1127
EDITORIALS
Should
we
case-control?
The case-control study is one of those methodological creeds that has enjoyed increasing popularity since the 1950s.1 In essence, it is merely the natural extension of that age-old cornerstone of medical progress, the case-series. Few things come more intuitively to a practising physician than to collect information on patients with a certain disease and to see whether these people have something strange in common in their antecedents. "Something strange" is a characteristic by which these individuals differ from others; "others" can be our armchair expectation of what is normal in our fellow men, or they can be formalised into a well-chosen control group that reflects how often the said characteristic might have been present in the absence of disease. These simple roots should not be forgotten in our admiration of tidy algebraic demonstrations that such studies can be construed as the looking-glass-world view of a follow-up study, however fruitful such a viewpoint might be
statistically. For such ease and simplicity one expects to pay a price. In rankings of methodological strength in books intended to engrain firm methodological truths in the mind of innocent physicians, the case-control study ranks second lowest5-it falls behind the randomised controlled trial and the prospective and retrospective follow-up study and barely overtakes the humble anecdote. The reason for this low esteem is the potential for bias that follows when an investigator has to choose a control group and to assess the antecedent characteristics after the disease has occurred. Exponents of this school of thought exhort young colleagues not to read anything on treatment efficacy, and to be wary of any claim of causality without evidence of proper randomisation, controls, and "blinding".6 A methodological leader of the profession who held 20 years ago that the case-control design was "simply illogical"’ later invented the term trohoc study, presumably to indicate that the casecontrol study was the literal inverse of a cohort study-ie, of a good study. Such ideas can have
far-reaching
consequences: some funding agencies will not countenance funding any studies that are not prospective. (Smart researchers promptly learned that the semantic confusion in epidemiology could be used to their advantage since a follow-up study may be
prospective, even if it describes a historical cohort.) Gradually some hardliners mellowed and deemed that a case-control study might be worthy of attention if certain very strict standards, mimicking those of randomised controlled trials, were adhered to.9 In a particular controversy-whether salicylates cause Reye syndrome-this attitude led to a rerun of a study that was already a much belaboured rerun, only to arrive at the same conclusion.lo,l1 This episode also led to some editorial frowning about methodological slandering between epidemiologists.12 Should our scepticism go as far as Lord Russell’s-ie, not believing anything unless we are forced to by the evidence?13 According to the philosopher this attitude would lead to a more tolerant, freer, and better world. Meanwhile, in the epidemiological community, called
tension has mounted. Protracted discussions about in case-control standards methodological investigations of postmenopausal oestrogen use and uterine cancer have now resulted in one group of epidemiologists making overt accusations against their conservative and over-sceptical brethren, in a vilifying paper with overtones of a religious war.14 The paper has drawn a response. 1-5 Nevertheless, the opportunity to wonder what the other side really has to say is dwindling fast. Epidemiology might lose sight of some reality. Impeccable methodology usually follows the development of new ideas with some delay. As a research cardiologist mused, real creativity in medicine lies in the hands of the world community of innovative investigators who together carry out hundreds of thousands of small research studies; confirmatory large randomised controlled trials become the task of bumt-out leaders who are only fit enough for administration and organisation.16 The lowliest forms of medical research, the case-history and case-series, have a long and distinguished record in medicine-in teaching and quality evaluation (the clinicopathological conference), as well as in new discoveries of aetiology, pathogenesis, and therapy. Long-term and short-term side-effects of drugs and many HLA-disease associations have first been
investigated in simple case-control studies; most new breakthroughs from the human genome front use material of one or a few patients and some expected DNA sequence. Thus the usual ranking of clinical
1128
