The case-control design and the assessment of the efficacy of cancer screening

Case-control studies have been used in recent years to evaluate the efficacy of cancer screening. However, relatively little work has been done to examine the methodology itself for this purpose. In this paper, it is demonstrated that because of self-selection bias the case-control study can yield a biased estimate of screening efficacy. Further, it is shown how this bias can be assessed using data from a randomized trial. Using data from the HIP breast cancer screening study, the magnitude of the self-selection bias is estimated and is seen to be substantial.     

Case-control prostate cancer screening studies should not exclude subjects with lower urinary tract symptoms

OBJECTIVE: Prostate-specific antigen (PSA)/digital rectal exam (DRE) screening for prostate cancer has become standard medical practice; however, its effectiveness in terms of reducing prostate cancer mortality remains undetermined. Case-control screening studies may help determine screening efficacy, though the proper disposition of symptomatic subjects is unclear. This paper presents a prostate cancer-specific methodological modification for analyzing symptomatic case-control screening subjects. METHODS: Prostate cancer detection studies and case-control studies of PSA/DRE screening were reviewed, and the results for symptomatic and asymptomatic subjects were compared. RESULTS: Most PSA/DRE detection studies have found that the prostate cancer detection rate among symptomatic patients is the same as or lower than that among asymptomatic patients. Lower urinary tract symptoms (LUTS), often referred to as early prostate cancer symptoms, occur more often in benign prostatic hyperplasia (BPH), a more commonly diagnosed, nonmalignant disease. Screened symptomatic subjects are usually removed from the "screened" category in case-control studies even though BPH-related symptoms do not confer increased prostate cancer risk and odds ratios do not change with inclusion of symptomatic subjects in the analysis. CONCLUSION: Screened subjects with LUTS should remain in the "screened" category in case-control prostate cancer screening studies since these symptoms may not be associated with increased risk of prostate cancer or validity of the odds ratio.     

Case-control studies of the efficacy of cancer screening: overcoming bias from nonrandom patterns of screening

A case-control study of the efficacy of cancer screening, like any other case-control study, must deal with potential confounding. There are 2 categories of confounding variables that pose special problems for studies of screening: (1) age and calendar time resulting from different temporal distributions of screening between cases and controls irrespective of whether the screening test leads to a reduction in mortality; and (2) the administration of other screening tests for the cancer in question when it is not clear whether the result of the other test had a bearing on the decision to order the test under study. We describe circumstances in which confounding from these sources can be dealt with satisfactorily by means of restriction or adjustment, and other circumstances in which it cannot.     

Measuring screening intensity in case-control studies of the efficacy of mammography

Of great interest in studies of screening for breast cancer is the relative efficacy of different screening frequencies (intensities). Prior work has suggested that estimates of the association between screening intensity and outcome in case-control studies would not produce valid results and that only binary indicators (no screens vs. one or more) of exposure can be used. Using case-control studies drawn from simulated cohorts of 30,000-40,000 women, the authors found that biases demonstrated in prior studies can be explained by 1) misclassification of true exposure groups by observed screening history, and 2) differential exposure misclassification of cases and controls. Binary as well as ordered categorical and interval measures can be biased unless they account for misclassification. By combining measurements of screening history from multiple periods of observation of varying lengths and using repeated-measures logistic regression models, the effect of screening intensity can be estimated in the presence of misclassification. Assessing the effect of screening intensity in case-control studies of mammography is possible if principles and methods for misclassification and measurement error guide the analysis.     

Prevention of colorectal cancer: guidelines based on new data. WHO Collaborating Center for the Prevention of Colorectal Cancer.

Recently published good quality data are the basis for this update. The newly reported studies include randomized trials, non-randomized cohort studies, and case-control studies; some of the data had mortality reduction as the endpoint. These guidelines, which were developed by the WHO Collaborating Center for the Prevention of Colorectal Cancer at Memorial Sloan-Kettering Cancer Center in conjunction with an International Advisory Committee, include primary prevention, screening of average-risk individuals, screening of individuals with heritable factors for colorectal cancer, surveillance of patients with colorectal polyps, and surveillance of patients with chronic ulcerative colitis. A list of papers reviewed for this update are cited, including recently published trials evaluating faecal occult-blood testing, case-control studies of sigmoidoscopy, the National Polyp study, and familial colon cancer studies. These guidelines will help inform patients and guide physicians in their approach to the prevention of colorectal cancer.     

