JMIR Publications

ABSTRACT

Background:

The incidence of early-onset colorectal cancer (EOCRC) rose abruptly starting in the mid-1990s. Inherited genes and inflammatory bowel disease (IBD) which are already known to be risk factors for colorectal cancer are not the causes of most cases of EOCRC. We hypothesized that the increasing incidence may be driven by an off-target effect of a medication, not previously widely used, in a genetically susceptible subgroup of young adults.

Objective:

To evaluate the possibility that pharmaceutical agents serve as risk factors for EOCRC using a novel machine learning methodology based on gradient boosted decision trees.

Methods:

Data were extracted from the database of Maccabi Healthcare Services (MHS), a state-mandated health provider covering 26% of the Israeli population. 941 EOCRC cases (<50 years of age) diagnosed during 2001-2019 were identified and were density matched with 9410 controls. IBD patients and patients with a known inherited cancer susceptibility syndrome were excluded. Complete medication dispensing history (Over the Counter [OTC] and prescription). A 2-year lag period prior to EOCRC diagnosis was used for ascertaining medication exposure to minimize the chance of reverse causation. Main Outcome(s) and Measure(s): An advanced machine learning algorithm based on gradient boosted decision trees coupled with Bayesian model optimization and repeated data sampling was used to sort through the very high-dimensional drug dispensing data to identify specific medication groups that were consistently linked with EOCRC while allowing for synergistic or antagonistic interactions between medications. Odds ratios for the identified medication classes were obtained from a conditional logistic regression model.

Results:

Of more than 800 medication classes, we identified several classes that were consistently (>50%) associated with EOCRC risk across independently trained models. Interactions between medication groups did not seem to substantially affect the risk. In our analysis drug groups that were consistently positively associated with EOCRC included beta blockers (OR=1.94, 95% CI=1.37-2.77, p<0.01) and valerian (Valeriana officinalis), OR=1.61, 95% CI=1.15-2.25, p=0.01); antibiotics were not consistently associated with EOCRC risk.

Conclusions:

Our analysis suggested that the development of EOCRC may be correlated with prior use of specific medications. Additional analyses should be employed to confirm the results and to decipher the underlying mechanisms driving the observed associations.