research techniques according to strength might only be applicable for certain well-defined problems. How could a study on HLA or DNA patterns be randomised, the genetic dice having already been thrown by the patient’s parents? Even in studies of efficacy, one sees more and more case-controlling: examples include the efficacy of vaccines," screening, 18 and prophylaxis.l9 The precept to read only randomised controlled trials might make the reader miss all that is new, innovative, and exciting in medicine. Since the progress of science seldom follows a very orderly path2O it might be unwise to maintain the pretence of some universal rank order preference in theory. A scientific fact always needs to be incorporated into the existing scientific framework of the reader. 21,22 As already stated in the 1930s, even if a perfectly conducted randomised controlled trial yielded a result such as that some much diluted coloured water would cure typhoid, we would tend to dismiss the findings as an unlucky draw and remain faithful to mainline pharmacology.23 Conversely, some case-control studies on smoking and cancer of the 1950s-— poorly conducted by today’s standards-might ultimately change the face of our civilisation. So, should we case-control? The most universal answer might be that "it depends..." on the question addressed, the state of the problem, the speed of answer and accuracy needed, and some practical constraints. Thereby epidemiologists might concentrate on medical and biological substance, rather than on dismissing one another’s studies out-of-hand because of alleged methodological shortcomings.26 Nevertheless, we should remain critical and watchful, but not in terms of a priori convictions, however logical they might seem to the believer. We urgently need to analyse instances in which epidemiological methodology has failed, in either direction-by being too sceptical or too lenient,
by underemphasis or by overinterpretation. 1. Cole Ph. The evolving case-control study. In: Ibrahim MA, Spitzer WO, eds. The case-control study: consensus and controversy. Oxford: Pergamon, 1979: 15-34. 2. Cornfield J. A method of estimating comparative rates from clinical data. Applications to cancer of the lung, breast and cervix. J Natl Cancer Inst 1951; 11: 1269-75. 3. Mantel N, Haenszel W. Stastical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959; 22: 719-48. 4. Miettinen OS. Theoretical epidemiology: principles of occurrence research in medicine. New York: Wiley, 1985. 5. Sackett DL, Haynes RB, Tugwell P. Clinical epidemiology: a basic science for clinical research. Boston: Little, Brown, 1985. 6. Bennet KJ, Sackett DL, Haynes RB, et al. A controlled trial of teaching critical appraisal of the clinical literature to medical students. JAMA 1987; 257: 2451-54. 7. Feinstein AR. Clinical judgment. Baltimore: Williams and Wilkins, 1967. 8. Feinstein AR. Clinical biostatistics. St Louis: Mosby, 1977. 9. Feinstein AR. Clinical epidemiology: the architecture of clinical research. Philadelphia: Saunders, 1985. 10. Hurwitz ES, Barrett MJ, Bergman D, et al. Public Health Service study on Reye syndrome and medication. JAMA 1987; 257: 1905-11. 11. Forsyth BW, Horwitz RI, Acompora D, et al. New epidemiologic evidence confirming that bias does not explain the aspirin/Reye syndrome association. JAMA 1989; 261: 2517-24.
Reye syndrome, salicylates, epidemiology and public health policy. JAMA 1987; 257: 1941. 13. Russell B. Sceptical essays. London: Allan and Unwin, 1977. 14. Savitz DA, Greenland S, Stolley PD, Kelsey JL. Scientific standards of criticism: a reaction to "Scientific standards in epidemiologic studies of the menace of daily life" by AR Feinstein. Epidemiology 1990; 1: 78-82 15. Feinstein AR. Scientific news and epidemiologic editorials: a reply to the critics. Epidemiology 1990; 1: 170-80. 16. Rahimtoola SH. Control of scientific research. In: Brugada P, Wellens HJJ. Cardiac arrythmias: where to go from here? Mount Kisco, NY:
12. Mortimer EA.
Future, 1987: 689-98. PG, Rodriguez LC, Fine PEM. Assessment of the protective
17. Smith
of vaccines against common diseases using case-control and cohort studies. Int J Epidemiol 1984; 13: 87-93. 18. Morrison AS. Screening in chronic disease. New York: Oxford University Press, 1985. 19. Imperiale TF, Horwitz RI. Does prophylaxis prevent postdental infective endocarditis? A controlled evaluation of protective efficacy. Am J Med 1990; 88: 131-36. 20. Feyerabend P. Against method. London: Verso, 1980. 21. Cornfield J. Statistical relationships and proof in medicine. Am Statistician 1954; 8: 19-21. (Reprinted in Greenland S, ed. Evolution of epidemiologic ideas: annotated readings on concepts and methods. Chestnut Hill, MA: Epidemiology Resources, 1987.) 22. Cornfield J. Recent methodological contributions to clinical trials. Am J Epidemiol 1976; 104: 408-21. 23. Woods HM, Russel WT. An introduction to medical statistics. London: Staples Press, 1948. 24. Doll R, Hill AB. Smoking and carcinoma of the lung. Br Med J 1950; 2: 739-48. (Reprinted in Buck C, Llopis A, Nájera E, Terris M, eds. The challenge of epidemiology: issues and selected readings. Washington, DC: Pan American Health Organization, 1988.) 25. Wynder EL, Graham EA. Tobacco smoking as a possible etiologic factor in bronchiogenic carcinoma: a study of six hundred and eighty-four cases. JAMA 1950; 143: 329-36. (Reprinted in Buck C, Llopis A, Nájera E, Terris M, eds. The challenge of epidemiology: issues and selected readings. Washington, DC: Pan American Health
efficacy
26.
