1306-1312.
12. Nolte E, Pitchforth E. What is the evidence on the economic impacts of integrated care [online]. 2014 [cited 2015 Nov 15]. Available at: http://webcache.googleusercontent.com/search?q=cache:r-o-XD1Oo-MJ:www.kingsfund.org.uk/sites/files/kf/media/Ellen%2520Nolte%2520-%2520What%2520is%2520the%2520evidence%2520on%2520the%2520economic%2520impacts%2520of%2520integrated%2520care.pdf+&cd=1&hl=en&ct=clnk&gl=uk
13. UK Parliament. The ageing population: key issues for the 2010 parliament [online]. 2010 [cited 2015 Nov 15]. Available at: http://www.parliament.uk/business/publications/research/key-issues-for-the-new-parliament/value-for-money-in-public-services/the-ageing-population/
14. Hassan M, Abedi-Valugerdi M. Hematologic malignancies in elderly patients. Haematologica 2014; 99: 1124-1127.
15. Golchin N, Frank S, Vince A, Isham L, Meropol S. Polypharmacy in the elderly. J Res Pharm Pract 2015; 4 (2): 85-88.
16. Milligan DW, Grimwade D, Cullis JO, Bond L, Swirsky D, Craddock C, et al. Guidelines on the management of acute myeloid leukaemia in adults. London, UK: British Society for Haematology; 2006.
17. World Health Organisation. Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity [online]. 2011 [cited 2015 Nov 17]. Available at: http://www.who.int/vmnis/indicators/haemoglobin.pdf
18. Suzman R, Beard JR, Boerma T, Chatterji S. Health in an ageing world – what do we know? Lancet 2015; 385: 484-6.
19. Centre for Workforce Intelligence. Haematology: CfWI medical fact sheet and summary sheet [online]. 2011 [cited 2015 Nov 17]. Available at: http://www.cfwi.org.uk/publications/haematology-cfwi-medical-fact-sheet-and-summary-sheet-august-2011
20. British Society for Haematology and Royal College of Pathologists. Haematology consultant workforce: the next 10 years [online]. 2008 [cited 2015 Nov 17]. Available at: https://www.rcpath.org/Resources/RCPath/Migrated%20Resources/Documents/P/pubhaem0108.pdf
There are 11.4 million people aged 65 or above in the UK and this figure is expected to increase by around 40 per cent to over 16 million by 2032.1 Many haematological diseases are more prevalent in the older population, particularly haematological malignancies and their prevalence is expected to increase as the population ages. In this essay I will describe how age associated changes in haematopoiesis, inflammation and protein levels influence the presentation and management of haematological cancer, anaemia and disorders of haemostasis. The treatment of these conditions is complicated by the presence of comorbidities and frailty. However, the health status of the over 65 age group is highly heterogeneous and mean life expectancy at the age of 70 ranges between 6.7 and 18 years.2 This creates the dilemma of deciding which patients will tolerate aggressive treatment with curative intent and in which a palliative approach is necessary.
Haematopoiesis and ageing
The increase in haematological dysfunction with age is associated with a reduction in the number of haematopoietic stem cells and a decline in haematological cell function. Additionally there is preferential production of cells of the myeloid lineage, which may explain why CML almost exclusively occurs in adults. These observations are thought to be due to a complex interaction between genetic changes, changes in the bone marrow microenvironment and systemic inflammation.3 Most of these changes have been demonstrated in mouse models but the existence of the decline in haematopoietic function and proliferative capacity in humans is supported by the observation that bone marrow donor age is a risk factor in recipients.4 These haematological changes form the basis of immunosenescence of ageing which in turn causes reduced immune surveillance of malignant cells and increased development of cancer, including haematological cancers. The impaired stem cell pool reduces tolerance to chemotherapy and necessitates less intense treatment.
Haemoncology and ageing
The cumulative exposure to environmental insults and progressive telomere shortening that occurs with ageing facilitates the process of clonal evolution, which is driven by mutational changes. The understanding that initiation and progression of cancer requires several mutations or ‘hits’ supports this. In addition to this mutation-centric view, age associated changes in selection pressures such as a modified bone marrow microenvironment and immunosenescence promotes expansion of malignant clones.3 It therefore follows that, with the exception of ALL, the incidence of all haematological cancers peak in the elderly population.
