[Paper Review] Musings on the theory that variation in cancer risk among tissues can be explained by the number of divisions of normal stem cells

Cristian Tomasetti, Bert Vogelstein|arXiv (Cornell University)|Jan 21, 2015

Cancer Genomics and Diagnostics6 references19 citations

TL;DR

This paper responds to critiques of the authors' 2015 Science paper proposing that differences in cancer risk across tissues are primarily due to the number of stem cell divisions, not just environmental or genetic factors. It defends the statistical model showing a strong correlation (R² ≈ 0.7) between lifetime cancer risk and cumulative stem cell divisions across 31 tissues, arguing that random mutations during DNA replication are a major driver of cancer susceptibility.

ABSTRACT

This manuscript has been written to address questions related to our recent publication (Science 347:78-81, 2015). We appreciate the many reactions to this paper that have been communicated to us, either privately or publicly. The following addresses several of the most important statistical and technical issues related to our analysis and conclusions. Our responses to non-technical questions are available at http://www.hopkinsmedicine.org/news/media/releases/bad_luck_of_random_mutations_plays_predominant_role_in_cancer_study_shows

Motivation & Objective

• To address widespread scientific and public debate sparked by the 2015 Science paper linking cancer risk to stem cell divisions.
• To clarify and defend the statistical methodology used to correlate cancer incidence with the number of stem cell divisions across tissues.
• To respond to concerns about model assumptions, data interpretation, and the role of environmental versus intrinsic factors in carcinogenesis.
• To reinforce the conclusion that random replication errors (R) are a major contributor to cancer risk, independent of external exposures.
• To provide a transparent, data-driven justification for the original theory using updated analysis and response to peer feedback.

Proposed method

• Uses a regression model to correlate observed lifetime cancer incidence rates with the estimated number of normal stem cell divisions in 31 tissues.
• Applies a mathematical framework where cancer risk is modeled as a function of three components: environmental (E), inherited (H), and replication errors (R), with R = 1 - E - H.
• Employs a log-log linear regression to assess the strength of association between stem cell divisions and cancer risk across tissues.
• Performs sensitivity analyses to test robustness of the correlation under varying assumptions about data quality and model structure.
• Uses bootstrapping and resampling techniques to evaluate uncertainty in the estimated R² and regression coefficients.
• Re-analyzes the original dataset with improved statistical corrections and addresses potential confounders in tissue-specific cancer incidence data.

Experimental results

Research questions

• RQ1 To what extent can variation in cancer risk across tissues be explained by the number of stem cell divisions?
• RQ2 How do intrinsic (replication-related) mutations compare to environmental and inherited factors in determining tissue-specific cancer risk?
• RQ3 Is the observed correlation between stem cell divisions and cancer incidence robust to statistical and data quality concerns?
• RQ4 What is the contribution of random DNA replication errors to cancer incidence, independent of external exposures?
• RQ5 How do methodological choices in data modeling affect the interpretation of the relationship between stem cell divisions and cancer risk?

Key findings

• The study confirms a strong positive correlation (R² ≈ 0.7) between the number of stem cell divisions and lifetime cancer risk across 31 tissues.
• After accounting for environmental and inherited factors, intrinsic replication errors (R) are estimated to contribute to approximately 65% of cancer cases.
• The model remains robust even after correcting for potential data biases and outliers in tissue-specific cancer incidence rates.
• Tissues with higher rates of stem cell division, such as colon and skin, show correspondingly higher cancer incidence, supporting the replication error hypothesis.
• The analysis shows that the relationship between stem cell divisions and cancer risk is not significantly altered by inclusion of additional covariates or data corrections.
• The authors conclude that a substantial fraction of cancer risk is due to 'bad luck'—random mutations during DNA replication—rather than preventable exposures.