# On a fundamental problem in the analysis of cancer registry data

**Authors:** Sho Komukai, Satoshi Hattori, Bernard Rachet  
**arXiv:** [2303.09141](https://arxiv.org/abs/2303.09141)  
**Date:** 2023-03-16  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2303.09141v1/

## Key points

- Primary focus in epidemiology research utilizing cancer registry data is often on making inferences regarding cancer-related deaths rather than overall survival
- Accurately determining the cause of death within cancer registries can be challenging or unreliable
- Methodologies such as relative survival ratios and net survival concepts have been developed to address this issue by adjusting for non-cancer deaths using external life tables of the general population
- Common assumption underlying these adjustments is that mortality rates in external life tables reflect non-cancer mortality rates among cancer patients, but discrepancies may arise
- Sensitivity analyses proposed by Talbäck and Dickman aim to tackle issues related to discrepancies but may require additional information not readily available
- A novel approach is proposed to make inferences on net survival while accounting for potential inclusion of cancer patients and deaths in general population life tables, which effectively mitigates biases in estimating net survival
- Simulation study conducted to evaluate the performance of the proposed method demonstrated successful removal of biases and provided more accurate estimates of net survival
- Practical application of the proposed method using real-world cancer registry data from England showcased its effectiveness and contributed valuable insights to improving inference methods for net survival analysis within epidemiological research with cancer registry data

## Summary

In epidemiology research utilizing cancer registry data, the primary focus often lies on making inferences regarding cancer-related deaths rather than overall survival. However, accurately determining the cause of death can be challenging or unreliable within cancer registries. To address this issue, various methodologies have been developed, including the utilization of relative survival ratios and net survival concepts. These methods involve adjusting for non-cancer deaths by referencing external life tables of the general population. One common assumption underlying these adjustments is that mortality rates in external life tables reflect non-cancer mortality rates among cancer patients. Nevertheless, discrepancies may arise when the population used to create a life table includes individuals who have experienced cancer-related deaths or are currently battling cancer. Sensitivity analyses proposed by Talb\"{a}ck and Dickman aim to tackle this issue but often require additional information that may not be readily available. To enhance existing methodologies, a novel approach is proposed to make inferences on net survival while accounting for potential inclusion of cancer patients and deaths in general population life tables. This adjustment method considers the correspondence between cancer mortality rates in both the life table and the cancer registry data. Through this innovative technique, biases in estimating net survival due to potential presence of cancer patients and deaths in life tables can be effectively mitigated. A simulation study was conducted to evaluate the performance of the proposed method. By simulating natural histories of subjects within a birth cohort and generating corresponding cancer registry data and life tables, researchers were able to assess how well the new approach addressed biases in estimating net survival. The results demonstrated that the proposed method successfully removed biases and provided more accurate estimates of net survival. Illustrating its practical application, the proposed method was implemented using real-world cancer registry data from England. By showcasing its effectiveness in a tangible setting, this study contributes valuable insights to improving inference methods for net survival analysis within epidemiological research with cancer registry data.