Overall survival and prognostic factors in metastatic colorectal cancer: a 10-year retrospective cohort study in Tunisia

¹Department of Medical Oncology, Fattouma Bourguiba Hospital, Monastir, Tunisia, ²Department of Visceral Surgery, Fattouma Bourguiba Hospital, Monastir, Tunisia

^&Corresponding author
Mabrouk Nada, Department of Medical Oncology, Fattouma Bourguiba Hospital, Monastir, Tunisia

Introduction    

Colorectal cancer is the third most frequently diagnosed malignancy and the second leading cause of cancer-related mortality both in Tunisia and worldwide, representing a major public health burden, with a continuously increasing incidence in Tunisia characterised by an annual percentage change of +4.5% [1]. At the time of diagnosis, approximately 20% of patients present with metastatic disease, and an additional 25% will develop metachronous metastases during follow-up [2], contributing substantially to cancer-related mortality. Despite significant advances in systemic treatments, including combination chemotherapy and targeted therapies, metastatic colorectal cancer (mCRC) remains a biologically heterogeneous disease with highly variable survival outcomes.

Numerous studies have investigated potential prognostic factors, including primary tumour location [3], tumour burden, lymph node involvement, baseline carcinoembryonic antigen (CEA) level [4], and therapeutic interventions, to better stratify patients and optimise treatment strategies. However, in resource-limited settings such as Tunisia, access to molecular profiling and innovative therapies remains inconsistent, and available local data on survival outcomes and prognostic determinants are scarce and heterogeneous, limiting their applicability to routine clinical practice. Therefore, the present study aimed to evaluate overall survival and to identify clinicopathological and therapeutic prognostic factors associated with survival in patients with metastatic colorectal cancer treated at a tertiary referral center in Tunisia.

Methods     

Study design and setting: this retrospective observational cohort study was conducted at the University Hospital Fattouma Bourguiba, Monastir, Tunisia, a tertiary referral center. Medical records of patients treated between January 2012 and December 2022 were reviewed using the institutional oncology database.

Study population and sampling: eligible patients were adults with histologically confirmed colorectal adenocarcinoma (colon or rectum) and documented metastatic disease, either synchronous (present at initial diagnosis) or metachronous (occurring during follow-up). Additional inclusion criteria included the presence of measurable metastatic disease based on radiological assessment and receipt of systemic anticancer therapy. Patients were excluded if essential clinical or follow-up data were missing or if they had received fewer than two cycles of systemic treatment. For patients with synchronous metastases, the index date was defined as the date of initial colorectal cancer diagnosis. For those with metachronous metastases, the index date corresponded to the date of metastatic recurrence. All patients meeting the eligibility criteria during the period of the study (January 2012 to December 2022) were included (n=120). This approach was adopted to reflect our real-world experience in mCRC management. In this observational study, a formal sample size was not calculated because all eligible patients during the 11-year study period were included. Patients with insufficient data or who had received fewer than 2 cycles of systemic therapy were excluded, and the final cohort consisted of 104 patients.

Data collection: demographic and clinical factors were age (<50 vs >=50), sex (male vs female), diabetes mellitus (yes/no), hypertension (HTN) (yes/no), and performance status based on the ECOG scale (0-1 vs >=2). Laboratory parameters comprised baseline carcinoembryonic antigen (CEA) levels recorded at the time of metastatic diagnosis and categorised as normal (≤ 5ng/mL) or elevated (> 5 ng/mL) based on the established clinical threshold. Pretreatment haemoglobin level was dichotomised as <10 or >=10 g/dl. Tumour location was classified by organ site (colon vs rectum) and by sidedness as right-sided (from the cecum to the transverse colon) or left-sided (from the splenic flexure to the rectum) or synchronous/bilateral (primary lesion in both right and left colon). Metastatic disease characteristics encompassed timing for metastases (synchronous, defined as metastases detected at the time of initial diagnosis, vs metachronous, defined as metastases identified after treatment of the primary tumour), number of metastatic sites (single or multiple), and specific metastatic patterns (liver-only or lung-only metastases). Pathological features included histological grade of differentiation (well/moderately differentiated vs poorly differentiated), nodal status according to TNM 8^th edition classification (N0, N1, N2 or NX when not assessed), RAS mutation status (wild type, mutated or unknown), and BRAF mutation status (wild-type, mutated or unknown). Molecular profiling for RAS and BRAF mutations has been available since 2015 and has not consistently been accessible.

