Rapid review of evaluation of interventions to improve participation in cancer screening services.

Objective Screening participation is spread differently across populations, according to factors such as ethnicity or socioeconomic status. We here review the current evidence on effects of interventions to improve cancer screening participation, focussing in particular on effects in underserved populations. Methods We selected studies to review based on their characteristics: focussing on population screening programmes, showing a quantitative estimate of the effect of the intervention, and published since 1990. To determine eligibility for our purposes, we first reviewed titles, then abstracts, and finally the full paper. We started with a narrow search and expanded this until the search yielded eligible papers on title review which were less than 1% of the total. We classified the eligible studies by intervention type and by the cancer for which they screened, while looking to identify effects in any inequality dimension. Results The 68 papers included in our review reported on 71 intervention studies. Of the interventions, 58 had significant positive effects on increasing participation, with increase rates of the order of 2%-20% (in absolute terms). Conclusions Across different countries and health systems, a number of interventions were found more consistently to improve participation in cancer screening, including in underserved populations: pre-screening reminders, general practitioner endorsement, more personalized reminders for non-participants, and more acceptable screening tests in bowel and cervical screening.

Effectiveness of timed and non-timed second appointments in improving uptake in breast cancer screening.

OBJECTIVES: To estimate the effect on breast screening uptake of delayed, targeted, second timed appointments in women who did not take up an initial breast cancer screening appointment offer. METHODS: Non-attending women received a four-month delayed second timed appointment following non-response to the initial invitation and the normal open invitation sent to non-attenders. A comparison group was sent a four-month delayed additional open invitation. RESULTS: Response to the second timed appointments (percentage of re-invited women eventually attending in this episode) was 20%, corresponding to an estimated increase on total uptake of 6%. Response was highest in women who had previously attended screens. Response in the women offered an additional delayed open invitation was 7.5%, corresponding to an estimated 2.3% increase in overall uptake. CONCLUSIONS: Second timed appointments were almost three times as effective as additional open invitation. They should be targeted at women most likely to attend. A randomized study of second timed appointments versus open invitations should be conducted.

Lung cancer risk prediction: a tool for early detection.

Although 45% of men and 39% of women will be diagnosed with cancer in their lifetime, it is difficult to predict which individuals will be affected. For some cancers, substantial progress in individual risk estimation has already been made. However, relatively few models have been developed to predict lung cancer risk beyond effects of age and smoking. This paper reviews published models for lung cancer risk prediction, discusses their potential contribution to clinical and research settings and suggests improvements to the risk modeling strategy for lung cancer. The sensitivity and specificity of existing cancer risk models is less than optimal. Improvement in individual risk prediction is important for selection of individuals for prevention or early detection interventions. In addition to smoking, factors related to occupational exposure, personal medical history and family history of cancer can add to the predictive power. A good risk prediction model is one that can identify a small fraction of the population in which a large proportion of the disease cases will occur. In the future, genetic and other biological markers are likely to be useful, although they will require rigorous evaluation. Validation is essential to establish the predictive effect and for ongoing monitoring of the model's continued relevance.

Overdiagnosis, sojourn time, and sensitivity in the Copenhagen mammography screening program.

The goal of this research was to estimate the overdiagnosis at the first and second screens of the mammography screening program in Copenhagen, Denmark. This study involves a mammography service screening program in Copenhagen, Denmark, with 35,123 women screened at least once. We fit multistate models to the screening data, including preclinical incidence of progressive cancers and nonprogressive (i.e., overdiagnosed) cancers. We estimated mean sojourn time as 2.7 years (95% confidence interval [CI] 2.2-3.1) and screening test sensitivity as 100% (95% CI 99.8-100). Overdiagnosis was estimated to be 7.8% (95% CI 0.3-26.5) at the first screen and 0.5% (95% CI 0.02-2.1) at the second screen. This corresponds to 4.8% of all cancers diagnosed among participants during the first two invitation rounds and following intervals. A modest overdiagnosis was estimated for the Copenhagen screening program, deriving almost exclusively from the first screen. The CIs were very broad, however, and estimates from larger datasets are warranted.

Defining high-risk individuals in a population-based molecular-epidemiological study of lung cancer.

Within the framework of the Liverpool Lung Project (LLP), population-based case-control and prospective cohort studies are in progress to identify molecular and epidemiological risk factors and define populations and individuals most at risk of developing lung cancer. This report describes a strategy for selection of a high-risk population and further provides support for the inclusion of occupational and genetic risk factors in future models. Data from the case-control study (256 incident cases and 314 population controls) were analysed to define a high-risk population. Detailed lifestyle and occupational information were collected during structured interviews. Models were constructed using conditional logistic regression and included terms for age, tobacco consumption and previous respiratory disease. Smoking duration was chosen as the most important predictor of lung cancer risk [>50 years (OR 15.65, 95% CI 6.10-40.15)]. However, such a model would preclude younger individuals. Several combinations of previous respiratory disease were also considered, of which a history of bronchitis, emphysema or pneumonia (BEP) was the most significant (OR 1.86, 95% CI 1.28-2.69). A high-risk subset (based on combinations of smoking duration and BEP) was identified, which have a 4.5-fold greater risk of developing lung cancer (OR 4.5, 95% CI 2.33-8.68). Future refinement of the risk model to include individuals occupationally exposed to asbestos and with the p21 genotypes is discussed. There is real potential for environmental and genetic factors to improve on risk prediction and targeting of susceptible individuals beyond the traditional models based only on smoking and age. The development of a molecular-epidemiological model will inform the development of effective surveillance, early detection and chemoprevention strategies.

