Mismeasured Covariate in the Long-Term Survival of Colorectal Cancer

Colorectal cancer (CRC) is one of the most important causes of morbidity and mortality worldwide. This study aimed to determine the effect of measurement error of risk factors on the cure fraction of CRC patients.

This study was conducted using the medical records of 346 patients with CRC, who were followed up between 2006 and 2017 in Shiraz, Iran. In our data, lymph node ratio (LNR) was a characteristic measuring with error. This variable was used in the model with 0.04 and 0.8 of error variance. Nonmixture nonparametric cure rate model and its corrected forms, simulation-extrapolation (SIMEX) and corrected score (CS), were applied to the data.

In noncured cases, the mean survival time was 1115.45 (95% confidence interval, 1043.60-1187.30) days. The 1-, 3-, and 5-year survival rates were 0.93, 0.71, and 0.65, respectively. The proportion of cured patients was 65.2%. The SIMEX method did not change the effect of LNR substantially on cure fraction as compared with the naive method when the variance of measurement error was 0.04 and 0.80. The CS method changed the effect of LNR on cure fraction even when the variance of measurement error was 0.04.

The best method to assess the effect of LNR on cure fraction was the naive method, and the CS method was not deemed to be a valid method to correct the measurement error in LNR.