Treatment-related variables included primary tumour resection and metastatic site resection. Systemic therapy comprised first-line chemotherapy regimens, which were grouped for descriptive purposes into the following categories (FOLFOX/CapOX (oxaliplatin-based), FOLFIRI/CapIRI (Irinotecan-based), Capecitabine, FOLFORINOX, and LV5FU(Fluorouracil/leucovorin). Targeted therapies included Cetuximab or Bevacizumab. Treatment response after three cycles of first-line chemotherapy was assessed radiologically and defined as objective response (complete or partial response vs no objective response) according to RECIST criteria.

Due to insufficient data, certain variables were excluded from prognostic analysis but were retained in baseline characteristics to document population characteristics and highlight gaps in molecular profiling availability: Targeted therapy was analysed as a binary variable (any targeted therapy (cetuximab or bevacizumab) vs none). BRAF mutation status was not included because only 5 patients had a known BRAF status. The chemotherapy regimen was analysed as oxaliplatin based (FOLFOX/CapOX) vs Irinotecan-based (FOLFIRI/CapIRI). All other variables described above were included in the univariate prognostic analysis.

Follow-up: the follow-up cut-off date was February 2024. Vital status was determined by medical record or by telephone contact for patients lost to follow-up. Patients lost to follow-up were censored at their last known contact date.

Outcome measures: the primary endpoint was overall survival (OS), defined as the time from the diagnosis of metastatic disease to death from any cause or the date of last follow-up. Secondary objectives included the identification of clinicopathological and therapeutic prognostic factors associated with OS.

Statistical analysis: to minimize selection bias, all patients meeting the inclusion criteria throughout the 11-year study period were enrolled. To address missing data bias, we reported the proportions of missing values in Table 1, and a complete case analysis was performed in the Cox multivariate regression model. Continuous variables were summarised using means with standard deviations or medians with ranges, as appropriate, while categorical variables were reported as frequencies and percentages. Overall survival was estimated using the KaplanMeier method, and survival curves were compared using the log-rank test. Univariate analyses were performed to identify factors associated with OS. variables with p≤0.20 in the univariate analysis were included in an initial multivariate Cox model. A manual backward elimination approach was then applied: the least significant variable (highest p-value >0.05) was sequentially removed, and the model was refitted until only variables with p<0.05 remained. The final model included only independent prognostic factors significantly associated with OS. Results were expressed as hazard ratios (HRs) with 95% confidence intervals (CIs). A two-sided p-value <0.05 was considered statistically significant. All statistical analyses were conducted using SPSS software, version 25.0 (IBM Corp., Armonk, NY, USA).

Ethical considerations: in accordance with institutional policy, informed consent was not required, as all data were collected retrospectively and anonymised.

