Survival of patients with stage III colon cancer varies widely according to T-N sub-stages. Estimating the benefit of each therapeutic option in each T-N subgroup may provide more accurate information helping doctors and patients in the complex shared decision-making process surrounding adjuvant therapy.

The outcomes data of 12,834 patients with stage III colon cancer enrolled in the IDEA trial served as our database. Patients were categorised in 16 sub-stages, based on T-N categories. We created a meta-regression model to predict the expected 5-year DFS within each T-N sub-stage. We then evaluated the efficacy of each therapeutic option in every sub-stage, working backward by subtraction, using an average of the HRs reported in pertinent trial publications as a conversion factor.

Large differences in 5-year DFS rate were observed among the subgroups, ranging from 89% (T1N1a) to 31% (T4N2b) in the overall population. The contribution to the outcome of each therapeutic option in this setting varied widely across sub-stages. According to our model, patients with T1N1a cancers have a projected 5-year DFS of 79.6% with surgery alone. Adjuvant fluoropyrimidine alone results in 5.6% absolute DFS gain; an additional 2.3% and 0.8% gain is seen with oxaliplatin for 3 and 6 months, respectively. Patients with T4N2b cancers show a 13.9% 5-year DFS with surgery alone, and an 11.2%, 6.4%, 2.5% increase with the aforementioned adjuvant options, respectively.

The resulting overlay bar graph gives patients and doctors the projected relative benefit of each treatment option and may substantially help the shared decision-making process, although caution must be exercised in using this model due to the significant variance of the estimates.