Age is a major risk factor for all cancers. Additionally, the manifestation of haematological cancers differs between young and elderly patients so that older patients have a biologically inferior prognosis. In AML ageing is associated with an increased occurrence of adverse cytogenetic changes while cytogenetically normal patients show decreased incidence of favourable mutations e.g. NPM1 and increased occurrence of those that confer a worse prognosis e.g. FLT3.5 Similarly, the subtype of DLBCL associated with unfavorable outcome, activated B-cell like DLBCL, is increasingly more common with advancing age.6
The poorer outcomes in elderly patients can also be attributed to their reduced tolerance of aggressive therapies. The diminished stem cell population in the elderly increases the risk of febrile neutropenia due to the myelopsuppressive effect of chemotherapy. Prophylactic granulocyte colony stimulating factor is recommended where the risk of febrile neutropenia is >20%.7 Increased toxicity of chemotherapy in the elderly also results from altered pharmacodynamics; the age-associated decline in renal function may lead to increased toxicity. Additionally age associated comorbidities carry risk; anthracycline, the first line therapy for Non-Hodgkin lymphomas, is cardiotoxic and should be avoided in those with increased cardiac risk.8
Guidelines for treatment of haematological malignancies are disproportionately based on outcomes in younger patients rather than the population age group most likely to be affected. This is due to exclusion of patients with comorbidities from clinical trials. However, one of a very small number of trials focussing on patients >80 years of age, the R-miniCHOP trial, showed that full dose targeted CD20 monoclonal antibody Rituximab combined with low dose CHOP immunochemotherapy provided a trade off between efficacy and toxicity in DLBCL patients. This was the first study to show that reduced dose chemotherapy could improve survival in addition to reducing toxicity.9
The addition of Rituximab to CHOP had previously been shown to improve rate of complete response and overall survival in patients aged 60-80 in comparison to CHOP alone.10 The success of this targeted therapy in the elderly population is mirrored in the tyrosine kinase inhibitor imatinib that has revolutionised the treatment of CML. Survival of patients on imatinib does not significantly differ below and above 60 years of age, most likely because targeted therapies have fewer side effects and are therefore better tolerated by the older population.11 However, no molecular response is achieved where adherence to imatinib is below 80%. Adherance to medication is an issue of particular importance in the elderly outside of clinical trials, particularly if they live alone or if they are taking multiple daily medications.12
Where targeted therapies are not yet readily available, for example multiple myeloma, the redeployment of already licenced drugs with a lower toxicity profile is providing a cost effective alterative to reducing toxicity of treatment. The use of thalidomide to treat myeloma has extended median survival to 4-5 years. More recently the screening of 100 off-patent drugs identified an antihelminth Niclosamide as a potential anti-myeloma therapy.13 Myeloma remains incurable and 85% of cases occur in over 60 year olds.13 Myeloma therefore represents a growing problem as the population ages and advances in treatment are highly anticipated.
Considerable debate surrounds the potential use of reduced intensity allogeneic stem cell transplant (SCT) in myeloma along with other haematological malignancies. Reduced intensity allogeneic SCT exploits the graft versus tumour effect of donor T lymphocytes not seen in autologous SCT but is preceded by considerably less myeloablative conditioning with chemotherapy and radiotherapy. It is therefore better tolerated in older patients. A study comparing reduced intensity allogeneic SCT with standard chemotherapy in patients aged 60-70 showed a 49% lower rate of relapse at 3 years in the SCT arm.14 However, graft versus host disease is still a significant risk.
Some caution should be executed in interpretation of trials in elderly patients. The sample may not be representative of the age group due to exclusion criteria although the therapy may be successful in a carefully selected group of patients. The challenge then lies in identifying which individuals are suitable for high intensity chemotherapy or SCT in a clinical setting. There is some evidence for the use of comprehensive geriatric assessment (CGA) to stratify patients. A retrospective study comparing survival of patients identified as ‘fit’ by CGA and survival of patients selected for treatment with curative intent by clinical judgement increased survival in those identified by CGA.15 However CGA can be time consuming and assessment of activities of daily living which have been shown to correlate strongly with overall survival in Non-Hodgkin Lymphoma, may be a more efficient and cost-effective alternative.2
Anaemia and ageing
Anaemia is extremely common in the elderly, occurring in 10% of adults over 65 and 20-30% of adults aged over 85. There is a strong link between anaemia and morbidity and mortality and it has been correlated with a decline in physical, functional and cognitive functioning. In the elderly this may translate to an increased fall risk. Anaemia is defined as a haemoglobin concentration of <120 g/l in women and <130 g/l in men. There has been considerable debate as to whether this figure should be adjusted for the elderly population. Some believe that the cut off should be raised given that some functional issues are noted at borderline anaemic levels while others suggest that the limit should be lowered to modulate the raised occurrence of anaemia with age. Further investigation into the effect on morbidity and mortality of a range of haemoglobin levels is required to settle this dispute.(16)
Around one third of anaemic individuals over 65 have anaemia due to a nutritional deficiency. Iron deficiency anaemia is the most common nutrient deficiency and is most commonly due to gastrointestinal bleeding. This may suggest malignancy and requires further investigation. Folate deficiency is common in the elderly due to relative malnutrition. Access to fresh fruit and vegetables may be difficult for elderly people particularly for those living alone or in residential care.(16) A further third of anaemia over 65 is due to anaemia of chronic disease, which occurs alongside conditions characterised by long-term inflammation such as malignancy or rheumatoid arthritis. The inflammatory cytokine IL6 stimulates release of hepcidin from the liver. Hepcidin triggers internalisation and degradation of ferroportin, the protein responsible for transporting iron out of cells. Iron becomes trapped, particularly in macrophages, and is therefore unavailable.17
The remaining third of the anaemic elderly are labelled with the term ‘unexplained anaemia of the elderly’ for which there is no guidelines for management. A proportion of this group may be explained by an inflammatory state of ageing which appears to have some overlap with anaemia of chronic disease. Some studies have noted raised hepcidin and ESR levels in unexplained anaemia of the elderly while IL6 levels have been found to be both raised and reduced compared to controls on separate occasions. Unexplained anaemia of the elderly may be partially explained by the observed insensitivity of HSCs to erythropoietin with ageing and the declining and dysfunctional haematopoietic stem cell population, which have reduced potential for erythropoiesis.16 These mechanisms are speculative and further research into the aetiology of anaemia in the elderly is necessary.
