UNDERSTANDING THE EFFECTS OF LEAD TIME AND OVERDIAGNOSIS IN MAMMOGRAPHY SCREENING: THE SCREENING ILLUSTRATOR

Purpose:

To illustrate the consequences of breast cancer screening in terms of detected cases and overdiagnosis by means of a simple spreadsheet model and compare to the 1.5 expected increase in incidence suggested by the International Agency for Research on Cancer (IARC).

Method:

We modeled hypothetical populations of women aged 50-79 years, in which population size and number of cancer cases  are recorded in five year age groups (50-54, 55-59, …, 75-79) and five year time periods (1985-9, 1990-4,…2010-4). First, we use a cross-sectional approach to illustrate how screening brings breast cancer cases forward in time both with respect to period and age allowing for underlying period and age trends. For each time period we compute incidence rates in the scenario with and without screening, respectively, and with and without incorporation of the compensatory drop among women leaving the screening program. Relative incidence rate ratios (RIRR) are used to compare changes in incidence rates over time across the two scenarios. Next, we investigate how stage specific incidence rates change when both localized and advanced cancers are brought forward in time by screening, and some advanced cancers are thereby diagnosed as localized. Finally, we incorporate overdiagnosis into the model as cases that are forwarded in time by screening, but would never have been detected without screening. We repeat these scenarios in a Lexis-based approach, which allows for an underlying birth cohort trend.

Result:

When including the compensatory drop, the cross-sectional model predicted a peak in incidence right after the introduction of screening (RIRR = 1.28). The overall incidence rate quickly stabilized at a slightly higher, but constant level than before introduction (RIRR = 1.01). Introducing 30% overdiagnosis resulted in incidence being 44% higher excluding the compensatory drop. The findings persisted when using the Lexis-based approach.

Conclusion:

A simple, user friendly model provides insights into the mechanics of screening in terms of changes in incidence and disease stage. It may assist students, clinicians and researchers in understanding phenomena such as lead time and overdiagnosis. The model results suggest that the expected increase in incidence suggested by IARC for good screening programs (RIRR=1.5) cannot be achieved by lead time effects amplified by underlying age and period trends alone, but requires overdiagnosis to be added to the model.