Prognostic factors and survival disparities in right-sided versus left-sided colon cancer

Right-sided colon cancer (RCC) and left-sided colon cancer (LCC) differ in features and outcomes because of variations in embryology, epidemiology, pathology, and prognosis. This study sought to identify significant factors impacting patient survival through Bayesian modelling. Data was retrospectively analysed from a colorectal neoplasia database. Data on demographics, perioperative risks, treatment, mortality, and survival was analysed from patients who underwent colon cancer surgery from January 2010 to December 2021. This study involved 2475 patients, with 58.7% having RCC and 41.3% having LCC. RCC patients had a notably higher mortality rate, and their overall survival (OS) rates were slightly lower than those with LCC (P < 0.05). RCC stages I–IV consistently exhibited worse OS and relapse-free survival (RFS) than LCC (P < 0.05). Factors like age, BMI, ASA score, cancer stage, and comorbidities had significant associations with OS and RFS. Poor and moderate differentiation, lower lymph node yield, and organ resection were linked to lower survival while receiving chemotherapy; higher BMI levels and elective surgery were associated with better survival (all P < 0.05). Our study reveals key differences between RCC and LCC, emphasising the impact of age, BMI, ASA score, cancer stage, and comorbidities on patient survival. These findings could inform personalised treatment strategies for colon cancer patients.


Inclusion and exclusion criteria
Patients were recruited in this study according to the selection criteria of 18 years old and above, diagnosed with colon cancer, and underwent resection for colon cancer.Patients undergoing trans-anal surgery were excluded.Patients with rectal cancer were excluded.Rectal cancer patients were excluded due to the complicated dataset of multiple neoadjuvant therapies and some patients not undergoing surgery in a 'watch and wait' treatment approach.

Main variables
Data on patient demographics, perioperative risks, treatment, mortality, morbidity, and survival were collected.The database has high levels of accuracy, completeness, and clinician-led patient follow-up 28,29 .Any additional data points were acquired directly from patients' hospital records.Based on the embryological origin of the involved colon, as previously reported by Meguid et al. in 2008, patients were divided into two groups according to the location of their tumour: RCC (tumours between the appendix and the splenic flexure) including appendix, caecum, ascending colon, hepatic flexure, and transverse colon, and LCC (tumours from the splenic flexure to the rectosigmoid junction) encompassing splenic flexure, descending colon, sigmoid colon, and rectosigmoid 30 .Urgent surgery was also defined as surgery happening within the same urgent hospital admission, and emergency surgery was defined as surgery within the same day of hospital admission.An anastomotic leak was defined as a clinical or radiological indication of a leak from the anastomosis 31 .As patients' pre-anaesthesia medical comorbidities, the ASA (American Society of Anaesthesiology) score was defined as a measure to decide if someone is healthy enough to tolerate surgery and anaesthesia 32 .Resections were carried out according to the clinical practice guidelines of the Cancer Council Australia on curative resection for colorectal cancer 33 .The patients' return to theatre was similarly defined as an indicator of whether a patient had a surgical procedure/operation and required an unplanned return to the operating theatre during the same episode of admitted patient care 34 .Adenocarcinoma, adenocarcinoma mucinous, adenocarcinoma signet, other tumour, no residual, and dysplastic adenoma were considered tumour types.The total number of regional lymph nodes examined by a pathologist and reported as containing tumour cells considered the lymph nodes positive 35 .Accordingly, the lymph node ratio was defined into four categories (0 to < 0.0825, 0.0825 to < 0.25, 0.25 to < 0.5, and 0.5 to 1).Following established medical criteria, the term 'lymph node yield' was operationally defined as categorising lymph nodes harvested into two groups: those with less than 12 lymph nodes and those with 12 or more lymph nodes.The grade was defined as undifferentiated, poorly differentiated, moderately differentiated, and well-differentiated 36 .OS was determined as the length of time after surgery that the patient survives until the last follow-up date.Relapse-free survival (RFS) was defined as the period from the time of surgery to either death or censoring for patients, excluding any time intervals during which metastasis or recurrence is observed for those who experience such events.It measures the duration during which patients remain free from signs or symptoms of disease recurrence following surgical intervention.Recurrence was defined as the detection of cancer (by means of biopsy or imaging) following surgery, either as a new colon cancer (local recurrence) or in a distant organ (metastasis).Follow-up was left to individual surgeons or public clinic units who adhered to the national (Cancer Council Australia) guidelines on follow-up after curative resection for colorectal cancer 37 .The primary outcomes were patient survival at 1-, 3-and 5-year post-surgery.