Anaemia appears to be a strong marker for physical decline but should anaemia of chronic disease and unexplained anaemia be treated? The most effective means of raising haemoglobin levels is through blood transfusion. However, in elderly patients with cardiac failure this is not appropriate given the risk of transfusion associated circulatory overload. Given that unexplained anaemia is typically more mild than iron deficiency anaemia, improving haemoglobin levels may not significantly improve function as any improvement may be negated by other factors associated with ageing such as reduction in muscle mass and decline in pulmonary and cardiac function.16 However, it may be dangerous to simply attribute anaemia to ageing without investigation, as a reduced haemoglobin level could be an indication of serious pathology.
Haemostasis and ageing
The increased risk of thrombosis with age is reflected in changes of components of the haemostatic system. Coagulation system proteins including FV, FVII, FVIII, FX and fibrinogen increase with age. Fibrinogen and VIII are acute phase proteins and fibrinogen production is in fact stimulated by IL6, suggesting that the inflammatory state of ageing may in part underlie these observed changes in coagulation factors. Fibrinogen levels have been identified as a risk factor for stroke and MI while increased FVIII has been shown to correlate with risk of venous thrombosis. In addition to these changes, fibrinolysis in the elderly is impaired as evidenced by the increased levels of plasminogen activator inhibitor (PAI-1) which inhibits fibrinolysis. PAI-1 is itself another acute phase protein. A corresponding increase in anticoagulant proteins with age has not been observed and therefore this imbalance could shift the haemostatic system towards a hypercoagulable state.18
The increased thrombotic risk with age necessitates continued emphasis of VTE prophylaxis in hospitals. Additionally a greater proportion of the population will require oral anticoagulants as the population ages. Elderly patients taking Warfarin require more frequent monitoring of INR as they are more susceptible to bleeding side effects.19 Increased monitoring by pharmacists and self-monitoring may be necessary to help the haematology profession deal with the increased demand on their anticoagulation clinics.
Contrastingly, population ageing also has implications for management of haemophilia patients. Previously a childhood disease, haemophilia now has an almost normal life expectancy due to advances in treatment with the development of recombinant factor VIII and IX in the 1990s. Therefore the haematology specialty must adapt to expect more elderly patients with haemophilic arthropathy and associated chronic pain. Additionally, there is a large population of patients who contracted hepatitis C and HIV via coagulation factor transfusion before viral inactivation of blood products became routine who are only now reaching old age. Before purification of blood products, 80-90% of haemophilia patients were infected with hep C. Patients coinfected with Hep C and HIV are twice as likely to develop cirrhosis and liver failure than those infected with Hep C alone. Studies suggest that around 1/3 of haemophilia patients are depressed and therefore there is a need for a psychiatric component of care.20
Elderly patients with haematological cancer have a worse prognosis due to comorbidities and their impaired tolerance of chemotherapy. Management of haematological malignancy in the elderly often requires less toxic treatment. However, management decisions should be made on a case-by-case basis and take into account the biological in addition to the chronological age of the patient. Formal geriatric assessment may aid selection of good candidates for intensive treatment and cooperation with geriatric doctors facilitates provision of sufficient nutrition and social support, which may influence outcome. Significant treatment advances have been made in some cancers; CML is now extremely treatable. However, others such as myeloma lag behind and will represent a significant burden as the population ages until more targeted therapies are available. There is desperate need for increased recruitment of patients over 75 into cancer trials. Only then can evidence based treatment decisions be made for this age group. Guidelines for target haemoglobin levels in the elderly and treatment of unexplained anaemia are required based on evidence of patient functional ability. The elderly are inherently thrombotic although age is a risk factor for bleeding for those anticoagulated on warfarin. An emerging cohort of elderly virally infected haemophilia patients exists and management must be adapted to accommodate their associated comorbidities. Inflammation and haematopoietic decline are recurrent themes that underlie haematological dysfunction in ageing and further research into these areas will aid management in the future.
Word count - 2307
1. UK A. Later Life in the United Kingdom 2015 [updated 7 December 2015]. Available from: http://www.ageuk.org.uk/Documents/EN-GB/Factsheets/Later_Life_UK_factsheet.pdf?dtrk=true. (accessed 30 December 2015)
2. Peyrade F, Gastaud L, Ré D, Pacquelet-Cheli S, Thyss A. Treatment decisions for elderly patients with haematological malignancies: a dilemma. The Lancet Oncology. 2012;13(8):e344-e52.
3. Henry CJ, Marusyk A, DeGregori J. Aging-associated changes in hematopoiesis and leukemogenesis: what's the connection? Aging (Albany NY). 2011;3(6):643-56.
4. Kollman C. Donor characteristics as risk factors in recipients after transplantation of bone marrow from unrelated donors: the effect of donor age. Blood. 2001;98(7):2043-51.
5. Kumar CC. Genetic abnormalities and challenges in the treatment of acute myeloid leukemia. Genes Cancer. 2011;2(2):95-107.
6. Mareschal S, Lanic H, Ruminy P, Bastard C, Tilly H, Jardin F. The proportion of activated B-cell like subtype among de novo diffuse large B-cell lymphoma increases with age. Haematologica. 2011;96(12):1888-90.
7. Aapro M CJ, Kamioner D. Prophylaxis of chemotherapy-induced febrile neutropenia with granulocyte colony-stimulating factors: where are we now? Support Care Cancer. 2010;18(5):529-41.
