Age-specific reduction in breast cancer mortality by screening: an analysis of the results of the Health Insurance Plan of Greater New York study.

The 14-year follow-up mortality results from the randomized breast cancer screening trial of the Health Insurance Plan of Greater New York (HIP) were analyzed with respect to the problem of age-specific screening effects. Mortality reduction was measured in three different ways and appears to be homogeneous across age groups. This finding challenges the widespread opinion that the results of the HIP study support the conclusion that breast cancer screening is not effective below age 50.

Model building on the basis of Dutch cervical cancer screening data.

A mass screening programme for cervical cancer is in progress in three pilot regions in The Netherlands. All women living in these regions aged 35-53 are invited to undergo screening at three-year intervals. The MISCAN simulation model was developed for the analysis and optimization of screening programmes. In this paper the model-based approach to evaluation is first outlined and then illustrated by analysing data from the first two screening rounds in the pilot regions. This analysis resulted in a rather restricted range of data-compatible assumptions for the mean duration of preclinical disease (14-19 yr) and the frequency of spontaneous regression of preinvasive lesions (45-65%), as well as a rather wide sensitivity range for the Pap smear (50-90%). These preliminary findings are compared with those of a previous MISCAN analysis of cervical cancer screening in British Columbia. On the basis of an assumed 18-yr duration, 50% regression and 70% sensitivity, a number of screening policies relating to the same age ranges but with different intervals are compared. Both the analysis and the policy comparisons are preliminary, but the findings are nevertheless reasonable and consistent with those of previous studies. A more complete MISCAN-based analysis of the Dutch screening programme and subsequent optimization of screening policies will be possible when further results become available and a cost-effectiveness analysis procedure has been incorporated into the MISCAN programme.

The MISCAN simulation program for the evaluation of screening for disease.

The computer program MISCAN is developed for use in evaluation of mass screening for disease. The program uses Monte Carlo simulation. It produces output on the results of screening projects and on the effects of screening on morbidity and mortality on the individual and population level. The calculations are based on models of the natural history of the disease and of the impact of screening on the natural history. The approach is such that considerable flexibility exists in specifying the structure of the model and its parameters. The program consists of two parts. The DISEASE part can be used for simulating the epidemiology of the disease when no screening is taking place; it requires input on the population and on the disease process. The SCREENING part is to be used in combination with the DISEASE part. It is intended for simulation of the results and effects of a screening project. It requires input on the properties of the screening tests, the consequences of early detection by screening, and the policy (ages and intervals between screens) of the project. MISCAN can be used for finding model assumptions regarding the disease process and the impact of screening that give a good explanation of the observed results of a screening project. Such an analysis proceeds in two steps. First, MISCAN is used to calculate simulated results of the project, based on specific assumptions. Next, these results are tested against the observed results, in order to assess the acceptability of the assumptions. MISCAN can also be used for optimization of the screening policy by simulating the cost and benefit components of a large number of different screening policies.

A model-based analysis of the HIP project for breast cancer screening.

A computer simulation approach is used to test assumptions about sensitivity of mammography and physical examination, and about the duration of preclinical screen-detectable breast cancer. Values between 50% and 80% for the combined sensitivity of the 2 tests give a good explanation of the results of the HIP randomized trial of breast cancer screening. The mean duration of the preclinical stage can vary from 1.6 years for high sensitivity values to 2.7 years for low values. In comparison with previous analyses of the HIP data our estimate for the sensitivity is lower, and the mean duration of the preclinical stage is longer. This is a consequence of the use of a more detailed model in our analysis, allowing for a more complete use of the HIP data in testing model assumptions. Similar analyses of data from recent screening projects in The Netherlands resulted in compatible estimates for the duration of preclinical breast cancer.

Preventive Pap-smears: balancing costs, risks and benefits.

The pattern of spontaneous screening for cervical cancer by general practitioners and gynaecologists in The Netherlands is compared with an efficient screening policy resulting from a cost-effective study. Spontaneous screening tends to start and stop too early in a woman's life, and leaves too many women overscreened or unprotected. The combination in young age of a low incidence of invasive cancer and a high incidence of regressive lesions explains relative ineffectiveness and harmfulness of present screening practice. When screening would take place between ages 30 and at least 60, with intervals of about 5 years, as many lives could be saved for half the costs and with only 60% of the unnecessary referrals and treatments. Much attention should be paid to the coverage of the target population. Therapeutic follow-up policies for dysplastic lesions should be restrained.

Desiderata for an adequate evaluation of cancer screening.

Three main problems in the evaluation of mass screening are obtaining reliable estimates of the impact of cancer screening on morbidity and mortality, taking account of the large number of factors that together determine the impact of cancer screening, and dealing adequately with all three stages (explanation, prediction, and planning) of evaluation. Mathematical models, implemented in computer programs, should preferably be used for the evaluation. Fourteen desiderata for adequate model-based evaluation of mass screening are formulated; they concern the model, the computer program, and the methodology used. Satisfying these desiderata is essential in overcoming the three main problems mentioned. As yet, no evaluation research group is close to fulfilling all 14 desiderata. Sticking to one model only should thus be avoided.