Ethical approval
The study protocol was approved and granted by Cabrini Research Governance Office (# 08-01-03-23).All patients on the database gave informed consent.The research registry unique identifying number for this study is #7558 (www.resea rchre gistry.com).

Statistical analyses
Data were expressed as mean (SD), median (percentile25-percentile75) for numeric normal and non-normal variables, and frequency (percentage) for categorical variables, respectively.Fisher's exact test was utilised to compare profiles across RCC and LCC patients.The mortality rates and 95% confidence interval (CI) for RCC and LCC patients were computed.1-, 3-, and 5-year overall survival (OS) and relapse-free survival (RFS) probabilities were computed along with their 95% CI, for RCC and LCC patients and across the patients' subgroups.The differences among subgroups were tested using the log-rank test.
The Bayesian models were compared using the deviance information criteria (DIC) Bayesian index to find the best-fit model on data to ascertain the factors affecting the OS and RFS.Smaller values of DIC signify a better model.All the supported models were considered: log-logistic, generalised gamma, Weibull, Gompertz, log normal and exponential models.In summary, each model was fitted using random-walk Metropolis-Hastings sampling with 12,500 Markov Chain Monte Carlo (MCMC) iterations and 2500 initial burn-in samples, which resulted in a 10,000 MCMC sample size.The parameter convergence was monitored by trace plots, which provide graphical summaries and convergence diagnostics for simulated posterior distributions (MCMC samples) of model parameters.Plots without any trends and specific patterns support the efficiency of the MCMC sampling in parameter estimates.We presented the model's accelerated failure time (AFT) form for a more appealing time ratio interpretation of the model compared to the hazard ratio interpretation 38

Patient cohort
To assess the clinical characteristics, pathology, and outcomes of patients with colorectal neoplasia, we utilised data from the Cabrini Monash colorectal neoplasia database.Data were collected between January 2010 and December 2021, encompassing 2475 patients with RCC and LCC.Of these patients, 58.7% were diagnosed with RCC and 41.3% with LCC.Out of the 1453 RCC patients, 288 deaths were observed, accounting for 19.8% of the patients during a total analysis time at risk of 3917 years.Similarly, out of the total 1022 LCC patients, 164 deaths were observed, accounting for 16.1% of the patients during a total analysis time at risk of 3082 years.

Comparing patients' profiles between RCC and LCC
Several differences were observed when comparing the characteristics of patients with RCC and LCC (Table 1).A higher percentage of females was found in the RCC group, and the prevalence of ASA score 3-4 was also higher in RCC.Anastomosis formation was more common in RCC, but stoma formation was less common.RCC patients also had a lower percentage of other organs resected and a lower incidence of adenocarcinoma.Poorly differentiated tumours were more common in RCC patients, and more RCC patients were in stage I-II.A lower percentage of RCC patients received chemotherapy, and lymph node yield > 12 was more common in RCC patients.In terms of patient comorbidities, RCC patients had a higher percentage of hypertension, ischemic heart disease, and cerebrovascular accidents.RCC patients were more likely to have deficient mismatch repair proteins (dMMR) and were less commonly detected through screening (Table 1).

Survival probabilities in RCC and LCC patients
Based on our analysis, the OS rates of RCC patients at 1-, 3-, and 5-year time points were estimated to be 0.93 (0.91-0.94), 0.80 (0.78-0.83), and 0.72 (0.68-0.74), respectively.These rates were slightly lower than those of LCC patients at the 1-year time point, which was 0.94 (0.93-0.96), but considerably lower at the 3-year (0.86 = 0.84-0.89)and 5-year (0.79 = 0.75-0.82)time points (Table 2).Figure 1 shows that the Kaplan-Maier (K-M) curves of OS for RCC patients were consistently lower than those for LCC patients across all analysed time points.Furthermore, a significant difference was observed between the two groups (P < 0.001).
When sub-grouped by stage, a similar pattern of difference was observed between the right and left sides from stage I to III, and the difference was significant (P < 0.05).However, a steeper slope was observed for stage IV, and the difference between the right and left sides was not significantly different (P > 0.05), as depicted in Fig. 2. For a more detailed comparison of OS between RCC and LCC patients across different predictors, please refer to Supplementary Table S1.
Additionally, when considering the time to recurrence, the 1-, 3-, and 5-year RFS rates were estimated to be 0.90 (0.88-0.91), 0.77 (0.75-0.80), and 0.71 (0.68-0.74), respectively, which were slightly lower than those of LCC patients at the 1-year (0.92 = 0.90-0.93)time point but considerably lower at the 3-year (0.82 = 0.79-0.84)and 5-year (0.78 = 0.74-0.81)time points (Table 2).Figure 3 shows that the K-M curves of RFS for RCC patients were consistently lower than those for LCC patients across all analysed time points.Furthermore, a significant difference was observed between the two groups (P = 0.005).Subgrouping the data by stage revealed a similar pattern of difference between the right and left sides for stages I and III, with a significant difference observed between the two sides (both P < 0.05).However, the difference between the two sides was not statistically significant for stage II.Furthermore, a steeper slope was observed for stage IV, and the difference between the right and left sides was not significantly different (P > 0.05), as shown in Fig. 4. For a more detailed comparison of relapse-free survival between RCC and LCC patients across different predictors, please refer to Supplementary Table S2.