8. Aapro M, Bernard-Marty C, Brain EG, Batist G, Erdkamp F, Krzemieniecki K, et al. Anthracycline cardiotoxicity in the elderly cancer patient: a SIOG expert position paper. Ann Oncol. 2011;22(2):257-67.
9. Peyrade F, Jardin F, Thieblemont C, Thyss A, Emile J-F, Castaigne S, et al. Attenuated immunochemotherapy regimen (R-miniCHOP) in elderly patients older than 80 years with diffuse large B-cell lymphoma: a multicentre, single-arm, phase 2 trial. The Lancet Oncology. 2011;12(5):460-8.
10. Coiffier B, Lepage E, Briere J, Herbrecht R, Tilly H, Bouabdallah R, et al. CHOP chemotherapy plus rituximab compared with CHOP alone in elderly patients with diffuse large-B-cell lymphoma. N Engl J Med. 2002;346(4):235-42.
11. Gugliotta G, Castagnetti F, Palandri F, Breccia M, Intermesoli T, Capucci A, et al. Frontline imatinib treatment of chronic myeloid leukemia: no impact of age on outcome, a survey by the GIMEMA CML Working Party. Blood. 2011;117(21):5591-9.
12. Marin D, Bazeos A, Mahon FX, Eliasson L, Milojkovic D, Bua M, et al. Adherence is the critical factor for achieving molecular responses in patients with chronic myeloid leukemia who achieve complete cytogenetic responses on imatinib. J Clin Oncol. 2010;28(14):2381-8.
13. Khanim FL, Merrick BA, Giles HV, Jankute M, Jackson JB, Giles LJ, et al. Redeployment-based drug screening identifies the anti-helminthic niclosamide as anti-myeloma therapy that also reduces free light chain production. Blood Cancer J. 2011;1(10):e39.
14. Champlin R. Reduced-intensity allogeneic hematopoietic transplantation should be considered a standard of care for older patients with acute myeloid leukemia. Biol Blood Marrow Transplant. 2011;17(12):1723-4.
15. Tucci A, Ferrari S, Bottelli C, Borlenghi E, Drera M, Rossi G. A comprehensive geriatric assessment is more effective than clinical judgment to identify elderly diffuse large cell lymphoma patients who benefit from aggressive therapy. Cancer. 2009;115(19):4547-53.
16. Merchant AA, Roy CN. Not so benign haematology: anaemia of the elderly. Br J Haematol. 2012;156(2):173-85.
17. Ganz T. Hepcidin, a key regulator of iron metabolism and mediator of anemia of inflammation. Blood. 2003;102(3):783-8.
18. Franchini M. Hemostasis and aging. Crit Rev Oncol Hematol. 2006;60(2):144-51.
19. Shoeb M, Fang MC. Assessing bleeding risk in patients taking anticoagulants. J Thromb Thrombolysis. 2013;35(3):312-9.
20. Canaro M, Goranova-Marinova V, Berntorp E. The ageing patient with haemophilia. Eur J Haematol. 2015;94 Suppl 77:17-22.
In 2014 the essay title was 'Does the 21st Century Haematologist need a Microscope?'. The winner was Tom Handley of the University of Oxford, and the runner-up was Kiloran Metcalfe, of the Oxford Medical School.
A haematologist, from the Greek haima "blood" and logos “to study”, studies the blood and haematopoietic tissues of the body, in order to diagnose disorders of blood components.1 The study of blood permeates medicine, and the microscope is an iconic tool that is bound to our perception of the field. The fate of this instrument is dependent on a technological arms race, where more recent technology has the potential to render the microscope’s place as a primary diagnostic tool obsolete. This essay will discuss haematology’s past, present and conceivable future. It will explore the microscope’s precarious position in haematology, representative of potential paradigmatic shifts in healthcare over the next hundred years.
What is the role of a haematologist in the 21st century?
A haematologist’s principal concern is the accurate identification of disease caused by (or relating to) disorders of the blood. A modern haematologist can split their time between the laboratory and the clinic, deriving the aetiology of blood-based disease through laboratory diagnostics, and implementing treatments as a clinician. In order to understand what is aberrant, they must first fully understand what is normal. This requires a comprehensive understanding of the cells and tissues of the blood system. In order for blood to fulfil its function in the circulatory system, the presence and relative abundance of the cellular and proteinaceous components of the solution must be maintained to ensure continued physiological function. However, these components are too small to be seen by the naked eye, and their interactions too fleeting and complex to be understood visually. We must therefore use tools to analyse the blood’s morphological and functional characteristics.
Microscope: a tool that founded a field
An informed diagnosis requires an understanding of the cellular appearance and physiological performance of the blood. The microscope enables haematologists to understand the constituent parts of the blood. It allowed the field to develop beyond philosophical ponderings about the substance of this life giving fluid, and visualise what lay within the murky liquid. The microscope gives an image of the subject at a higher magnification and resolution than can be seen by the naked eye, founding a morphological understanding of the cellular composition of blood.
Microscopes were discovered in the late 16th century, but the name was first used to describe Galileo’s compound optical microscope in the 17th century.2 When a generation of scientists led by Van Leeuwenhoek in the 17th century studied living organisms using this optical microscope, a new world opened before their eyes. The “mikros” (small) “skopeîn” (to view) allowed them to discover the world of microbiology3, and with it tools and information that would become pivotal to haematology. Being able to image cells for the first time became one of the most important developments in life-sciences, and allowed scientists to push the boundaries of how they understood the world.