Validation of self-reported screening mammography histories among women with and without breast cancer

As part of a case-control study of the efficacy of screening mammography, the authors validated the mammography histories of 2,495 women aged 40-64 years with incident breast cancer diagnosed in 1994-1998 and a 25% random sample of 615 controls never diagnosed with breast cancer, all reporting a mammogram in the past 5 years. Subjects from five metropolitan areas of the United States were cross-classified by facility records ("gold standard") and self-report according to history of a recent screening mammogram (within 1 year or within 2 years). Sensitivity and specificity of self-reported screening at 1 year were 0.93 and 0.82, respectively, for cases and 0.92 and 0.80 for controls. At 2 years, sensitivity and specificity were 0.97 and 0.78 for both cases and controls. Confidence intervals for the differences in sensitivity and specificity were narrow and included zero. Scant evidence was found of telescoping (recollection of events as more recent than actual). Findings suggest that, in an interview-based case-control study of the efficacy of screening mammography, 1) estimated true prevalences of recent screening mammography adjusted for sensitivity and specificity will be slightly lower than self-reported prevalences, and 2) differential misclassification of exposure status is slight. Therefore, odds ratios will likely be biased toward the null, underestimating screening efficacy.     

Systematic review of the epidemiologic and trial evidence of an association between antidepressant medication and breast cancer

Biologic studies have suggested that antidepressant use may increase breast cancer risk. We conducted a systematic review of trials and controlled epidemiologic studies to assess this association. Pooled data from 31 primary efficacy drug company trials of fluoxetine suggested no increased risk but the short duration of these trials may have been insufficient to detect an association. In one prospective cohort study antidepressant use was associated with breast cancer, but this study was conducted among women attending for breast screening, and only limited data on antidepressant use were available. In a second large prospective drugs screening study no association was found between either amitriptyline or imipramine and breast cancer. In a large well-conducted retrospective cohort study there was no association between antidepressant use and breast cancer. A second retrospective cohort study was flawed, with exposure in those who developed breast cancer being measured over a shorter time period than in those who remained disease free. Two of four case-control studies found no association between antidepressant use and breast cancer after control for a number of potential confounding factors. We conclude that epidemiologic evidence does not support an association between antidepressant use and breast cancer.     

Case-control studies of screening

The application of case-control studies to the evaluation of screening data presents a number of potential problems. In particular, the results are liable to be affected by selection bias. Published case-control studies of breast cancer screening are reviewed in the light of the results of a recent UK study of somewhat different design which attempts to estimate the effect of selection bias. Case-control studies of cervical screening involve additional difficulties, such as the handling of cases of invasive disease diagnosed by screening, and the distinction between diagnostic and screening smears. The extent to which published studies have addressed these is discussed. It is concluded that case-control studies of screening should be interpreted with caution, particularly in situations where comparable data from controlled trials are not available.     

Investigation of design and bias issues in case-control studies of cancer screening using microsimulation

Using a microsimulation approach, the authors examined design and bias issues in case-control studies of cancer screening. Specifically, they looked at the impact on the odds ratio of the way in which exposure to screening is defined, the type of age matching, the time scale used, and the criteria used to determine control eligibility. The results showed that defining exposure as "ever/never" screened produced, as expected, a serious bias in favor of screening. Defining exposure as being screened no later than the time the case's cancer is diagnosed has a serious bias against screening. An alternative exposure definition--screening can occur no later than the time the case would have been clinically diagnosed--eliminates the bias against screening. Further, the results showed that the type of age matching and the time scale used can produce a bias against screening and that this bias can be quite strong when case-control studies are performed in populations with a periodic screening program that is the only source of screening. Finally, control eligibility criteria had little effect.     

A case-control evaluation of the effect of breast cancer screening in the United Kingdom trial of early detection of breast cancer.

