# Understanding the Will Rogers Phenomenon in Cholangiocarcinoma Research and Beyond

## TL;DR

This paper examines how reclassifying cholangiocarcinoma subtypes affects survival outcomes and questions the statistical validity of the Will Rogers phenomenon in cancer research.

## Contribution

The study introduces a novel reclassification analysis to assess the impact of subtype misclassification on survival estimates in cholangiocarcinoma.

## Key findings

- No significant survival differences were found between intrahepatic and extrahepatic cholangiocarcinoma subtypes after reclassification.
- Reclassification of Klatskin tumors did not statistically confirm the Will Rogers phenomenon in this cancer context.
- Perihilar cholangiocarcinoma showed increased mortality risk compared to distal cholangiocarcinoma.

## Abstract

The authors performed a comparative analysis of different cholangiocarcinoma (CC) subtypes and provided a reclassification analysis to investigate the impact of misclassifications of Klatskin tumors in CC research. Overall, our study revealed weak evidence of survival differences between the subtypes after curative liver resection. We also questioned whether the Will Rogers phenomenon is supported by statistical evidence when contrasting changes that come from multiple reclassifications in the cancer literature.

Background. The existing literature highlights a lack of comparative studies between subtypes of cholangiocarcinoma (CC) and the impact of misclassification on the epidemiological parameters. Methods. A retrospective study was conducted to evaluate the surgical outcomes. The authors used Poisson regression with modified errors to calculate the risk ratios (RR) and reported post-estimation marginal effects. Coefficient estimates, variance inflation factors, and Pearson’s goodness-of-fit test statistics were used to check for multicollinearity and model fit, respectively. We also performed a reclassification analysis by modeling Klatskin tumors (PCC) as extrahepatic (ECC), reclassifying them as intrahepatic (ICC), and comparing the corresponding changes in estimates. Results. Regression analysis revealed an increased risk of death in patients with ICC (RR = 2.05, 95% CI 1.11–3.78) and PCC (RR = 2.03, 95% CI 0.97–4.24) compared to those with DCC. When PCC was analyzed as an ECC, the ICC revealed an RR of 1.52 (95% CI 0.84–2.73). Further reclassification of PCC showed an RR of 2.04 for ICC (95% CI: 1.53–3.53). The adjusted marginal effects saw a reduction in the death probability for both ICC and ECC. However, post hoc analyses revealed insufficient evidence for differences between the reclassified models. Conclusions. Patients with DCC had slightly better prognosis compared to ICC and PCC. We found no differences in survival between ICC and ECC (combining DCC and PCC). The decrease in mortality risk due to reclassification in both groups was not confirmed statistically. Future studies should focus on statistical evidence when referring to the Will Rogers phenomenon, instead of inferring from raw comparisons.