However, it would be another 150 years before other scientists used the microscope to further their understanding of the component parts of the blood beyond Leeuwenhoek’s ‘Red Corpuscles’.4 The 19th century saw the discovery of white cells and platelets. Osler, along with Schultze and Bizzozero are among a number of scientists who contributed to the discovery of platelets.5 The gap between the first discovery of the cellular organisation of the blood and a more comprehensive understanding of its component parts was due in part to the development of the microscope as a tool to study the blood. Traditional light microscopy is limited by the quality of illumination, the magnification capacity of the lens, and the staining techniques that are employed. Biological tissue has little inherent contrast under a microscope6, and thus requires staining to both give contrast to the tissue and highlight particular features of interest. Improvements in the technological capabilities of the microscope enabled a much greater morphological understanding of the blood, its appearance in vitro, and some of the basic pathologies.
How is a microscope used today?
Microscopes are used to view cells in an in vitro environment (for example in a thin blood smear). This allows some conclusions to be drawn about their physiological state in the body (for example, an elevated number of white cells in a sample may be indicative of an infection in vivo). Today, the microscopic appearance of cells is used to create a picture of how the cells may be performing in the body. Aberrant shapes characterise a multitude of diseases, and fill our textbooks and journals with examples of different states of disease. Microscopes are particularly useful in circumstances where a rapid diagnosis is important, for example, the initial stage of diagnosing acute leukaemias uses a morphological analysis, enabling fast initiation of treatment.7
The 20th century brought great technological advancements, including the development of a new way of studying cells with microscopes using electrons, rather than beams of light. As microscopy techniques moved toward greater levels of magnification and higher levels of resolution we began to understand more about the morphological characteristics of the cells and tissues of the body. Clinicians and scientists also began to employ diagnostic techniques that did not require imaging of the cells, but were based on molecular signals too small and complex to be detected by the microscope.
What technologies may supersede the microscope?
A modern haematologist’s task is to quickly detect a wide range of diseases and disorders, often in a large number of patients. In order to do this, machines have been designed to detect the presence of molecular markers correlated with sequences that are specific to a certain condition, or to a mutation that is related to a certain prognosis (for example JAK2 mutations linked with poor prognosis in myeloproliferative disorders8). The myriad of machines that have been developed to assess such molecular parameters have helped us to develop better markers, tactics and techniques for detecting disease. A molecular approach enables ‘personalised medicine’, where the treatment of disease is specific to the patient.
This is useful, for example, in some cancers where specific protein mutations can lead to different treatment regimes. Perhaps the classic example of this hails from haematology – specifically the treatment Chronic Myeloid Leukaemia. David Hungerford’s discovery of the bcr/abl (Philadelphia) chromosomal translocation allowed the development of a targeted treatment for CML. In the 1990’s the tyrosine kinase imatinib was developed, which blocks the BCR-Abl enzyme, and stops it from adding phosphate groups. As a result, these cells stop growing, and can commit apoptosis9,10. One could argue that the goal of personalised medicine is to re-create this achievement for other conditions.
Genomics uses gene sequencing and bioinformatics to analyse genomes and investigate the roles different polymorphisms can play in disease. Genomics may well be the next iteration of personalised medicine, and could enable the development of treatments specific to an individual’s disease. The recent announcement of a program to sequence 100,000 genomes within three years at 11 centres across the U.K. aims to develop new tests and drugs and is a sign of rapidly developing research on a molecular level11.
This program is symbolic of a movement in the medical field, which is no longer so focused on a morphological relationship between form and function, but rather has adopted a ‘black box’ methodology. A doctor or scientist may take a sample from a patient and present this sample to a machine. This machine then runs a number of tests on the sample designed to detect minute differences between cells, and generate a (usually quantitative) response. Even the strongest microscopes are unable to resolve differences at these levels. Such machines generate a series of results from an array of molecular diagnostic machinery to come to a relatively binary decision as to whether a patient has ‘disease X’ or not.
However, these machines are not always available to clinicians, and as such haematology is in a transitional period. Exclusive use of the microscope is not practical in some clinical settings, as better molecular options can provide more conclusive and specific diagnoses; however these molecular techniques cannot be solely used, as they are expensive and slow, and not yet widely distributed. Will the 21st century change the way technology is used, and will there continue to be a place in the haematologist’s toolbox for the microscope?
The microscope’s uncertain future
Regardless of the obvious merits of molecular diagnostics, microscopes are still used in clinics and continue to be critical to the cutting edge of research. The 2014 Nobel Prize for Chemistry was jointly awarded to Hell, Betzig & Moerner for the development of super-resolved fluorescence microscopy, surpassing the previous limitations of the light microscope12. They had separately found different solutions to pass Abbe’s diffraction limit, the point that had previously limited the resolution of light microscopes. By bypassing this physical limitation on imaging, higher resolution images can now be made of living cells.
Such advances in microscope technology indicate that the potential exists for even greater development in this area. There are numerous ways that microscopy could be manipulated to provide novel findings, from advances in staining techniques to further revisions in the limits of resolution. Indeed, in some areas of research intra-vital microscopy is enabling real-time in vivo visualisation of the interactions between cells. Ronald Germain and his colleagues have been able to challenge the immune system and visualise the consequences in the lymph nodes13. Perhaps advances in techniques in these areas may enable us to image cells in their physiological environment.