OBJECTIVE--The aim was to assess the extent to which selection bias affects a case-control study of breast cancer screening in which attenders and non-attenders for screening are compared. DESIGN--There were two retrospective case-control studies, one estimating the risk of death from breast cancer in women in the screening district relative to those in the comparison district (study A), the second estimating the relative risk for women who had ever been screened compared with women who had never been screened in the screening district alone (study B). For cases and controls in study B, the women's screening history was summarised for the time period from date of entry to diagnosis of the case, or the equivalent time from date of entry for the matched controls. For cases detected by screening, the screen at which cancer was detected was included in the screening history. SUBJECTS--Cases were deaths from breast cancer in women with disease diagnosed after entry to the trial, up to 31 December 1986 or a maximum of seven years from date of entry, in one of the screening districts (Guildford) and one of the comparison districts (Stoke) participating in the UK Trial of Early Detection of Breast Cancer: study A: 198 deaths in Guildford and Stoke; study B: 51 deaths in Guildford only. There were five age matched controls for each case, with length of follow up at least as great as the time from entry to death of the case. MAIN RESULTS--The estimate of the risk of death from breast cancer in the screening district relative to the comparison district from study A was 0.76, thus implying a reduction of 24% in the screening district, similar to that obtained from a cohort analysis of data from the two districts. In contrast, the relative risk in study B for ever v never screened women was 0.51, which, taking the 72% compliance into account, would result in a relative risk of 0.65 for the screening district if there were no selection bias. The risk of breast cancer mortality in the never screened relative to the comparison district was 1.13, despite the fact that incidence rates in the two populations were similar. This suggested that cancers in the never screened group had a particularly poor prognosis, contributing to selection bias. CONCLUSIONS--The possible existence of selection bias should lead to caution in interpretation of the results of case-control studies of the effect of breast cancer screening on mortality.     

Nonexperimental evaluation of the effectiveness of a screening program for lung cancer.

The effectiveness of screening to control lung cancer was examined in the German Democratic Republic by analyzing data from a cancer registry and incidence and mortality rates for lung cancer relative to different screening policies, and by two case-control studies. Mortality from lung cancer did not appear to be affected by the screening programs studied. The high cost of mass screening, combined with uncertainty about the benefits of early treatment of lung cancer, outweigh the vague advantages of such screening.     

Validity of Outpatient Medical Records in Identifying Prostate-Specific Antigen (PSA) Tests Administered for Screening Purposes

A case-control study of cancer screening efficacy seeks to compare cases and controls for the receipt of only those tests done when there were no symptoms or signs of the cancer that is the target of the screening. We conducted a study to determine the accuracy of Kaiser Permanente Northwest (KPNW) medical records in identifying the reason(s) underlying testing for prostate-specific antigen (PSA). The authors reviewed the medical records and then conducted telephone interviews in 1996–1997 with 97 randomly-selected men whose PSA tests were categorized as screening according to the record. Subjects were asked why the tests were ordered and whether signs or symptoms of prostate cancer were present. Ninety-six men (agreement 99 percent, 95 percent confidence interval: 94, 100) reported no symptoms suggestive of prostate cancer. Our results suggest that the outpatient medical record is reliable when it indicates that a PSA test was done in a man with no symptoms of prostate cancer. Thus, a medical record-based study of the efficacy of PSA screening in this health plan should be biased to only a small degree as a result of underascertainment of symptoms as the reason(s) for receiving a PSA test. Presumably, these results could apply to studies in other health plans in which a similar level of detail on reason for screening is present in the medical record; further investigation of this issue is needed.     

Risk factors for breast cancer with applications to selection for the prevalence screen.

There have been many studies of individual risk factors for breast cancer; most of the factors concerned may be broadly grouped into demographic and dietary, reproductive history, endocrine related, family history of breast cancer, and previous history of breast disease. Some of these studies have examined the combined effect of these factors. The present case-control study does this in the context of a randomised controlled trial of breast cancer screening. The relative risks that we have obtained are, in general, of similar magnitude to those in other reports. The relevance of the results to a screening programme is discussed.     

Vaccination against human papillomavirus infection: a new paradigm in cervical cancer control

Universal deployment of organized or opportunistic screening with Pap cytology in high and middle income countries has been the primary reason for the substantial reductions in cervical cancer morbidity and mortality during the last 50 years. However, in many low income countries Pap cytology screening is yet to be effectively implemented or has failed to reduce cervical cancer rates to an appreciable extent. Cervical cancer thus remains a critical public health problem that is second only to breast cancer in overall disease burden for women throughout the world. The fact that infection with certain human papillomavirus (HPV) types is now recognized as a necessary cause of this disease has led to new research fronts on the prevention of cervical cancer. Recent research on the safety and efficacy of candidate prophylactic vaccines against HPV have shown very promising results with nearly 100% efficacy in preventing the development of persistent infections and cervical precancerous lesions. Ongoing clinical studies are expected to provide further evidence of efficacy and will form the basis for licensing of candidate vaccines by the major pharmaceutical companies within 3-6 years. Although the future seems bright on the HPV vaccine front policy makers are strongly cautioned to avoid scaling back cervical cancer screening. It will take many years before we can rationally develop cervical cancer screening strategies that will be cost-effective for the proper surveillance of women protected by HPV vaccination.     