Overdiagnosis and overtreatment of breast cancer: estimates of overdiagnosis from two trials of mammographic screening for breast cancer.

Randomised controlled trials have shown that the policy of mammographic screening confers a substantial and significant reduction in breast cancer mortality. This has often been accompanied, however, by an increase in breast cancer incidence, particularly during the early years of a screening programme, which has led to concerns about overdiagnosis, that is to say, the diagnosis of disease that, if left undetected and therefore untreated, would not become symptomatic. We used incidence data from two randomised controlled trials of mammographic screening, the Swedish Two-county Trial and the Gothenburg Trial, to establish the timing and magnitude of any excess incidence of invasive disease and ductal carcinoma in situ (DCIS) in the study groups, to ascertain whether the excess incidence of DCIS reported early in a screening trial is balanced by a later deficit in invasive disease and provide explicit estimates of the rate of 'real' and non-progressive 'overdiagnosed' tumours from the study groups of the trials. We used a multistate model for overdiagnosis and used Markov Chain Monte Carlo methods to estimate the parameters. After taking into account the effect of lead time, we estimated that less than 5% of cases diagnosed at prevalence screen and less than 1% of cases diagnosed at incidence screens are being overdiagnosed. Overall, we estimate overdiagnosis to be around 1% of all cases diagnosed in screened populations. These estimates are, however, subject to considerable uncertainty. Our results suggest that overdiagnosis in mammography screening is a minor phenomenon, but further studies with very large numbers are required for more precise estimation.

Prospective estimation of rates of change in mammographic parenchymal patterns: influence of age and of hormone replacement therapy.

The objective of this study was to assess the effect of age, breast size and use of hormone replacement therapy (HRT) on the rate of change of mammographic parenchymal patterns, and the effect of age on the probability of misclassification between patterns. It was designed as a longitudinal study of the members of the treatment arm of a non-randomized screening trial in which subjects were assigned to screening or not by year of birth. The subjects were women in the Kotka district of Finland, each of whom attended for four or five mammographic screens. Participants were all women living in the district who were born in the relevant years and accepted an invitation to screening. A model was fitted to the longitudinal data comprising the observed Wolfe patterns on each woman, with age and breast size as predictors of breast density at first screen, age and HRT use as predictors of change in density at future screens, and age as a predictor of misclassification of true density between favourable (non-dense) and unfavourable (dense) patterns (according to the Wolfe classification). Relevant posterior probability estimates (with 95% credible intervals) were as follows. The probability that a woman of age 43.5 is truly in the favourable state ranged from 0.35 (0.34-0.37) for smallest breast size to 0.74 (0.72-0.76) for the largest. The probability that a woman is truly in the favourable state at first screen increased from 0.37 (0.36-0.38) at age 40 to 0.59 (0.58-0.60) at age 47. The probability that a woman having a later screen who had truly been in the unfavourable state at her previous screen changed to the favourable state increased from 0.12 (0.11-0.13) at age 42 to 0.48 (0.46-0.50) at age 55 for a woman not taking HRT, and from 0.10 (0.09-0.11) to 0.43 (0.40-0.45) at the same ages for a woman taking HRT. The probability that a woman would have changed from being truly in the favourable state to the unfavourable state was 0.003 (0.001-0.003) for any age and HRT use. The probability that a woman truly in a favourable state would be correctly classified rose from 0.87 (0.85-0.89) at age 40 to 0.998 (0.997-0.998) at age 55. The probability that a woman truly in the unfavourable state would be correctly classified decreased from 0.96 (0.95-0.97) to 0.93 (0.91-0.94) between the same ages. The probability of being in a non-dense, favourable state increases with age, as does the rate of change from dense to non-dense patterns. These are consistent with previous work. The probability of non-dense patterns and the rate of change to non-dense patterns are reduced with HRT use. Errors of classification are relatively rare, but are dependent on the age of the subject.

Misclassification in a matched case-control study with variable matching ratio: application to a study of c-erbB-2 overexpression and breast cancer.

We provide a simple analytic correction for risk factor misclassification in a matched case-control study with variable numbers of controls per case. The method is an extension of existing methodology, and involves estimating the corrected proportions of controls and cases in risk factor categories within each matched set. These estimates are then used to calculate the Mantel-Haenszel odds ratio estimate corrected for misclassification. A simulation-based interval estimate is developed. An example is given from a study of risk factors for progression of benign breast disease to breast cancer, in which the risk factor is a biological marker measured with poor sensitivity.

Bayesian evaluation of breast cancer screening using data from two studies.

The mean sojourn time (the duration of the period during which a cancer is symptom free but potentially detectable by screening) and the screening sensitivity (the probability that a screen applied to a cancer in the preclinical screen detectable period will result in a positive diagnosis) are two important features of a cancer screening programme. Little data from any single study are available on the potential effectiveness of mammographic screening for breast cancer in women with a family history of the disease, despite this being an important public health issue. We develop a method of estimation, from two separate studies, of the two parameters, assuming that transition from no disease to preclinical screen detectable disease, and from preclinical disease to clinical disease, are Poisson processes. Estimation is performed by a Markov chain Monte Carlo algorithm. The method is applied to the synthesis of two studies of mammographic screening in women with a family history of breast cancer, one in Manchester and one in Kopparberg, Sweden.