Organization, 1988.) Stolley PD. Faith, evidence and the epidemiologist. J Publ Hlth Policy 1985; 6: 37-42.
Soluble CD4: anothertherapeutic option in HIV infection Very high affinity binding of the gp 120 envelope glycoprotein of human immunodeficiency virus (HIV) to the CD4 surface antigen of T "helper" lymphocytes, macrophages, and related cells provides the main (although not the only) route of cell attachment and entry for the virus. Binding also disrupts the interaction of T-cell CD4 with major histocompatibility complex (MHC) class II determinants on antigen-presenting cells, an essential early step in the generation of a normal immune response.1-5 Furthermore, CD4-positive cells with bound viral gpl20 may be targets for destruction mediated by cytotoxic T cells or by antibodydependent cellular cytotoxicity (ADCC) .2,6Thus there are at least three mechanisms whereby binding of gpl20 to CD4 may contribute to the pathogenesis of AIDS-enhancement of spread of infection from cell to cell, impairment of immune recognition, and triggering of destruction of helper T cells. For these reasons the concept of a "decoy" in the form of CD4 molecules in solution in the plasma or carried, for example, on red cell membranes has its attractions as a means of diverting at least a proportion of HIV particles and of free gp 120 molecules away from more vulnerable sites.
EDITORIALS
Should
we
case-control?
The case-control study is one of those methodological creeds that has enjoyed increasing popularity since the 1950s.1 In essence, it is merely the natural extension of that age-old cornerstone of medical progress, the case-series. Few things come more intuitively to a practising physician than to collect information on patients with a certain disease and to see whether these people have something strange in common in their antecedents. "Something strange" is a characteristic by which these individuals differ from others; "others" can be our armchair expectation of what is normal in our fellow men, or they can be formalised into a well-chosen control group that reflects how often the said characteristic might have been present in the absence of disease. These simple roots should not be forgotten in our admiration of tidy algebraic demonstrations that such studies can be construed as the looking-glass-world view of a follow-up study, however fruitful such a viewpoint might be
statistically. For such ease and simplicity one expects to pay a price. In rankings of methodological strength in books intended to engrain firm methodological truths in the mind of innocent physicians, the case-control study ranks second lowest5-it falls behind the randomised controlled trial and the prospective and retrospective follow-up study and barely overtakes the humble anecdote. The reason for this low esteem is the potential for bias that follows when an investigator has to choose a control group and to assess the antecedent characteristics after the disease has occurred. Exponents of this school of thought exhort young colleagues not to read anything on treatment efficacy, and to be wary of any claim of causality without evidence of proper randomisation, controls, and "blinding".6 A methodological leader of the profession who held 20 years ago that the case-control design was "simply illogical"’ later invented the term trohoc study, presumably to indicate that the casecontrol study was the literal inverse of a cohort study-ie, of a good study. Such ideas can have
far-reaching
consequences: some funding agencies will not countenance funding any studies that are not prospective. (Smart researchers promptly learned that the semantic confusion in epidemiology could be used to their advantage since a follow-up study may be
prospective, even if it describes a historical cohort.) Gradually some hardliners mellowed and deemed that a case-control study might be worthy of attention if certain very strict standards, mimicking those of randomised controlled trials, were adhered to.9 In a particular controversy-whether salicylates cause Reye syndrome-this attitude led to a rerun of a study that was already a much belaboured rerun, only to arrive at the same conclusion.lo,l1 This episode also led to some editorial frowning about methodological slandering between epidemiologists.12 Should our scepticism go as far as Lord Russell’s-ie, not believing anything unless we are forced to by the evidence?13 According to the philosopher this attitude would lead to a more tolerant, freer, and better world. Meanwhile, in the epidemiological community, called