Advances in these areas may be symbolic of revolutions in the field of microscopy. The microscope of the future may be as different to today’s as our microscopes, with their beams of electrons and high fluorescence multi-coloured images, would seem to Galileo or Van Leeuwenhoek. The advances that the next hundred years may bring could allow us to visualise not only cells, but proteins on cells. Enabling analysis of morphological appearance, and real-time changes in response to physiological or test interventions. A “mikros” (small) “skopeîn” (to view) may enable us to see things that today we cannot even imagine.
The response to this question also depends on where one sits in the world. In many Western medical systems a movement toward increasingly complex and expensive technologies persists, with complex molecular analyses of many conditions leading the spear-tip of research. However, in the developing world access to haematological facilities and molecular diagnostic laboratories is rare, or even non-existent. Many developing countries use basic equipment to provide the fundamentals of haematological analysis and diagnosis. Microscopes are essential under these circumstances. They are cheap, easily transportable, and efficient in the hands of a skilled technician. However, if molecular diagnostics were to continue to miniaturise and become even cheaper, there is a chance that they could leapfrog microscopy to become the most prevalent diagnostic technique.
The fate of the microscope in haematology raises a more philosophical point about how we are going to treat patients in the future. If a haematologist no longer needs to use a microscope, they may no longer be basing their diagnoses upon the relationship between morphological form and pathological dysfunction that the field was founded upon. Instead, they may come to rely entirely upon a series of machines and analytical tests. The laboratory-based side of a haematologist’s work would move further towards the bioinformatic world, caught up in the rush of a new era of genomic science.
Currently, it is vital for a haematologist to learn how to use a microscope, as a great deal of the diagnostics and protocols currently in place require an ability to act rapidly and effectively. As healthcare in the world develops, different healthcare systems will progress from using the microscope at different rates, and thus its use will certainly not cease in the immediate future. Some older medical technologies have been cast aside once they were superseded by new inventions (such as the cane once used by physicians to ward off the miasmas of their patients)14. However, unlike some of the technologies history has cast aside, microscopes have had decades of empirically proving their worth in the medical field, and so are unlikely to meet the same obscure fate.
Does the 21st century haematologist need a microscope? At the moment, they do. However, over the course of the next 85 years of our century, a technological arms race may change the face of medicine. A new, exclusively molecular, branch of medicine may appear in the place of the discipline we currently call haematology. New technologies within such a discipline may entirely supersede the humble microscope, or they may simply work alongside it. Science is a discipline in constant flux, where it is nigh on impossible to stipulate the importance of a tool in a future environment that cannot be predicted. In the immediate and foreseeable future, however, microscopes will continue to be used by haematologists, inform a large portion of their diagnostic work, and remain rooted in the field of haematology.
1. NHS cited from [Internet].; cited [November 2014]Available from: www.NHScareers.nhs.uk
2. Gould, Stephen Jay (2000). "Chapter 2: The Sharp-Eyed Lynx, Outfoxed by Nature". The Lying Stones of Marrakech: Penultimate Reflections in Natural History. New York, N.Y: Harmony.
3. Wootton, David (2006). Bad medicine: doctors doing harm since Hippocrates. Oxford [Oxfordshire]: Oxford University Press. ISBN 0-19-280355-7
4. M. Bessis and G. Delpech.(1981), Discovery of the Red Blood Cell with notes on priorities and credits of discoveries. Blood Cells 7:447-48
5. Max Schultze (1865), G. Bizzozero (1882) and the discovery of the platelet, 2006 Blackwell Publishing Ltd, British Journal of Haematology, 133, 251–258
6. Young, B., Heath, J. W., Stevens, A., Lowe, J. S., Wheater, P. R., & Burkitt, H. G. (2000). Wheater's functional histology: A text and colour atlas. Edinburgh: Churchill Livingstone.
7. Bennet, J (1985) Proposed Revised Criteria for the Classification of Acute Myeloid Leukemia: A Report of the French-American-British Cooperative Group. Ann Intern Med. 1985;103(4):620-625
8. Peter J. Campbell (2006) V617F mutation in JAK2 is associated with poorer survival in idiopathic myelofibrosis; Blood: 107 (5)
9. Hungerford et al. (1960) Chromosome preparations of leukocytes cultured from human peripheral blood. Experimental Cell Research Volume 20, Issue 3, Pages 613-616
10. Goldman JM, Melo JV (October 2003). "Chronic myeloid leukemia – advances in biology and new approaches to treatment". N. Engl. J. Med. 349 (15): 1451–64
11. Genomics England project cited from [Internet].; cited [December 2014]Available from: http://www.genomicsengland.co.uk/
12. Nobel Prize cited from [Internet],; "The Nobel Prize in Chemistry 2014". Nobelprize.org. Nobel Media AB 2014. Web. 31 Dec 2014.
13. Bajenoff, M. & Germain, R. N (2007).Seeing is believing: a focus on the contribution of microscopic imaging to our understanding of immune system function. Eur. J. Immunol. 37, S18–S33
14. Entry to the Royal College of Physicians, based upon an engraving by William Hogarth, (1736) “Consultation of physicians or the arms of the undertaker”, cited from: https://www.rcplondon.ac.uk/update/physicians-cane-and-secrets-it-contained
The microscope has been a mainstay of haematological investigation for hundreds of years. Microscopy was first used to observe red blood cells as early as 1658, but it took until the mid-19th century before the components of blood and their role in disease were more fully examined. Using stains developed by Ehrlich, microscope wielding researchers started to link what they saw in their blood films to clinical syndromes. Different forms of anaemia were identified and their causes classified. Identification of leukocytes and their role in diseases such as leukaemia came later. Microscopic examination of blood films and bone marrow samples now forms an important part of haematological diagnosis. However, new technologies such as automated cell counting, flow cytometry and genetic analysis may provide alternatives to the traditional method. Here we will examine what such techniques bring to the process of haematological diagnosis and whether they can truly replace the microscope.
