Projection of cancer screening on metastasis: A prostate cancer case study

From the Etzioni Group, Public Health Sciences Division

One of the main purposes of cancer screening is to detect localized tumors early to reduce cancer mortality. The belief or prediction that early detection will decrease the likelihood of advanced-stage cancer is called “stage shift”. Unfortunately, stage shift doesn’t consider progressive cases after diagnosis. Additionally, the diagnosis of advanced-stage cancer with screening does not always reduce disease-specific deaths. The reduction of advanced-stage cases at diagnosis will become organ-confined cases and increase the number of cases eligible for metastasis. Also, progressive cases may reduce the incidence of metastatic disease at diagnosis, resulting in the same number of metastatic cases in a screened and non-screened group.

To date, most of the population-based screening trials have not analyzed the overall risk of metastasis in the screening and control arm during follow-up but at diagnosis.  Determining the effect of screening on metastasis can be difficult and time-consuming; it requires follow-up on localized cases in both trial arms. Projecting an observed stage shift and case differences in advanced-stage disease aid in developing practical, realistic projections and the effect of the long-term screening. The Etzioni Group (Fred Hutch, Division of Public Health Sciences) developed a modeling framework to project the decrease in metastasis at diagnosis to the decrease in overall metastasis under screening. This framework included data and assumptions important to the realistic projection of the long-term effects of screening. In the study, published in the Journal of Medical Screening, the investigators utilized the framework to project how lower incidence of metastasis diagnosis would translate into incidence of metastasis overall and compare it with the observed pattern of overall metastasis from the trial.

The modeling framework includes three major components: stage-specific incidence of disease at diagnosis with and without screening, the division of excess localized cases in the screen arm into overdiagnosed and non-overdiagnosed cases, and the distribution of time from diagnosis to metastasis of non-overdiagnosed, localized cases. The European Randomized Screening for Prostate Cancer (ERSPC) data was utilized for stage shift and the Scandinavian Prostate Cancer Group-4 trial data was utilized for the risk of metastatic progression. The ERSPC randomized controlled screening trial began in 1993 and compared Prostate-Specific Antigen (PSA) screening every 2-4 years for adult men aged 55-69; this analysis used a total of 76,813 men (N=36,270 screening; N=40,543 control) from four centers of the ERSPC study. The SPCG-4 randomized control trial, began in 1989, observed the progression of prostate cancer cases; a total of 695 men (N=347 intervention; N=348 control) were used in the study.

Fred Hutch’s microsimulation model of prostate cancer natural history was used to project the incidence of de novo metastatic cancer. This model links an individual’s PSA growth to the start and progression of a preclinical prostate tumor. For example, a man with a faster PSA growth rate may have a shorter time to clinical diagnosis and a higher likelihood of metastatic stage at diagnosis. Metastatic progression was projected at clinical diagnosis (screen arm). In order to analyze metastasis at diagnosis and overall metastasis, the hazard ratio (time until metastasis), relative risk (metastasis rate ratio measured at 12 years), absolute risk (difference at 12 years) were utilized.

According to the simulation, there were fewer de novo metastatic cases at diagnosis than the control arm. In the screen vs. control arm of the 12-year simulated trial, the relative risk was 0.57, the hazard ratio was 0.60, and the absolute risk reduction was 39.6/10,000 men. The modeled effect of screening on progressive metastasis in the screen and controls arms showed a decrease in absolute and relative benefit for overall metastatic disease. Compared to each other, the simulated trial and ERSPC metastasis (at diagnosis and overall) were in agreement. 

In conclusion, the Etzioni group created a novel modeling framework to examine the long-term impact of cancer screening on metastasis incidence. This model is more accurate than other predictions that only use stage shift at diagnosis because the researchers measured de novo and progressive cases. We know that screening decreases the amount of advanced-stage cases at diagnosis, but it also increases the amount of non-overdiagnosed localized cases with the eligibility to progress. Therefore, the expectations of screening benefits based on stage shift should be mitigated when evaluating disease mortality. 

This research was supported by the National Cancer Institute, Cancer Intervention and Surveillance Modelling Network.

Fred Hutch/UW Cancer Consortium member Ruth Etzioni contributed to this work.

Lange J, Remmers S, Gulati R, Bill-Axelson A, Johansson JE, Kwiatkowski M, Auvinen A, Hugosson J, Hu JC, Roobol MJ, Carlsson SV, Etzioni, R. Impact of cancer screening on metastasis: A prostate cancer case study. Journal of Medical Screening. 2021 Feb 9.