Peter G. Warmerdam

Quantitative Interpretation of Age-Specific Mortality Reductions From the Swedish Breast Cancer-Screening Trials

BACKGROUND Results from five Swedish randomized trials may provide the most conclusive evidence on the effect of mammographic screening and have been used to forecast the expected reduction in breast cancer mortality in other programs. However, those trials demonstrated different degrees of reduction. The interpretation of observed mortality reduction after long follow-up for women aged 40-49 years at trial entry is both important and controversial. PURPOSE We estimated what percentage of the observed mortality reduction for women aged 40-49 years at entry into the five Swedish screening trials might be attributable to screening these women at 50 years of age or older. Moreover, we calculated the most likely percentage mortality reduction for specific screening programs if the Swedish results were generalized and analyzed whether characteristics of each trial might at least partly explain the observed differences in reductions among the trials. METHODS Each Swedish trial was simulated with one underlying computer simulation model (MISCAN--MIcrosimulation SCreening ANalysis) of the natural history of the disease and the performance of screening, taking into account nine important trial characteristics. Improvement in prognosis for screen-detected case patients was estimated with age-specific reduction for all trials and each trial design as a reference. RESULTS An expected 7% reduction in breast cancer mortality for women aged 40-49 years at trial entry (relative risk [RR] = 0.93) was determined by computer modeling, assuming no improvement in prognosis for cancers that are screen detected before 50 years of age. This result indicates that, of the overall 10% observed reduction (RR = 0.90) in the five Swedish trials analyzed, most (70%) of this reduction might be attributable to screening these women in later rounds after their 50th birthday. Using additional trial information, predictions of breast cancer mortality reduction in women 50 years or older might be 11% larger than previously expected, assuming that high-quality mammographic screening can be achieved in nationwide programs. For women aged 50-69 years at trial entry, the differences in expected versus observed mortality reduction among the trials are estimated to be relatively small. (Expected mortality reductions range from 24% to 32%). CONCLUSIONS Results from the Swedish randomized breast cancer-screening trials should be seen as more favorable regarding the effect of mammographic screening in reducing breast cancer mortality for women aged 50-69 years than was estimated earlier. Our analyses also suggest that the improvement in prognosis due to screening for women aged 40-49 years is much smaller than that for women aged 50 years or older. Approximately, 70% of the 10% observed reduction in breast cancer mortality (i.e., 7%) for women aged 40-49 years at trial entry might be attributable to a reduction due to screening these women after they reach age 50. IMPLICATIONS Detailed screening data for the 40- to 49-year age group of all Swedish trials should be analyzed to specifically estimate the natural history and performance of screening in this age group.

Cost effectiveness of shortening screening interval or extending age range of NHS breast screening programme: computer simulation study

Abstract Objective : To compare the cost effectiveness of two possible modifications to the current UK screening programme: shortening the screening interval from three to two years and extending the age of invitation to a final screen from 64 to 69. Design : Computer simulation model which first simulates life histories for women in the absence of a screening programme for breast cancer and then assesses how these life histories would be changed by introducing different screening policies. The model was informed by screening and cost data from the NHS breast screening programme. Setting : North West region of England. Main outcome measures : Numbers of deaths prevented, life years gained, and costs. Results : Compared with the current breast screening programme both modifications would increase the number of deaths prevented and the number of life years saved. The current screening policy costs £2522 per life year gained; extending the age range of the programme would cost £2612 and shortening the interval £2709 per life year gained. The marginal cost per life year gained of extending the age range of the screening programme is £2990 and of shortening the screening interval is £3545. Conclusions : If the budget for the NHS breast screening programme were to allow for two more invitations per woman, substantial mortality reductions would follow from extending the age range screened or reducing the screening interval. The difference between the two policies is so small that either could be chosen.

Radiation risk of mammography related to benefit in screening programmes: a favourable balance?

Objectives To estimate the number of breast cancer deaths induced by low dose radiation in breast cancer screening programmes compared with numbers prevented. Methods A computer simulation model on the natural history of breast cancer was combined with a model from BEIR-V on induced breast cancer mortality from low levels of radiation. The improvement in prognosis resulting from screening was based on the results of the Swedish overview of the randomised screening trials for breast cancer and the performance of screening in the Netherlands. Different scenarios (ages and intervals) were used to explore the objectives. Sensitivity analyses were carried out for latency period, dose of mammography, sensitivity of the screening test, early detection by screening of induced breast tumours, and new 1996 risk estimates by Howe and McLaughlin. Results For a screening programme, age group 50–69, two year interval, 2 mGy per view, the balance between the number of deaths induced versus those prevented was favourable: 1:242. When screening is expanded to the age group 40–49 with a one or two year interval the results may be less favourable, that is, 1:66 and 1:97. According to these scenarios and with the Dutch scenario as reference, one breast cancer death from radiation may be expected to occur to save eight extra deaths from breast cancer. If screening was equally effective in young women as in women aged 50–69, the marginal value was 1:±30. Assuming detection of induced cancers by screening could influence the ratios by about 30%, but did not substantially change the conclusions. The new risk estimates by Howe and McLaughlin resulted in five times to eight times favourable ratios breast cancer deaths induced to prevented. Besides age group of screening, dose of mammography is the other determinant of risk. Conclusions For screening under the age of 50, the balance between the number of breast cancer deaths prevented by screening compared with the number induced by radiation seem less favourable. Credibility intervals were however wide, because of many uncertainties of radiation risk at very low doses.

Stage Distribution at First and Repeat Examinations in Breast Cancer Screening

Objectives To investigate observed stage distributions at first and repeat screenings. To compare the observed outcomes with expected values based on simulation modelling, varying the assumptions about the natural history of the disease. Methods An overview is made of observed data on stage distribution at first and repeat screenings and the difference between those distributions is summarised in a Gini coefficient. Four possible explanations for the observations are considered, two of these are worked out as Miscan simulation models, and the outcomes are compared with observations. Results Often the reported stage distributions at repeat screenings are not or only slightly more favourable than at first screenings and, in the ones that are more favourable, the difference is relatively small. If, in the Miscan model, it is assumed that there is no correlation between the duration of preclinical breast cancer in consecutive tumour size categories and that there is a strong influence of latent cancers, it is not possible to reproduce the observed outcomes. Conclusions The two modelled explanations are not sufficient. Decreasing sensitivity seems an unlikely explanation for the discrepancy in many screening programmes. The possibility that the observations may be explained because false reassurance has been given should be seriously considered and investigated.

A comparison of disease specific survival of patients who died of and who had newly diagnosed prostate cancer.

PURPOSE We investigated the validity of calculating prostate specific survival from a population of deaths occurring during a given period. MATERIALS AND METHODS Stochastic simulation was used to generate a large number of life histories of men with prostate cancer. RESULTS The investigated method of calculating survival can lead to different outcomes compared to the standard method. In an example with a Dutch population structure the investigated method leads to strong underestimation of 25 years of survival. CONCLUSIONS The method of calculating survival studied is theoretically not valid. Several probable changes in the population can produce results that are different from the standard method of calculating survival.