The role of clinical breast examination and breast self-examination

The efficacy of screening by clinical breast examination (CBE) and/or breast self-examination (BSE) is reviewed using indirect evidence from randomized breast screening trials and that from observational studies. In countries where breast cancer is diagnosed at an advanced stage, screening by CBE with the teaching of BSE as an integral component will probably be effective in reducing breast cancer mortality. However, in technically advanced countries where adequate treatment is given, no screening modality is likely to be sufficiently beneficial to outweigh the harms of screening, especially false positives and over-diagnosis.     

Case-control evaluation of breast cancer screening efficacy

The feasibility and validity of using case-control methods to evaluate the efficacy of breast cancer screening were investigated using data from the Health Insurance Plan of Greater New York Breast Cancer Screening Trial. Women who died of breast cancer were compared with individually matched controls of the same age who had equal or greater survival times. To minimize the effects of self-selection bias, most analyses focused on the 95 cases and 380 matched controls who had been screened at least once and for whom covariate data were available. A statistically significant effect for being screened more than once versus being screened once, uncorrected for healthy-screenee bias, was found (odds ratio = 0.13, p less than 0.001). After correction for healthy-screenee bias, the effect was not statistically significant (odds ratio = 0.54, p = 0.15). Possible explanations for this statistical nonsignificance are: 1) correction for healthy-screenee bias entails loss of power and 2) the correction procedure used may, under some circumstances, result in bias toward the null hypothesis. Further research into means of eliminating healthy-screenee bias is needed to improve the applicability of case-control methods to the evaluation of screening efficacy.     

Diet and breast cancer

Obesity, overweight, and a sedentary lifestyle-all common conditions in breast cancer patients-are likely to be associated with poor survival and poor quality of life in women with breast cancer. Diet-related factors are thought to account for about 30% of cancers in developed countries. Most studies of diet and healthcare have focused on the role of single nutrients, foods, or food groups in disease prevention or promotion. Recent cancer guidelines on nutrition and physical activity emphasize diets that promote maintenance of a healthy body weight and a prudent dietary pattern that is low in red and processed meats and high in a variety of vegetables, fruits, and whole grains. Except for dietary fat, few nutritional factors in adult life have been associated with breast cancer. Extensive data from animal model research, international correlations linking fat intake and breast cancer rates, and case-control studies support the hypothesis that a high-fat diet is conducive to the development of breast cancer in postmenopausal women. Conflicting findings from cohort studies, however, have created uncertainty over the role of dietary fat in breast cancer growth and recurrence. Results from large-scale nutritional intervention trials are expected to resolve such issues. As new and improved data on dietary factors and patterns accumulate, dietary guidelines for cancer risk reduction will become more focused.     

Costs of breast cancer and the cost-effectiveness of breast cancer screening.

Breast cancer is the second leading cause of death by cancer among women in the United States. The total cost of illness for breast cancer has been estimated to be $3.8 billion, of which $1.8 billion represents medical care costs. It has been estimated that breast cancer detected early is considerably less expensive than when the tumor is discovered at a later stage. Mass screening using mammography can improve early detection by as much as 15-35%. Cost-effectiveness studies have estimated that the costs of breast cancer screening range between $13,200 and $28,000 per year of life saved. The cost-effectiveness of breast cancer screening in the 40-49-year-old age group is controversial. Mass screening for breast cancer will probably increase total health care costs, but when all economic costs are included, screening appears to be more cost-effective than not screening.     

A novel form of ascertainment bias in case-control studies of cancer screening.

In case-control studies of cancer screening, some have generally admonished investigators against case definitions based on diagnosis dates because of lead-time bias. However, perhaps partly due to vagueness, the admonitions have been frequently ignored. A recurrence-time model simulates case ascertainment when diagnosis must occur within a specific calendar period. The model depends on screening test sensitivity and rate, age-specific preclinical incidence rates, and preclinical duration time and survival time distributions. For one study of sigmoidoscopic screening for colorectal cancer, when the true odds ratio is 1, its estimate is 0.50 to 0.75 under plausible assumptions. This bias can affect any observational study wherein case definition depends on diagnosis times (e.g., health-plan enrollment data). To avoid bias in observational investigations of cancer screening wherein the case definition depends on the diagnosis date, one must ensure that both screening and preclinical incidence do not occur before the case definition period.     

Interview-based case-control studies of screening efficacy

Some studies of the efficacy of cancer screening are obliged to obtain interviews from persons with cancer who are likely to die as a result of their disease, and from controls. We review the interview-based approaches that have been developed and the purpose that each one can achieve and offer suggestions for increasing the validity of these studies.