This was a retrospective analysis of data from the SACR. The registry collates cancer-related information, including the type of cancer, date of first and subsequent primary cancer diagnoses, cancer sites and morphology. It also includes other relevant details such as place of residence, sex, usual occupation, birth and death dates, and cause of death where available. The SACR collects information on all invasive cancers but does not collect data on non-melanoma skin cancers, except for squamous cell carcinoma of the lip and the skin of the genitalia and perineum and basal cell carcinoma of the skin of the genitalia and perineum.

The primary sources of cancer information for SACR include reports from pathology laboratories, hospitals, radiology departments, and other supplementary sources such as clinicians. The SACR also obtains death data from the registry of births, deaths, and marriages, including the cause and date of death, shared through death certificates or electronic records. The SACR reviews, verifies, codes, and enters the primary cause of death as either cancer or non-cancer. If the primary cause of death is cancer, the registry codes the site and morphology based on the International Classification of Diseases for Oncology, 3rd edition (ICD-O-3) coding rules. For coding causes of death, the SACR used ICD-9 (International Classification of Diseases, 9th edition) until 2011, ICD-10 (10th edition) from 2011 to 2014, and ICD-O-3 since 2015. Available tools facilitate both automated and manual mapping and conversion between various disease coding systems and their versions, including SEER*Stat software, International Association of Cancer Registries (IARC)/International Agency for Research on Cancer (IACR) tools, and ICD-O-3/World Health Organisation conversion tools (Ferlay J et al. 2005).

All cases with an invasive CRC with ICD-O-3 topography codes of C18 to C20 and C21.8, diagnosed between 1 January 1982 and 31 December 2017, were extracted from the SACR. To minimise the potential confounding effects of the COVID-19 pandemic-related disruptions in cancer diagnosis, treatment, and follow-up, and to ensure data consistency and quality (Canfell et al. 2023), the analysis was restricted to data from the pre-pandemic period. Individuals diagnosed with invasive CRC who met the following inclusion criteria were included: (1) survived at least two months after index CRC diagnosis, (2) had no invasive cancer prior to CRC, (3) were aged 20–89 years old at the time of index CRC diagnosis, and (4) had complete information on date of cancer diagnoses, age at diagnosis, and death date and cause of death if deceased. Cases of index CRC diagnosed as primary colorectal sarcoma, lymphoma, or leukemia and those diagnosed with synchronous cancers within two months of the index CRC diagnosis were excluded. The eligible study population were followed until 31 December 2019, at least two years from the diagnosis of index CRC.

According to the IARC/IACR rules, MPCs are defined as the presence of histologically distinct types of invasive cancers originating in a primary site or tissue. These cancers should not be extensions, recurrences, or metastases of other primary site cancers (Report 2005). While the IARC rules for defining multiple primaries are not time-dependent, invasive cancers diagnosed within two months of the index CRC diagnosis were excluded to minimise detection bias (Ye et al. 2018). Causes of death were categorised as index CRC, MPCs, and non-cancer related. For cases where cancer was the cause of death, the primary cause of cancer-specific death was identified based on the topography and histology code of the cancer following ICD-O-3 guideline.

The outcome measures assessed in this study were the cumulative mortality and the risk of cancer-specific mortality associated with the presence of MPC compared to cancer mortality in the general population. In addition, the study assessed the effect of MPC on all-cause mortality among individuals who were first diagnosed with CRC.

As described previously(Deng et al. 2022; Gonçalves et al. 2024), for individuals who were not diagnosed with MPC, the time at risk was defined as the time in years from the diagnosis of the index CRC to death or the end of the follow-up period. Whereas for the group with MPC, the time at risk began at the diagnosis of the first MPC and continued until the date of death or end of the follow-up period, whichever came first. Participants were censored at the date of death from non-cancer causes or on December 31, 2019, whichever occurred first.

Cumulative all-cause and cause-specific mortality rates were estimated using a cumulative incidence function. The cumulative mortality from MPCs under the competing risks was estimated at 37 years after the index CRC diagnosis, with deaths from any non-MPC causes treated as competing events, as previously described (Sung et al. 2022). The indirect standardisation method was employed to standardise cancer-specific mortality rates using age, sex, and cancer-specific mortality rates from the South Australian population between 1982 and 2019 (Naing 2000). The excessive mortality risk associated with MPCs was assessed for both overall and cancer-specific mortality, reported as the SMR and the absolute excess mortality (AEM) (Sung et al. 2022). The SMRs and their 95% confidence intervals (CI) were calculated by dividing the observed number of deaths by the expected number of deaths, assuming that MPC-associated mortality follows a Poisson distribution. Similarly, AEMs and their 95% CIs were calculated as the difference between observed and expected deaths, divided by the person-years at risk and multiplied by 10,000, assuming a normal distribution of differences. SMRs and AEMs were calculated for overall MPC-associated mortality and cancer-specific mortality, and the results were stratified by sex. The proportion of deaths attributed to CRC, MPCs, and non-cancer causes was calculated by dividing the number of deaths from each cause by the total number of deaths, as previously described (Liang et al. 2025). Stata version 18 software was used to analyse the cumulative mortality rate, as well as SMRs and AEMs.

Propensity score analysis was used to ensure group comparability before estimating the effect of MPC on all-cause mortality. The propensity scores were generated based on basic demographics, calendar year of index CRC diagnosis, and location of the index cancer (colon vs rectum) to balance the distribution of covariates between individuals diagnosed with MPC and those without MPC diagnosis. The inverse probability of treatment weighting (IPTW) method was then applied to assign weights to each individual based on their likelihood of being assigned to the MPC group, given the observed covariates. The propensity scores were estimated using the “WeightIt” package in R. A Cox proportional hazards regression model was then fitted using the weighted dataset to estimate hazard ratios (HR) and 95% CI for all-cause mortality. The propensity score weights generated by the IPTW method were included in the model to account for covariate imbalances. Analysis was performed in R using the “survival” package. The proportional hazards assumption was tested using the global Schoenfeld residual test, which indicated a violation for the variable age at the diagnosis of index CRC, suggesting a time-varying effect on all-cause mortality. To address this, a Cox regression with time-varying spline analysis was fitted, treating age as a time-dependent covariate. Various spline methods, including natural spline, B-spline, penalised spline, and restricted cubic spline, were applied to check the robustness of the estimates. The estimates were consistent, and for ease of interpretation, the Cox proportional hazards model with a natural spline was applied. Knots were automatically determined using the default settings in R software and placed at the 25th, 50th, and 75th percentiles of the age distribution.

To assess the influence of different time cut-offs for defining MPC on the estimates, sensitivity analysis was conducted by calculating mortality rates, SMRs, and HRs using two distinct time intervals: two months and 6 months following the index CRC diagnosis.