tension has mounted. Protracted discussions about in case-control standards methodological investigations of postmenopausal oestrogen use and uterine cancer have now resulted in one group of epidemiologists making overt accusations against their conservative and over-sceptical brethren, in a vilifying paper with overtones of a religious war.14 The paper has drawn a response. 1-5 Nevertheless, the opportunity to wonder what the other side really has to say is dwindling fast. Epidemiology might lose sight of some reality. Impeccable methodology usually follows the development of new ideas with some delay. As a research cardiologist mused, real creativity in medicine lies in the hands of the world community of innovative investigators who together carry out hundreds of thousands of small research studies; confirmatory large randomised controlled trials become the task of bumt-out leaders who are only fit enough for administration and organisation.16 The lowliest forms of medical research, the case-history and case-series, have a long and distinguished record in medicine-in teaching and quality evaluation (the clinicopathological conference), as well as in new discoveries of aetiology, pathogenesis, and therapy. Long-term and short-term side-effects of drugs and many HLA-disease associations have first been
investigated in simple case-control studies; most new breakthroughs from the human genome front use material of one or a few patients and some expected DNA sequence. Thus the usual ranking of clinical
1128
research techniques according to strength might only be applicable for certain well-defined problems. How could a study on HLA or DNA patterns be randomised, the genetic dice having already been thrown by the patient’s parents? Even in studies of efficacy, one sees more and more case-controlling: examples include the efficacy of vaccines," screening, 18 and prophylaxis.l9 The precept to read only randomised controlled trials might make the reader miss all that is new, innovative, and exciting in medicine. Since the progress of science seldom follows a very orderly path2O it might be unwise to maintain the pretence of some universal rank order preference in theory. A scientific fact always needs to be incorporated into the existing scientific framework of the reader. 21,22 As already stated in the 1930s, even if a perfectly conducted randomised controlled trial yielded a result such as that some much diluted coloured water would cure typhoid, we would tend to dismiss the findings as an unlucky draw and remain faithful to mainline pharmacology.23 Conversely, some case-control studies on smoking and cancer of the 1950s-— poorly conducted by today’s standards-might ultimately change the face of our civilisation. So, should we case-control? The most universal answer might be that "it depends..." on the question addressed, the state of the problem, the speed of answer and accuracy needed, and some practical constraints. Thereby epidemiologists might concentrate on medical and biological substance, rather than on dismissing one another’s studies out-of-hand because of alleged methodological shortcomings.26 Nevertheless, we should remain critical and watchful, but not in terms of a priori convictions, however logical they might seem to the believer. We urgently need to analyse instances in which epidemiological methodology has failed, in either direction-by being too sceptical or too lenient,
by underemphasis or by overinterpretation. 1. Cole Ph. The evolving case-control study. In: Ibrahim MA, Spitzer WO, eds. The case-control study: consensus and controversy. Oxford: Pergamon, 1979: 15-34. 2. Cornfield J. A method of estimating comparative rates from clinical data. Applications to cancer of the lung, breast and cervix. J Natl Cancer Inst 1951; 11: 1269-75. 3. Mantel N, Haenszel W. Stastical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959; 22: 719-48. 4. Miettinen OS. Theoretical epidemiology: principles of occurrence research in medicine. New York: Wiley, 1985. 5. Sackett DL, Haynes RB, Tugwell P. Clinical epidemiology: a basic science for clinical research. Boston: Little, Brown, 1985. 6. Bennet KJ, Sackett DL, Haynes RB, et al. A controlled trial of teaching critical appraisal of the clinical literature to medical students. JAMA 1987; 257: 2451-54. 7. Feinstein AR. Clinical judgment. Baltimore: Williams and Wilkins, 1967. 8. Feinstein AR. Clinical biostatistics. St Louis: Mosby, 1977. 9. Feinstein AR. Clinical epidemiology: the architecture of clinical research. Philadelphia: Saunders, 1985. 10. Hurwitz ES, Barrett MJ, Bergman D, et al. Public Health Service study on Reye syndrome and medication. JAMA 1987; 257: 1905-11. 11. Forsyth BW, Horwitz RI, Acompora D, et al. New epidemiologic evidence confirming that bias does not explain the aspirin/Reye syndrome association. JAMA 1989; 261: 2517-24.