Automated blood cell counting
The full blood count is one of the most routine tests done in any diagnostic setting. It provides a breakdown of the numbers of the different blood cells present in a sample. This is important in the diagnosis of conditions such as anaemia or haematological malignancies but is also used in other conditions such as infection or in inflammatory disorders. Traditionally, a full blood count is done manually. A blood film preparation is examined to provide an estimate of the numbers of different cells. This process is laborious, time consuming and only examines a small number of cells. While manual counting remains the gold standard, automated analysers are now used routinely to perform blood counts. Improved software and new technologies have allowed automated blood cell counters to evolve dramatically over the last two decades. New analysers are able to perform not only simple cell type counts, but also identify cells not normally found in peripheral blood, e.g. blasts, atypical lymphocytes and nucleated red cells.
While analytic performance is good for leukocyte and red cell counts, it is less satisfactory with some aspects of the leukocyte differentiation count and reticulocyte or platelet counts particularly when cell numbers are low3. However the technology continues to improve and newer analysers are already showing comparable results to manual counting in areas such as immature granulocyte and nucleated red cell counts. One area where analysers have continued to have poor performance is in platelet counts. While some studies have indicated that automated platelet counts are a reliable predictor of clinical bleed risk it has been shown that analysers are inaccurate in measuring platelets in severe thrombocytopenia. They produce variable results and often wrongly over estimate platelet counts, which could clinically result in under perfusion of platelets. For low platelet counts external quality control, such as manual counting, is still needed until analyser accuracy is improved.
For simple blood cell counts, analysers do seem able to replace microscopy. They are able to provide rapid results and examine a far larger number of cells but there are still doubts about the efficacy of automated analysers where there is abnormal cell morphology. While it has been shown that counters are still able to count the abnormal and immature cells that can be seen in peripheral blood in disease states such as haematological malignancies2, other abnormalities will still disrupt the count. Analysers can, in some circumstances, produce spurious counts e.g. if blood sample is inadequate, if there is platelet agglutination (such as that caused by EDTA), or if the patient’s pathology leads to fragmented red cells, other cell fragments, bacteria, fungi or abnormal lipids . In these cases a manual count is still often required. So while analysers can relieve much of the burden of blood counting, the microscope is still required in many abnormal states.
Analysis of Cell Morphology
One major advantage manual counting has over automated counting is that it allows the study of abnormal cellular characteristics at the same time. For example abnormal red cells such as spherocytes and poikilocytes are readily observed down the microscope but may be missed in an automated cell count and could even cause an error in the count. Some analysers are able to detect cellular abnormalities and can ‘flag’ these samples for further examination5. But there is variability in how analysers are calibrated and not all will flag every sample that needs further examination8 meaning that some samples will be missed or subject to error prone counting.
It is likely that newer analysers will be able to detect and report on more abnormal cell types. Already in the last few decades, they have gone from just being able to detect red cells and leukocytes, to performing differential counts, reticulocyte counts and identify immature white cells. However, histological examination remains important. In many haematological malignancies, differential diagnoses can often only be ruled out after histological examination of a bone marrow or lymph node sample. A histological examination can also often give a very rapid diagnosis, for example the presence of Reed-Sternberg cells in a lymph node biopsy is diagnostic for Hodgkin’s lymphoma. Indeed, for some conditions e.g. Myelodysplasia, blood film analysis and histological examination of a bone marrow sample are actually required for diagnosis and classification .
Blood films also allow use of techniques such as immunohistochemistry where presence or absence of specific cell markers can be analysed. For example, presence or absence of specific markers can be used to predict prognosis in diffuse large B-cell lymphoma . However, technological advances have led to the development of flow cytometry as an alternative to immunohistochemistry, allowing more sophisticated analysis.
Flow cytometry is a laser based technique which can be used to measure properties of individual cells such as size or surface antigens. Examination of surface antigens allows leukaemia or lymphoma typing. It can also be used to look for antibodies e.g. against platelets or neutrophils which is an important diagnostic feature of certain autoimmune diseases. DNA content can also be measured, allowing examination of aneuploidy seen in some cancers e.g. multiple myeloma and acute lymphoblastic anaemia11.
Flow cytometry can also be used for cell counting or sorting. For example, it allows the detection and quantification of fetal red cells in maternal blood. This can be important clinically when there is a need to prevent maternal exposure to fetal blood cells such as in the case of an Rh- mother carrying a Rh+ fetus. Flow cytometry also allows counting of reticulocytes, in a manner that is more sensitive and precise than manual counting. However, this method does not distinguish Howell-Jolley bodies from reticulocytes11. So while it is more accurate than other methods of automated counting, it is not infallible.
Flow cytometry has also become an integral part of diagnosis of leukaemias and lymphomas . It is as effective as manual histological examination and allows more subtle differentiation between different antigen profiles. It is thus becoming a very useful diagnostic tool and in certain aspects of cellular analysis it is already replacing microscopy.