Subgroup analysis in patients undergoing an elective operation
The results presented in Supplementary Table S3, pertaining to patients undergoing elective colon cancer surgery, demonstrate a high degree of comparability with those observed in Table 3. Notable results include patients aged 50-75 showing a TR indicating a higher duration until event occurrence for RFS than those under 50.Elevated BMI levels (> 25 kg/m 2 ) demonstrated higher TRs, signifying longer durations until events for both OS and RFS.Higher ASA scores were associated with significantly shorter durations until events for both OS and RFS.Patients with stoma formed showed a shorter TR for RFS.Mucinous adenocarcinomas exhibited higher TRs for  www.nature.com/scientificreports/RFS, but signet cell adenocarcinomas exhibited lower TRs for both OS and RFS compared to adenocarcinoma.Undifferentiated tumours were associated with longer durations until events for RFS, while poorly differentiated tumours indicated shorter durations.Increasing lymph node ratio categories were linked to significantly shorter durations until events for both OS and RFS.Positive circumferential margins significantly increased durations until events for both OS and RFS.Higher overall stages were associated with significantly lower TRs for both OS and RFS.Patients experiencing medical complications had significantly longer durations until events for RFS,

Subgroup analysis in patients with stages I-III
Supplementary Table S4 shows patients undergoing colon cancer surgery within stages I-III and unveils a multitude of statistically significant associations akin to those observed in Table 3. Noteworthy observations included a shorter TR for both OS and RFS among patients aged 75 years and above compared to those under 50 years.Elevated BMI levels (30 and higher) were associated with longer durations until events for OS and RFS.Higher ASA scores were linked to significantly shorter durations for both OS and RFS.CVA was linked with a significantly shorter TR for OS.Patients with stoma formed and with synchronous organ resection showed shorter TRs for both OS and RFS.Signet cell adenocarcinomas were associated with lower TRs for both OS and RFS Increasing lymph node ratio categories were significantly associated with shorter durations for both OS and RFS.Overall, stage II was associated with significantly lower TRs for both OS and RFS than stage I. Receiving chemotherapy was consistently associated with longer durations until events for both OS and RFS.However, contrary to the whole dataset analyses, patients experiencing medical complications had significantly shorter durations until events for both OS and RFS.

Optimal Bayesian model selection
Supplementary Figure 1 summarises the model selection process based on the deviance information criterion (DIC).While the minimum DIC value was obtained for the log-logistic model, the other models showed only a negligible difference in DIC compared to the log-logistic model.However, due to its superior fit, the log-logistic model was ultimately selected as the optimal model.This decision was made because the log-logistic model had the lowest DIC value among all the models tested, indicating a better overall fit and greater predictive power.Therefore, the log-logistic model was deemed the most appropriate choice for this study.

