Mammographic screening: no reliable supporting evidence?

Mammographic screening: no reliable supporting evidence?

CORRESPONDENCE 39% disease rate was reported in women older than 40 years, none of whom died from breast cancer.4 Therefore, a proportion of individu...

58KB Sizes 0 Downloads 18 Views

CORRESPONDENCE

39% disease rate was reported in women older than 40 years, none of whom died from breast cancer.4 Therefore, a proportion of individuals who have breast cancer detected by screening will never die from the disease. This theory is supported by the necropsy results from the four Swedish mammography screening trials.5 Only 58% of all deceased breast-cancer patients included in one of those studies died from breast cancer. A population of women older than 55 years includes individuals who will die within 8–11 years, irrespective of whether breast cancer is diagnosed as a concomitant disease—ie, the 8–11-year life expectancy of the subgroup analysed by Miettinen and colleagues is less than 100%. These patients will never experience any delayed benefit. An assumption of breast cancer leading to death in every patient and the assumption that every woman of age 55 years not having breast cancer will survive 8–11 years might lead to an overestimation of the mortality reducing effect of mammography screening. Andreas du Bois Comprehensive Breast Care Centre Wiesbaden, Department of Gynaecology and Gynaecological Oncology, Dr Horst Schmidt Kliniken, 65199 Wiesbaden, Germany (e-mail: [email protected]) 1

2

3

4

5

Miettinen OS, Henschke CI, Pasmantier MW, Smith JP, Libby DM, Yankelevitz DF. Mammographic screening: no reliable supporting evidence? Lancet 2002; 359: 404–06. Olsen O, Gøtzsche PC. Cochrane review of screening for breast cancer with mammography. Lancet 2001; 358: 1340–42. Andersson I, Aspegren K, Janzon L, et al. Mammography screening and mortality from breast cancer: the Malmö mammographic screening trial. BMJ 1988; 297: 943–48. Nielsen M, Thomsen JL, Primdahl S, Dyreborg U, Anderson JA. Breast cancer and atypia among young and middle-aged women: a study of 110 medicolegal autopsies. Br J Cancer 1987; 56: 814–19. Lindgren A. Autopsy and cause of death in randomized mammography studies. Qual Assur Health Care 1993; 5: 303–07.

Sir—Olli Miettinen and colleagues1 are right to conclude that, in the pursuit of reliable evidence on the usefulness of screening mammography, the use of wrong measures of effectiveness is misleading. They inadvertently illustrate this conclusion by introducing a misleading measure, the proportional reduction of the conditional casefatality rate. The aim of breast cancer screening is to reduce breast cancer mortality by early diagnosis and treatment. Since diagnosis and treatment are advanced in time (and age), early morbidity and

720

early mortality in the screened group are part and parcel of the intervention. Astonishingly, Miettinen and colleagues themselves have noted that during the necessary delay an increase in mortality can be expected.2 They redefine the case-fatality rate as a measure conditional on survival for an extended period (chosen akin to the method of the Texas sharpshooter: draw a target around the bullet hole). If screening increases the risk of death among breast cancer patients in the early period, the conditional casefatality rate in the later period will decrease automatically; you can die but once. In the control group there were 25 deaths in the first 6 years, in the screening group of identical size there were 35 deaths in that same period.2 If no effect whatsoever is assumed, in the remaining years there will be ten fewer deaths in the screened group. The conditional case-fatality rate decreases, but the screened group will only lose life-years because of earlier mortality. The fundamental measure of outcome of an intervention is its reduction in absolute risk over a stated period.3–5 There were 84 deaths in the control group of 21 142 women in the first 11 years. Whatever way these numbers are turned, in the screened population of the same size there were 15 fewer deaths, or 21 in Miettinen and colleagues’ opportunistic time window (8–11 years). The more than 24 120 women who did not benefit might have some policy interest too. Cancer risk management should learn from cardiovascular risk management, for which the ruling paradigm is to identify groups at higher absolute risk.4 In the Malmö trial, the risk of death from breast cancer was 0·4% over 11 years—excessively small compared with the vantage point of cardiovascular risk management. Relative measures of effect confuse this issue central to health policy: at what levels of absolute risk and absolute risk reduction are interventions among healthy people justified? Luc Bonneux Julius Centre for Health Sciences and Primary Care, HP D 01.335, PO Box 85500, 3508 GA Utrecht, Netherlands (e-mail: [email protected])

1

2

Miettinen OS, Henschke CI, Pasmantier MW, Smith JP, Libby DM, Yankelevitz DF. Mammographic screening: no reliable supporting evidence? Lancet 2002; 359: 404–06. Miettinen OS, Henschke CI, Pasmantier MW, Smith JP, Libby DM, Yankelevitz DF. Mammographic screening: no reliable supporting evidence? http://image.thelancet.com/extras/1093web. pdf (accessed April 9, 2002).

