Assessing the magnitude of surveillance bias in prostate cancer, melanoma and lung cancer
DOKPE
- Tancredi, Stefano ORCID University of Fribourg
- Cullati, Stéphane ORCID University of Fribourg
- van der Linden, Bernadette W.A. ORCID University of Fribourg
- Braggion, Axelle ORCID University of Fribourg
- Chiolero, Arnaud University of Fribourg
- 2026
Published in:
- European Journal of Cancer. - Elsevier BV. - 2026, vol. 245, p. 116942
English
Background
Changes in cancer incidence can result from changes in screening and diagnostic practices rather than changes in the true occurrence of cancer, leading to surveillance bias. Quantitative approaches to estimate this bias are lacking.
Objectives
To develop an approach to estimate surveillance bias in prostate cancer, melanoma, and lung cancer.
Methods
We used population-based data from Swiss cancer registries on incidence and mortality from 1989 to 2021. Age-standardized incidence was analyzed using Joinpoint regression to identify periods with distinct trends. The same periods were used to segment mortality trends. The magnitude of surveillance bias was assessed for each period by computing the absolute normalized difference between the mean annual changes in age-standardized incidence and mortality rates, since mortality is less affected by screening and diagnostic practices than incidence. Larger differences indicated greater bias. We defined three cut-offs to categorize the bias into low, moderate and high. Analyses were also conducted by cancer stage.
Results
Our estimator of surveillance bias for prostate cancer points to a high bias in 1989–2004 (absolute normalized difference= 5%), low in 2004–2011 (absolute normalized difference = 0.3%), and high in 2011–2014 and 2014–2021 (absolute normalized difference = 6% and 5%). For melanoma, the bias was high from 1989 and 2003 and moderate between 2003 and 2021 (absolute normalized difference = 5.5% and 1.6%). For lung cancer, it was low over the entire study period (absolute normalized difference = 0.5%). In stage-specific analyses, surveillance bias was greater for early-stage than advanced-stage cancers.
Conclusions
We estimated surveillance bias using a simple approach that can be used in daily monitoring activities. Further studies are needed to refine these estimates.
Changes in cancer incidence can result from changes in screening and diagnostic practices rather than changes in the true occurrence of cancer, leading to surveillance bias. Quantitative approaches to estimate this bias are lacking.
Objectives
To develop an approach to estimate surveillance bias in prostate cancer, melanoma, and lung cancer.
Methods
We used population-based data from Swiss cancer registries on incidence and mortality from 1989 to 2021. Age-standardized incidence was analyzed using Joinpoint regression to identify periods with distinct trends. The same periods were used to segment mortality trends. The magnitude of surveillance bias was assessed for each period by computing the absolute normalized difference between the mean annual changes in age-standardized incidence and mortality rates, since mortality is less affected by screening and diagnostic practices than incidence. Larger differences indicated greater bias. We defined three cut-offs to categorize the bias into low, moderate and high. Analyses were also conducted by cancer stage.
Results
Our estimator of surveillance bias for prostate cancer points to a high bias in 1989–2004 (absolute normalized difference= 5%), low in 2004–2011 (absolute normalized difference = 0.3%), and high in 2011–2014 and 2014–2021 (absolute normalized difference = 6% and 5%). For melanoma, the bias was high from 1989 and 2003 and moderate between 2003 and 2021 (absolute normalized difference = 5.5% and 1.6%). For lung cancer, it was low over the entire study period (absolute normalized difference = 0.5%). In stage-specific analyses, surveillance bias was greater for early-stage than advanced-stage cancers.
Conclusions
We estimated surveillance bias using a simple approach that can be used in daily monitoring activities. Further studies are needed to refine these estimates.
- Faculty
- Faculté des sciences et de médecine
- Department
- Section de médecine
- Language
-
- English
- Classification
- Pathology, clinical medicine
- Other electronic version
- License
-
CC BY
- Open access status
- hybrid
- Identifiers
-
- DOI 10.1016/j.ejca.2026.116942
- ISSN 0959-8049
- Persistent URL
- https://folia.unifr.ch/unifr/documents/336396
Statistics
Document views: 17
File downloads:
- 1-s2.0-s0959804926007239-main: 11