The effect of predictors on survival time using the optimal model
For our modelling analysis, we utilised 2475 observations, of which 351 (14.2%) resulted in death, and the total time at risk was equal to 6998.2 person-years.To ensure the reliability and accuracy of our findings, we assessed the convergence of the model parameters using trace plots, which indicated that the Markov Chain Monte Carlo (MCMC) sampling method was effective in accurately estimating the model parameters.However, we have not included the trace plot results in this report for brevity.By using this approach, we ensured that our findings were robust and based on reliable estimates.The results are presented in Table 3.
Based on the optimal univariable Bayesian log-logistic model with AFT parametrisation, patients with LCC had significantly higher overall (around 41%) and relapse-free (around 42%) survival than those with RCC.Besides cancer location, other predictors were also found to have a significant relationship with OS and RFS.Patients over 75 had lower OS and RFS than those under 50 in univariable (around half) and multivariable analyses (around three-quarters).Higher BMI levels were linked with higher OS and RFS in univariable (around 21-38%) and multivariable analyses (around 50-85% more) compared to the normal BMI level.However, significantly poorer OS (53%) and RFS (55%) were observed in the underweight patients compared to the normal BMI level.Emergency and urgent operations showed lower OS (around 75% and 67%) and RFS (around 76% and 68%) than elective operations, only in univariable analysis.
Conversely, the ASA score was inversely related to OS and RFS, with higher scores showing a significant decreasing trend in survival probabilities, both in univariable (ending in around one-tenth for both OS and RFS) and multivariable analyses (ending in around one-fifth and one-third for both OS and RFS).Moreover, patients who underwent anastomosis had higher OS and RFS in univariable (around 3-4 times) and multivariable analyses (around 47% more).However, patients with a stoma had lower OS and RFS only in univariable analysis (around half and one-third).Additionally, patients with organ resection had lower OS and RFS in univariable (around one-quarter and one-third) and multivariable analyses (around 30% less).Return to theatre was associated with around 30% lower OS and RFS, only in univariable analysis.
The study found poor and moderate differentiation was associated with lower OS and RFS in univariable and multivariable analyses.In addition, a higher overall stage was inversely related to OS and RFS, with increasing stages showing a significant decreasing trend in survival probabilities in both analyses.
While chemotherapy was initially linked to lower OS and RFS in the univariable analysis, adjusting for other predictors in the multivariable analysis showed around 78% higher survival probabilities.Lymph node yield < 12 was associated with lower OS and RFS probabilities, ranging from 29 to 52%.On the other hand, higher lymph node ratios were related to lower OS and RFS in both types of analyses, with decreases of around 88% and 73% in OS and around 93% and 80% in relapse-free survival.Patients who experienced surgical, medical, or any complications had around half the overall relapse-free survival probabilities in the univariable analysis, with medical complications also being related to lower survival probabilities in both analyses.
Moreover, length of hospital stay (LOS) was inversely related to OS and RFS, with increasing LOS quartiles showing a significant decreasing trend in survival probabilities in univariable (ending in around one-fifth) and multivariable analyses (around 40% less).Higher T-stage and N-stages were also associated with lower OS and RFS in univariable analyses, with a decreasing trend observed in both types of survival.
In univariable analyses, M-stage was inversely related to OS and RFS, with decreases ranging from around 30% to 90%.Some comorbidities, such as congestive heart failure, ischemic heart disease, myocardial infarction, peripheral vascular disease, cerebrovascular accident, and hypertension, were inversely associated with OS and RFS based on univariable or multivariable analyses, with survival rates ranging from TR = 0.5 to TR = 0.83.Lymphovascular invasion was linked with 65% lower OS and 73% lower relapse-free survival based on univariable analysis, while positive circumferential margins were linked with 84% lower OS and 89% lower relapse-free survival based on univariable analysis.Interestingly, detection through screening was associated with higher OS and RFS in both univariable (around 3 and 4 times) and multivariable analyses (around 22% and 48% more) (Table 3).

