Commentary
Prognosis research: Why is Dr. Lydgate still waiting?

https://doi.org/10.1016/j.jclinepi.2006.02.005Get rights and content

Abstract

Background

Understanding prognosis—the future risk of adverse outcomes among people with existing disease—plays third fiddle behind clinical research into therapeutic interventions and novel diagnostic technologies.

Methods and Results

Diseases show marked variations in a wide range of prognostic outcomes, yet these variations have seldom been the subject of systematic and sustained epidemiologic and multidisciplinary research. This is important to prioritize hypotheses for testing in intervention studies in groups, and to refine tools for prognostication in individuals. Methodologic standards for the design, conduct, analysis and reporting of prognosis research are required. Training is needed for the clinicians, policymakers, and payers who use prognostic information.

Conclusion

Here, arguments detracting from the potential scope of prognosis research are rebutted and misconceptions addressed with the aim of stimulating debate on the evolving role of prognosis research.

Introduction

“Diseases of the heart are eminently difficult to form predictions on. A good deal of experience—a more lengthened observation—is wanted on the subject. But it is my duty to tell you that death in this disease is often sudden. At the same time, no such results can be predicted. Your condition may be consistent with a tolerably comfortable life for another 15 years or more.” Dr. Lydgate

George Elliott “Middlemarch” 1872

Is prognosis research moribund? Despite more than 130 years of “lengthened observation,” Dr. Lydgate's words remain largely true today. Everyone can cite recent advances in treating and diagnosing disease. Fewer would be able to give examples of progress in our understanding of the future course of established diseases—prognosis. Fewer still would agree on the definition (Box 1) and scope (Box 2) of prognosis research. Prognosis occupied a central role in Greek medicine, yet this has been dying out over the last century [1]. The repeated calls for improved prognostic understanding in the first half of the 20th century [2] went largely unheeded in the second half [3]. Currently, the broad scope of prognostic enquiry (Box 2) is being pursued in a piecemeal, parallel endeavor. Few systematic and sustained programs of prognosis research are funded to investigate the range of factors (host, disease, environment) associated with the range of outcomes constituting prognosis.

Yet by many indicators, prognosis research is vibrant. The questions have never been more important: the length of time that people live with disease is increasing [4]; up to one-third of the life expectancy at age 40 is estimated to be lived with one or more manifestation of cardiovascular disease [5]. International scientific meetings and high-impact journals [6] abound with presentations and publications on prognosis research. Variations in prognosis are increasingly the subject of debate for policy markers, health care professionals, third-party payers and patients. There is an abundance of novel putative prognostic factors, and a proliferation of risk prediction scores. Prognosis of certain conditions is particularly well established; for cancer, national survival by site, age, and year of diagnosis are freely available on the internet [7].

So why then is Dr. Lydgate still waiting for answers? The objective here is to discuss reasons for the possible discordance between the large amount of research activity and slow advances in prognostic understanding. Here it is argued that the potential of a concerted, coordinated research endeavor of cognate disciplines is yet to be realized, in part due to misconceptions of the scope of prognosis research. There are wide variations in prognostic outcomes (see Fig. 1), yet these remain poorly understood. Many arguments have been made that detract from the importance, or minimize the scope, of prognosis research and support the status quo. A polemic is offered, giving the counterperspective. It would be hard to find an individual who was a proponent of all these shibboleths. But to this person, should they exist, and to all others an invitation to debate is extended.

Section snippets

Medical training does not need to train doctors in prognosis

Doctors throughout the world qualify from medical school with scarcely a single seminar, lecture, or practical devoted to prognosis. Prognosis is now virtually extinct as a discrete classroom subject in medical schools. Yet prognosis research is intrinsically related to the practice of clinical medicine—most doctors collect data on patients at one time point and observe them over follow-up for health outcomes. These observations are inchoate prognostic studies. By contrast, randomization in

Prognosis research is a subject without a discipline

False. Prognosis research is necessarily multidisciplinary, including epidemiologists, clinicians, health services researchers, basic biologic scientists, geneticists, social scientists, psychologists, and health economists, among others. However, clinical epidemiology is primus inter pares among the disciplines in prognosis research [23], [24], [25], [26], [27], [28].

Prognosis research is an oxymoron

Many influences on prognosis are contingent on time and place, threatening the generalizability—and hence scientific validity—of

Conclusion

Progress in prognostic understanding can be gauged using measures such as the amount of variation in prognostic outcomes explained, discovery, and establishment of new prognostic factors, or moving to the top left-hand corner of the ROC curve. Integral to achieving advances across the broad scope of prognosis research is the need to consider the interrelated perspectives of the person experiencing disease, through the clinician, basic scientist, health care policy maker, to the public policy

Acknowledgments

H.H. is supported by a public health career scientist award from the Department of Health.

References (81)

  • P.M. Rothwell

    External validity of randomised controlled trials: “To whom do the results of this trial apply?”

    Lancet

    (2005)
  • P.A. Poole-Wilson et al.

    Effect of long-acting nifedipine on mortality and cardiovascular morbidity in patients with stable angina requiring treatment (ACTION trial): randomised controlled trial

    Lancet

    (2004)
  • S. Philpott et al.

    Gender differences in descriptions of angina symptoms and health problems immediately prior to angiography: the ACRE study. Appropriateness of Coronary Revascularisation study

    Soc Sci Med

    (2001)
  • E. Nikolsky et al.

