General demographic and clinicopathological characteristics of M-CLM
A total of 7179 patients were retrieved from the SEER database according to the inclusion and exclusion criteria. Then, according to the SEER Combined Metastasis at DX-liver (2010+) code, a total of 5816 CLM patients from 2010 to 2015 were enrolled, including 306 M-CLM patients and 5510 A-CLM patients. The results showed that M-CLM patients had the general features of MCs, such as larger tumour sizes, more localizations to the right colon, and higher pT and pN stages than A-CLM patients (P < 0.05 each, Table 1). In addition, the results also showed that the M-CLM group had more female patients and more retrieved and positive lymph nodes and accounted for a higher proportion of surgeries than the A-CLM group (P < 0.05 each, Table 1). Other variables, such as race, age, CEA level, number of primary tumours and tumour differentiation, were comparable between the two groups (P > 0.05 each, Table 1). In order to reduce the possible statistical biases, we performed 1:1 PSM analyis as described in methods and produced 306 patients in the A-CLM group and the M-CLM group respectively. Results showed that the clinicopathological characteristics and surgery information of the A-CLM and M-CLM group patients after PSM were strongly in line with the original data before PSM (Table S1), which strengthened the fingdings.
Long-term survival in M-CLM
We then analysed the potential survival difference between M-CLM and A-CLM patients via Kaplan-Meier analysis and log-rank tests. The results showed that the follow-up of the whole study cohort was 0-83 months, and the median follow-up was 17.0 months. The OS of M-CLM patients was comparable to A-CLM patients (22.59±1.24 vs. 25.65±0.36 months, P = 0.088, Figure 1A). The CSS of M-CLM patients was also similar to that of A-CLM patients (24.33±1.33 vs. 28.19±0.39 months, P = 0.053, Figure 1B); although the actual values of the OS and CSS of M-CLM were lower than those of A-CLM, the difference was not statistically significant. The finding of OS and CSS of M-CLM patients were similar as A-CLM patients was also comfirmed after PSM (Figure S1A, B).
Long-term survival in M-CLM classified by surgery type
Furthermore, we explored the potential advantage of different surgery types for long-term survival. The results showed that the cohort who underwent resection for both the primary tumour and liver metastases had the best OS (41.15±0.96 months, P < 0.001), followed those who underwent resection only for the primary lesion (26.79±0.47 months) and for metastatic lesions (21.44±4.22 months), which had similar OS (P = 0.388), and the patients who did not undergo any surgery had the poorest OS (13.08±0.39 months, P < 0.001) (Figure 1C). These results were also confirmed for the CSS analysis (Figure 1D). Then, we classified and analysed the effect of surgery on the survival of M-CLM and A-CLM patients. The results showed that M-CLM patients who underwent any type of surgery (primary or metastatic lesion resection or both) had significantly better OS and CSS than those who did not undergo any type of surgery (P < 0.001 for all, Figure 2A-B). The survival analyses in the A-CLM group also yielded similar results (P < 0.001, Figure 2C-D).
Survival differences between M-CLM and A-CLM stratified by surgery type
We previously found that M-CLM had comparable OS and CSS to A-CLM (Figure 1A-B), since surgery could result in survival benefits for both cancers, and so we further analysed the potential survival differences between M-CLM and A-CLM via stratification of surgery types. The results showed that among all patients who underwent any kind of surgery, M-CLM patients had poorer OS (P < 0.001, Figure 3A) and CSS (P < 0.001, Figure 3B) than A-CLM patients. However, the OS and CSS were not significantly different between M-CLM and A-CLM patients who did not undergo surgery (P = 0.394 and P = 0.404, respectively, Figure 3C-D). Kaplan–Meier OS and CSS curves after PSM also indicated the similar results (Figure S1C-F).
Then, we continued to explore the survival differences via stratification of surgery into primary or metastatic lesion resection. The results showed that among patients who underwent surgery for primary lesion resection, M-CLM patients had poorer OS and CSS than A-CLM patients (P P < 0.05 each, Figure 4A-B). Among patients who underwent surgery for metastatic lesion resection, M-CLM patients also had poorer OS and CSS than A-CLM patients (P = 0.044 and P = 0.011, respectively, Figure 4C-D).
Effect of surgical option for the primary lesion on survival in M-CLM
There is also controversy regarding the selection of surgical option for the primary lesion in CLM to date; thus, we further analysed the surgical options in terms of survival in M-CLM. A total of 272 (88.89%, 272/306) M-CLM patients underwent surgery in this study, partial colectomy (26.10%, 71/272) and hemicolectomy or more extensive colectomy (72.06%, 196/272) were the most common options. The results showed that partial colectomy had a similar OS to hemicolectomy or more extensive colectomy (P = 0.240) but better OS than the no surgery group (P < 0.001, Figure 5A). The CSS analyses also showed similar results (Figure 5B).
Prognostic risk factors for survival in M-CLM
Survival for M-CLM is poor, and we need to explore the potential prognostic risk factors for survival for this condition. We analysed the risk factors for OS and CSS of M-CLM by univariable and multivariable Cox proportional hazards regression models in this study. The univariable analysis results showed that black race, pT3-4 stage and surgery for either or both lesions (Either lesion HR = 0.506, 95% CI: 0.349-0.734; both lesions HR = 0.314, 95% CI: 0.198-0.497) were associated with better OS in M-CLM (P < 0.05 for all, Table 2). Black race, pT3-4 stage, and surgery for either or both lesions were also associated with better CSS in M-CLM (P < 0.05 for all, Table 2). The multivariable analysis demonstrated that only surgery type was an independent prognostic factor for better OS, and black race, pT3-4 stage and surgery type were associated with better CSS in M-CLM (P < 0.05, Table 3). .