Discussion
The study analysed data from the Cabrini Monash colorectal neoplasia database, comparing outcomes of rightsided colon cancer (RCC) and left-sided colon cancer (LCC) after surgery in 2475 patients.RCC patients had higher mortality rates and lower overall survival (OS) and relapse-free survival (RFS) compared to LCC patients.Bayesian log-logistic modelling identified LCC patients as having significantly better OS and RFS.Various www.nature.com/scientificreports/predictors such as age, BMI, cancer stage, and comorbidities impacted OS and RFS.The findings were consistent with previous research 10,21,22 .Several studies have addressed the differences between RCC and LCC in survival outcomes [21][22][23] .Populationbased studies have shown variations in the OS rates between different stages of patients with RCC and LCC 23 .Despite our results, this study involving a large group of patients demonstrated longer OS in stage I and II RCC compared to stage I and II LCC 23 .Another study focusing on older patients reported longer OS for RCC in stage II but worse OS for RCC in stage III 39 .However, in line with our results, another study found that OS was worse for stage II and III RCC than for stage II and III LCC 30 .An additional investigation encompassing a substantial patient cohort (n = 167,606) and employing propensity score matching to equate LCC and RCC patients revealed enhanced OS in patients with LCC across all disease stages (Duke's A HR = 0.845; Duke's B HR = 0.947; Duke's C HR = 0.783).These observations align with our study's findings regarding OS; however, it is noteworthy that the effect sizes observed in this study were of a relatively lesser magnitude than ours 21 .In addition, like our results, the age at diagnosis for RCC is older than for LCC, and a relatively small proportion of patients with stage I-III colon cancer have been found to die from the disease 40 .Our study confirms the conclusion of Shida et al. that tumour location in stage III colon cancer patients should be a stratification parameter for prognosis 41 .In a comprehensive systematic review encompassing 87 studies, a discerning analysis was undertaken to elucidate the myriad factors contributing to variations in the clinical presentation and outcomes of CRC, contingent upon its subsite.These disparities can be ascribed to the distinctions in anatomical characteristics and the interplay of acquired and hereditary factors, culminating in divergent molecular pathways implicated in the pathogenesis of CRC 9 .Our study also observed variances between RCC and LCC, affirming the disparities in the outcomes between RCC and LCC.
Our study showed that collecting more than 12 LN benefited OS and RFS and that the greater the LNR, the worse the OS and RFS.This corroborates with previous studies showing that adequate LN harvest leads to better survival 42 , and another study suggested that extended lymphadenectomy in colon cancer patients would be recommended especially for left-sided cancers 43 .
In recent years, Bayesian modelling has gained popularity in cancer patients' survival analysis.For example, a study conducted to evaluate the best model for CRC survival analysis reported that the Bayesian log-normal model with AFT parametrisation provided the most accurate results and has found that gender, age at diagnosis, T-stage, N-stage, tumour size, grade of differentiation, and the number of chemotherapies were significant predictors of recurrence outcome 44 .Furthermore, age at diagnosis, metastasis to other sites, T-stage, grade of differentiation, tumour size, and the number of chemotherapies were significant predictors of death without recurrence 44 .Age at diagnosis and the number of chemotherapies were also significantly associated with death after recurrence 44 .Similarly, our study found that non-modifiable factors such as age and cancer stage and modifiable factors such as BMI, ASA score, and comorbidities were significant predictors of OS and RFS in colon cancer patients.
This study conducted a comprehensive analysis of clinical characteristics, pathology, and outcomes of patients with RCC and LCC after curative resection, using data entered prospectively into a large custom-designed colorectal neoplasia database.This study included more patients than others 16,17 , strengthening the analysis's statistical power.In addition, the study was conducted in private and public Victorian centres, assuring the generalisability of the findings.This study used a Bayesian approach to explore the relationship between various predictors and survival outcomes, which has advantages over frequentist approaches, especially in the presence of high correlations and multiple predictors and represents a significant advancement in methodology.The study identified several key variables, such as age, BMI, ASA score, cancer stage, synchronous organ resection and comorbidities such as ischemic heart disease, myocardial infarction, peripheral vascular disease, cerebrovascular accident, and hypertension, that had significant relationships with OS and RFS, providing important insights for clinical decision-making.The subgroup analysis by stage revealed a consistent pattern of difference between RCC and LCC for stages I-IV, which further supports the validity of the study results.
Additionally, this approach goes beyond conventional statistical methods by employing advanced Bayesian statistical techniques, allowing for a more nuanced exploration of risk factors and their impact on patient survival.This study also examined a wide range of risk factors, including demographics, perioperative risks, treatment modalities, and comorbidities.This comprehensive analysis delves deeper into the complex interplay of variables that influence patient outcomes, providing a more thorough understanding of the unique challenges posed by RCC and LCC.Therefore, this study confirms the differences between RCC and LCC and offers a more comprehensive and nuanced perspective identifying modifiable and non-modifiable factors supported by a rigorous methodology and a wealth of data.These distinctive attributes contribute significantly to the scientific discourse surrounding these distinct forms of colon cancer.
The study has a few limitations to consider when interpreting the results.First, the study did not account for potential confounding factors, such as socioeconomic status, ethnicity, and lifestyle factors, which may impact survival outcomes.Second, the study did not consider the impact of all types of therapies on survival outcomes, which may significantly affect the results.Third, the study did not include quality-of-life measures or patientreported outcomes (now being collected 45 ), which are important considerations for clinical decision-making.Fourthly, patients not undergoing primary tumour resection for stage IV disease are currently not entirely captured on the database.Finally, the study did not address the impact of all types of molecular differences between RCC and LCC, which may have important implications for treatment and prognosis 10 .
In conclusion, this study found that patients with RCC had a significantly higher mortality rate and slightly lower OS rates than those with LCC.The Bayesian log-logistic model was found to be the most appropriate model for the study, and based on this model, patients with LCC had significantly higher OS and RFS rates than those with RCC.Age, BMI, ASA score, cancer stage, and comorbidities such as ischemic heart disease, myocardial infarction, peripheral vascular disease, cerebrovascular accident, and hypertension were also found to

Figure 2 .
Figure 2. Overall survival probability across left and right colons by stage.Log-rank test, Stages 1 to III; p < 0.05, Stage IV; p > 0.05.