Reye syndrome, salicylates, epidemiology and public health policy. JAMA 1987; 257: 1941. 13. Russell B. Sceptical essays. London: Allan and Unwin, 1977. 14. Savitz DA, Greenland S, Stolley PD, Kelsey JL. Scientific standards of criticism: a reaction to "Scientific standards in epidemiologic studies of the menace of daily life" by AR Feinstein. Epidemiology 1990; 1: 78-82 15. Feinstein AR. Scientific news and epidemiologic editorials: a reply to the critics. Epidemiology 1990; 1: 170-80. 16. Rahimtoola SH. Control of scientific research. In: Brugada P, Wellens HJJ. Cardiac arrythmias: where to go from here? Mount Kisco, NY:
12. Mortimer EA.
Future, 1987: 689-98. PG, Rodriguez LC, Fine PEM. Assessment of the protective
17. Smith
of vaccines against common diseases using case-control and cohort studies. Int J Epidemiol 1984; 13: 87-93. 18. Morrison AS. Screening in chronic disease. New York: Oxford University Press, 1985. 19. Imperiale TF, Horwitz RI. Does prophylaxis prevent postdental infective endocarditis? A controlled evaluation of protective efficacy. Am J Med 1990; 88: 131-36. 20. Feyerabend P. Against method. London: Verso, 1980. 21. Cornfield J. Statistical relationships and proof in medicine. Am Statistician 1954; 8: 19-21. (Reprinted in Greenland S, ed. Evolution of epidemiologic ideas: annotated readings on concepts and methods. Chestnut Hill, MA: Epidemiology Resources, 1987.) 22. Cornfield J. Recent methodological contributions to clinical trials. Am J Epidemiol 1976; 104: 408-21. 23. Woods HM, Russel WT. An introduction to medical statistics. London: Staples Press, 1948. 24. Doll R, Hill AB. Smoking and carcinoma of the lung. Br Med J 1950; 2: 739-48. (Reprinted in Buck C, Llopis A, Nájera E, Terris M, eds. The challenge of epidemiology: issues and selected readings. Washington, DC: Pan American Health Organization, 1988.) 25. Wynder EL, Graham EA. Tobacco smoking as a possible etiologic factor in bronchiogenic carcinoma: a study of six hundred and eighty-four cases. JAMA 1950; 143: 329-36. (Reprinted in Buck C, Llopis A, Nájera E, Terris M, eds. The challenge of epidemiology: issues and selected readings. Washington, DC: Pan American Health
efficacy
26.
Organization, 1988.) Stolley PD. Faith, evidence and the epidemiologist. J Publ Hlth Policy 1985; 6: 37-42.
Soluble CD4: anothertherapeutic option in HIV infection Very high affinity binding of the gp 120 envelope glycoprotein of human immunodeficiency virus (HIV) to the CD4 surface antigen of T "helper" lymphocytes, macrophages, and related cells provides the main (although not the only) route of cell attachment and entry for the virus. Binding also disrupts the interaction of T-cell CD4 with major histocompatibility complex (MHC) class II determinants on antigen-presenting cells, an essential early step in the generation of a normal immune response.1-5 Furthermore, CD4-positive cells with bound viral gpl20 may be targets for destruction mediated by cytotoxic T cells or by antibodydependent cellular cytotoxicity (ADCC) .2,6Thus there are at least three mechanisms whereby binding of gpl20 to CD4 may contribute to the pathogenesis of AIDS-enhancement of spread of infection from cell to cell, impairment of immune recognition, and triggering of destruction of helper T cells. For these reasons the concept of a "decoy" in the form of CD4 molecules in solution in the plasma or carried, for example, on red cell membranes has its attractions as a means of diverting at least a proportion of HIV particles and of free gp 120 molecules away from more vulnerable sites.