In certain haematological conditions, diagnosis could theoretically be made solely on the basis of a genetic test. This applies to certain inherited disorders such as heamoglobinopathies and clotting disorders and also to mutations found in certain myeloid malignancies. For example, Sickle cell anaemia can be diagnosed based on a blood film but equally, one could be just as sure of the diagnosis after a genetic test. Such genetic tests even allow a prenatal diagnosis . In sickle cell anaemia, genetic testing is relatively simple, as there is a recognised single point mutation. In other conditions e.g. thalassemia, the mutation is more variable, but within the last decade DNA sequencing has become relatively cheap, so genetic testing is also readily available.
Genetic testing is also used in malignancy. Particular translocations or mutations are characteristic of certain diseases. For example, in chronic myelogenous leukaemia (CML) there is the BCR-ABL1 translocation, which is easily detected using real time quantitative PCR (though it can also be detected using fluorescence microscopy). This translocation is diagnostic for CML and the mutated gene product produced can now be targeted with specific therapies. Detection of BCR-ABL1 positive cells using sensitive PCR can also be used for disease monitoring e.g. to detect malignant cells after transplant . Similarly, Polycythaemia Vera can be diagnosed solely on the basis of a high haematocrit and a JAK2 mutation in a non-hypoxic patient.
In certain conditions, genetic testing is therefore a viable alternative to microscopy. It can provide rapid results at an ever lower cost. But it is only relevant for inherited conditions with a known gene mutation or in malignancies where there is a characteristic genetic mutation. In other malignancies it may provide prognostic information, but is unlikely to be diagnostic and so in these conditions and other haematological disorders, histological examination is still likely to be required.
In conclusions, it appears that the 21st century haematologist does still require a microscope. While new technologies are providing alternatives and are likely to dramatically reduce the amount of time spent on histological examination, for some things the human eye is still the best we’ve got. Automated cell counting can manage most routine blood counts but errors still occur and in abnormal blood films, further manual examination is required, particularly when it comes to cell morphology. Flow cytometry allows rapid analysis of cellular characteristics and its uses are ever increasing. Similarly, genetic analysis also has the potential to replace the microscope in certain fields and is a very important diagnostic and prognostic marker.
It seems probable that, over time, use of the microscope will decrease as many aspects of microscopic examination are superseded by newer technologies. Perhaps a haematologist at the end of the 21st century will have no use for a microscope, but that seems unlikely. At the moment, manual histological examination remains a gold standard for many aspects of haematological diagnosis and even if the machines can take over some of that role, there will always be occasions when something goes wrong and it’s just easier to take a look yourself.
Rolleston, R. The History of Haematology (Section of the History of Medicine). Proc R Soc Med 27, 1161-1178 (1934)
Hijiya, N. et al Utility of automated counting to determine absolute neutrophil counts and absolute phagocyte counts for paediatric cancer treatment protocols. Cancer 101, 2681-6 (2004)
Buttarello, M., Plebani, M. Automated Blood Cell Counts: State of the Art. Am J Clin Pathol 130, 104-116 (2008)
Field, D., Taube, E. and Heumann, S. Performance evaluation of the immature granulocyte parameter on the Sysmex XE-2100 automated hematology analyser. Lab Hematol 12, 11-14 (2006)
Ruzicka, K., Veitl, M., Thalhammer-Scherrer, R. and Schwarzinger, I. The new hematology analyzer Sysmex XE-2100: performance evaluation of a novel white blood cell differential technology. Arch Pathol Lab Med 125, 391-6 (2001)
Lawrence, J. B. et al Reliability of automated platelet counts: comparison with manual method and utility for prediction of clinical bleeding. Am J Hematol 48, 244-250 (1995)
Segal, H. C. et al Accuracy of platelet counting haematology analysers in severe thrombocytopenia and potential impact on platelet transfusion. Br J Haematol 128, 520-525 (2005)
Zandecki, M., Genevieve, F., Gerard, J. and Godon, A. Spurious counts and spurious results on haematology analysers: a review. Part I: platelets. Int J Lab Hematol 29, 4-20 (2007)
Killick, S. B. et al Guidelines for the Diagnosis and Management of Adult Myelodysplastic Syndromes. British Committee for Standards in Haematology (2013)
Sjo, L. D. et al Profiling of diffuse large B cell lymphoma by immunohistochemistry: identification of prognostic subgroups. Eur J Haematol 79, 501-507 (2007)
Brown, M. and Wittwer, C. Flow Cytometry: Principles and Clinical Applications in Hematology. Clinical Chem 46, 1221-1229 (2000)
Kawano-Yanamoto, C. et al Two-color Flow cytometry with a CD19 Gate for the Evaluation of Bone Marrow Involvement of B-cell Lymphoma. Leuk Lymphoma 43, 2133-2137 (2002)
Saiki, R. K. et al Enzymatic amplification of beta-globin genomic sequences and restriction site analysis for diagnosis of sickle cell anaemia. Science 230, 1350-1354 (1985)
Eder, M. et al Monitoring of BCR-ABL expression using real time RT-PCR in CML after bone marrow or peripheral blood stem cell transplantation. Leukemia 13, 1383-1389 (1999)
Spivak, J. L and Silver, R. T. The revised World Health Organisation diagnostic criteria for Polycythaemia vera, essential thrombocytosis, and primary myelofibrosis: an alternative proposal. Blood 112, 231-239 (2007)