3

4

5

Cook DJ, Sackett DL. The number needed to treat: a clinically useful measure of treatment effect. BMJ 1995; 310: 452–54. Jackson R. Guidelines on preventing cardiovascular disease in clinical practice. BMJ 2000; 320: 659–61. Lubsen J, Hoes A, Grobbee D. Implications of trial results: the potentially misleading notions of number needed to treat and average duration of life gained. Lancet 2000; 356: 1757–59.

Sir—Olli Miettinen and colleagues1 cogently discuss the low mortality ratios in years 6–10 after entry into the Malmö trial,2 but say little about the high ratios in years 2–4. I believe the latter might have important implications. I suggest, as a possible explanation for these high ratios, that screening may have identified advanced breast cancers that would not have been detected as early in the unscreened cohort. Therefore, certain treatments of advanced breast cancers might promote systemic disease progression by some unknown mechanisms, thus explaining the high mortality rate ratios in years 2–4. If this hypothesis were supported by a more detailed analysis of the early years of the Malmö cohorts, reassessment of the assumed benefits of treating advanced breast cancers through various interventions might be needed. This effect might also complicate the interpretation of the low mortality rate ratios seen in years 6–10, which might be due to a combination of causes: prolonged life by finding and treating early-stage tumours and shortened life (to years 2–4) from treating later-stage tumours among the screened. Much could be learned from a closer look at the Malmö data. John Werth 204 Bertrand Drive, Princeton, NJ 08540, USA 1

2

Miettinen OS, Henschke CI, Pasmantier MW, Smith JP, Libby DM, Yankelevitz DF. Mammographic screening: no reliable supporting evidence? Lancet 2002; 359: 404–06. Anderson I, Aspegren K, Janzon L, et al. Mammographic screening and mortality from breast cancer: the Malmö mammographic screening trial. BMJ 1988; 297: 943–48.

Authors’ reply Sir—Andreas du Bois attributes to us the amazing idea that breast cancer leads inevitably to death (100% mortality), unless it is detected by mammographic screening and cured. We actually stated that any cancer’s case-fatality rate in the absence of curative treatment would be 100% if there were no role for other causes of death. In the complete version of the report on the web, we added that even

THE LANCET • Vol 360 • August 31, 2002 • www.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.

CORRESPONDENCE

in the absence of screening the ase-fatality rate of breast cancer is as low as about 30%. This addition was, however, followed by a gratuitous falsehood—irrelevant to our arguments—about the relative roles of cures and deaths from unrelated causes in making the no-screening case-fatality rate that low. Just as amazingly, Luc Bonneux claims we redefined the concept of case-fatality rate. With no redefinition, we merely estimated the case-fatality rate on the premise that the early excess in breast-cancer mortality is negligible relative to the deficit that continually prevails after the requisite delay, given continuing screening. Bonneux seems to think that the early increase in mortality is appreciable, and that it leads to an appreciable reduction in the likelihood of later death from breast cancer; John Werth seems to share this view. An appreciable early increase is not, however, suggested by the few early deaths in the Malmö trial, nor is it plausible without compelling evidence. Irrespective of this, the early increase is a consequence of early intervention and not a part and parcel of it. Bonneux also does not appreciate that the early excess is limited in its duration, whereas the ensuing reduced risk is continual, given continued screening. Nor does he seem to appreciate that experimental screening in trials is generally stopped much earlier than screening in actual practice. Bonneux demands research focus on absolute measures of consequence, but he does not distinguish between the appropriate focus in research and that in practice. Research focus on relative proportional reduction in case-fatality rate is appropriate because this measure can reasonably be presumed to be almost constant over varying degrees of background case-fatality rate, prevailing in the absence of screening. In practice, then, the general proportional reduction in case-fatality rate is coupled with the ad hoc background case-fatality rate to derive the ad hoc difference (absolute) in it. This difference, together with other inputs, translates to the ad hoc value for whatever absolute measure Bonneux actually advocates, and to ones that are genuinely meaningful. The realistically hoped-for consequence of cancer screening on causespecific death risk is substantially a function of time since the start of continuing screening, and, therefore, the need is to study this temporal pattern, especially the asymptote of the mortality-density ratio, which in screening compared with no-screening

trials requires duration.