    Impact of anemia in patients with acute myocardial infarction undergoing primary percutaneous coronary intervention: Analysis from the controlled abciximab and device investigation to lower late angioplasty complications (cadillac) trial

    J Am Coll Cardiol

    (2004)
  • E.E. Ntzani et al.

    Predictive ability of DNA microarrays for cancer outcomes and correlates: an empirical assessment

    Lancet

    (2003)
  • J.P. Ioannidis et al.

    Genetic associations in large versus small studies: an empirical assessment

    Lancet

    (2003)
  • G.L. Grunkemeier et al.

    Receiver operating characteristic curve analysis of clinical risk models

    Ann Thorac Surg

    (2001)
  • H.L. Elsner

    The prognosis of internal diseases

    (1923)
  • Mackenzie J. Angina pectoris. London: Oxford Medical Publications;...
  • A. Peeters et al.

    A cardiovascular life history. A life course analysis of the original Framingham Heart Study cohort

    Eur Heart J

    (2002)
  • SEER Cancer statistics review, 1975–2001

    (2004)
  • D.A. Warrell et al.

    Oxford textbook of medicine

    (2003)
  • Anon

    Clinical evidence: the international source of the best available evidence for effective healthcare

    (2004)
  • M.K. Gospodarawicz et al.

    Prognostic factors in cancer

    (2001)
  • J.M. Gilchrist

    Prognosis in neurology

    (2000)
  • P. Jallon

    Prognosis of epilepsies

    (2003)
  • R.W. EVANS et al.

    Prognosis of neurological disorders

    (2000)
  • A.S. Kapoor et al.

    Prognosis and risk assessment in cardiovascular disease

    (1993)
  • J.A. Knottnerus

    Challenges in dia-prognostic research

    J Epidemiol Community Health

    (2002)
  • R.F. Nease et al.

    Variation in patient utilities for outcomes of the management of chronic stable angina. Implications for clinical practice guidelines. Ischemic Heart Disease Patient Outcomes Research Team

    JAMA

    (1995)
  • J. Hilden et al.

    Prognosis in medicine: an analysis of its meaning and roles

    Theor Med

    (1987)
  • C.M. Otto et al.

    The Cardiovascular HS. Association of aortic-valve sclerosis with cardiovascular mortality and morbidity in the elderly

    N Engl J Med

    (1999)
  • T.R. Golub et al.

    Molecular classification of cancer: class discovery and class prediction by gene expression monitoring

    Science

    (1999)
  • A. Edwards et al.

    Explaining risks: turning numerical data into meaningful pictures

    BMJ

    (2002)
  • E.C. Schneider et al.

    Use of public performance reports: a survey of patients undergoing cardiac surgery

    JAMA

    (1998)
  • J.R. Paul

    Clinical epidemiology

    (1958)
  • A.R. Feinstein

    Clinical judgement

    (1967)
  • D.R. Peterson

    Epidemiology and clinical problems: with coordinated readings

    (1973)
  • R.H. Fletcher et al.

    Clinical epidemiology: the essentials

    (1984)
  • A.R. Feinstein

    Clinical epidemiology: the architecture of clinical research

    (1985)
  • Cited by (49)

    • Prognostication in neurology

      2022, Handbook of Clinical Neurology
      Citation Excerpt :

      Patients flow through a pyramidal health care system ranging from primary care to specialized centers. Prognosis severity is different at each level of the health care pyramid, usually worse at the apex (Hemingway, 2006). Studies with patients exclusively derived from specialized centers are susceptible to a referral bias.

    • How to conduct a high-quality original study on a prognostic research topic

      2018, Surgical Oncology
      Citation Excerpt :

      In medicine, prognosis is forecasting the likelihood of a future health state of a patient based on his or her clinical and non-clinical characteristics. An accurate prediction of prognosis may lead to more effective, efficient and targeted care options that can overcome or optimize characteristics that influence future health care states [3–5]. Within a cancer context, for instance, people not only hope to live longer, but also to live with a high-quality of life.

    • Prognosis as an Intervention

      2018, Clinics in Perinatology
      Citation Excerpt :

      Moreover, it is not always clear how to overcome such issues. Despite the improvement of scientific methods in several areas of medicine during recent decades, such as with randomized controlled trials for neonatal and perinatal therapies, some investigators have argued that methods to support prognosis research remain poorly implemented and understood.30,31 Nonetheless, producing an evidence-based prognosis is worth striving for.

    • Moderate efficiency of clinicians' predictions decreased for blurred clinical conditions and benefits from the use of BRASS index. A longitudinal study on geriatric patients' outcomes

      2016, Journal of Clinical Epidemiology
      Citation Excerpt :

      This could be a valid aid for complex clinical situations, where clinical judgment might not be accurate enough. Although epidemiological research (as in the Perdove-Anziani study) is more an etiological than a prognostic discipline [56], our data suggest that more high-quality research is required in this direction [8]. This could optimize health service use and functioning and improve clinicians' predictive skills.

    • Development and validation of an epidemiologic case definition of epilepsy for use with routinely collected Australian health data

      2015, Epilepsy and Behavior
      Citation Excerpt :

      Therefore, in developing our model, we included codes for epilepsy (G40) and status epilepticus (G41) and added at least AED monotherapy. This improves the accuracy of new-onset epilepsy ascertainment and minimizes some of the selection bias from hospital-based sampling [33]. However, to distinguish incident from prevalent cases, an important issue in prognosis and comorbidity linkage studies, cross-linking with AED prescriptions needs to be comprehensive, occur early after treatment initiation and have a sufficient treatment-free period prior.

    View all citing articles on Scopus
    View full text