screening

of

long

Olli S Miettinen, *Claudia Henschke Departments of Epidemiology and Biostatistics, and Medicine, Faculty of Medicine, McGill University, Montreal, Canada; and *Department of Radiology, Weill Medical College of Cornell University, New York, NY 10021, USA (e-mail: [email protected])

Schizophrenia and velocardio-facial syndrome Sir—One of the valuable lessons from Kieran Murphy’s (Feb 2, p 426)1 review of schizophrenia and velo-cardio-facial syndrome (VCFS) is that minor physical anomalies may represent the final common pathway for various pathological processes, involving differing levels of genetic and environmental inputs. Minor physical anomalies are minor dysmorphic features that are generally of no medical or cosmetic importance but serve as a useful index of neural maldevelopment. Recognition of the prevalence of minor physical anomalies in schizophrenia is increasing.2 Moreover, there is substantial overlap between the minor physical anomalies in schizophrenia and those in VCFS.3 Minor physical anomalies in schizophrenia are probably related to environmental insult between weeks 8 and 22 of gestation, whereas in VCFS the pathology is more clearly determined genetically. The relative weighting of gene-environment interactions may explain why VCFS patients are at increased risk of schizophrenia—their genetic predisposition to the disorder is raised, as evidenced by their high degree of characteristic minor physical anomalies, so that a lower level of additional environmental insult is required to trigger a first episode of psychosis. The prognostic implications of the relation between VCFS and schizophrenia are equally important. There is substantial evidence to support the existence of a neurodevelopmental subtype of schizophrenia.4 In this context, the neurobiological links between VCFS and schizophrenia provide a credible underpinning for the findings that neurodevelopmental schizophrenia seems to have an especially poor prognosis and that schizophrenia in combination with VCFS presents a particular therapeutic challenge, above and beyond that reported with other comorbidities. The coexistence of these two disorders may reflect a highly neurodevelopmentally weighted process, with a proportionally negative impact on prognosis. Finally, biological psychiatry notwithstanding, it is worth remembering that

THE LANCET • Vol 360 • August 31, 2002 • www.thelancet.com

the paradigm that best explains the pattern of relapse and long-term outcome in schizophrenia is still a substantially psychological one. Relapse is related to levels of expressed emotion and adherence to treatment, which is in turn closely related to insight, another psychological construct. Furthermore, the raised risk of schizophrenia among migrants shows a powerful inverse relation with the size of the migrant group, a finding that, at the present stage in the evolution of biological psychiatry, lends itself more readily to social or psychological explanations than to biological ones.5 These social and psychological factors are very probably caused or mediated by biological factors that we have yet to identify in schizophrenia. There is much to be done. Brendan D Kelly Stanley Research Centre (Ireland), Cluain Mhuire Family Centre, Blackrock, Co Dublin, Ireland (e-mail. [email protected]) 1 2

3

4

5

Murphy KC. Schizophrenia and velo-cardiofacial syndrome. Lancet 2002; 359: 426–30. Lane A, Kinsella A, Murphy P, et al. The anthropometric assessment of dysmorphic features in schizophrenia as an index of its developmental origins. Psychol Med 1997; 27: 1155–64. Murphy KC, Owen MJ. Velo-cardio-facial syndrome: a model for understanding the genetics and pathogenesis of schiophrenia. Br J Psychiatry 2001; 179: 397–402. Murray RM, O’Callaghan E, Castle DJ, et al. A neurodevelopmental approach to the classification of schizophrenia. Schizophr Bull 1992; 18: 319–32. Boydell J, van Os J, Mckenzie K, et al. Incidence of schizophrenia in ethnic minorities in London: ecological study into interactions with environment. BMJ 2001; 323: 1336–38.

Sir—Kieran Murphy1 describes VCFS as a powerful model for identifying susceptibility genes for schizophrenia. Molecular studies now suggest common genetic liability between schizophrenia and bipolar (affective) disorder at several susceptibility loci, including the chromosome 22q region implicated in VCFS.2 Although modern diagnostic systems differentiate between these disorders, psychotic symptoms (delusions and hallucinations) are common to both, and a subgroup of patients who meet criteria for both disorders are diagnosed as having schizoaffective disorder. Since the two most common functional psychotic disorders, schizophrenia and bipolar disorder, share common epidemiological features, risk factors, and response to antipsychotic treatment. Differences between them detected by imaging or neuropathology are qualitative rather than quantitative. As an alternative, a continuum of psychosis has been proposed, in which patients described as having bipolar

721

For personal use. Only reproduce with permission from The Lancet Publishing Group.