Results     

Patient selection and baseline characteristics: during the study period, 310 patients identified from our institutional database were evaluated for eligibility. The study flow is illustrated in Figure 1. Of these, 190 patients (61.3%) did not meet the inclusion criteria: 173 patients had non-metastatic colorectal cancer, 13 patients (4.2%) did not receive systemic therapy, and four (1.3%) had other histological tumour types. Of the 120 remaining patients, 16 were further excluded: 10 patients (3.2%) received fewer than two cycles of treatment, and 6 patients (1.9%) had incomplete data. The final cohort included 104 patients. The mean age was 60.5±12.13 years (median 61; range: 32-87), and 51 patients (49%) were male. Comorbidities included diabetes in 33 patients (31.7%) and hypertension (HTN) in 28 (26.9%). A good performance status (ECOG 0-1) was observed in 76.9% of the population. Carcinoembryonic antigen (CEA) was elevated (> 5 ng/ml) in 60 patients (57.7%), and anaemia (haemoglobin <10 g/dL) was present in 18 (17.3%). Regarding the primary tumour location, 72.1% had colon cancer and 27.9% had rectal cancer. Tumour sidedness was classified as right-sided in 19.2% and left-sided in 77.8% of patients; three patients (1.9%) presented with synchronous lesions involving both right and left colon. Synchronous metastases were present in 73.1% of cases, and single-organ metastatic disease was observed in 60.6%. The most common metastatic sites were the liver (40.4%), peritoneum (10.6%), and lung (7.7%). Histologically, most tumours were well or moderately differentiated (82.7%). Pathologically, 33.7% of patients were classified as pN2 and 17.3% as pN1 according to lymph node involvement. RAS mutational status, available for a subset of patients (46.1%), was wild-type in 28.8% of cases and mutated in 17.3%. BRAF status was only available for 5 patients (4.8% of cases) and was wild-type. Regarding therapeutic interventions, 60.6% of patients underwent primary tumor resection, and 17.3% had metastatic site resection. The most frequently administered chemotherapy regimens were FOLFOX or CapOX (51.9%) and FOLFIRI or CapIRI (29.8%), and less frequently capecitabine (13.5%). Targeted therapy, primarily bevacizumab or cetuximab, was used in 19.2% of cases (9.6% each). An objective response after 3 cycles of first-line chemotherapy was observed in 27.9%. Detailed patient characteristics are summarised in Table 1.

Overall survival: safter a median duration of follow-up of 16 months (range: 164 months),76% of patients (n=78) had died, and 3 patients were lost to follow-up. The median OS was estimated at 18 months (CI 95%: 14.06-21.93). The estimated survival rate at 24 months was 31.7% (Figure 2). The presence of comorbidities (diabetes, HTN), performance status, haemoglobin level, baseline CEA levels, lung-only metastases, nodal involvement, primary site resection, metastatic site resection, and objective response after 3 cycles of first-line therapy were factors associated with survival in the univariate analysis (Table 2).

Multivariate analysis showed that diabetes (HR= 1.78; CI 95%: 1.08-2.91; p=0.02) and a high CEA baseline level (HR = 2.77; CI 95%: 1.634.71; p < 0.00) were associated with increased mortality. Surgery of the primary tumour and resection of metastatic disease were each associated with improved outcomes (HR= 0.34; CI 95%: 0.20-0.58; p=0.00 and HR=0.40; CI 95%: 0.18-0.91; p=0.03, respectively). The use of targeted therapy also conferred better survival (HR= 0.42; CI 95%: 0.20-0.86; p=0.017). In addition, having an objective response after 3 cycles of first-line therapy was a predictor of favourable results (HR=0.40; CI 95%: 0.22-0.71; p=0.002). Results of multivariate analyses are detailed in Table 3. Survival curves of independent prognostic parameters are presented in Figure 3.

Discussion     

In this retrospective real-world study conducted in a resource-limited setting, we identified a median OS of 18 months in patients with mCRC. Diabetes and elevated baseline CEA levels were independently associated with worse survival, whereas surgical resection of the primary tumour, resection of metastatic sites, and the use of targeted therapies were associated with improved outcomes. These findings highlight the prognostic relevance of clinical biological markers and treatment-related factors in mCRC within the Tunisian context.

Colorectal cancer remains a major global health burden with substantial mortality. At the time of diagnosis, 1530% of patients present with metastatic disease, and up to half will develop distant metastases during follow-up [5]. Metastatic colorectal cancer represents a biologically heterogeneous disease characterised by diverse clinicopathological features and a highly variable prognosis. Unlike localised colorectal cancer, the prognostic landscape of mCRC remains incompletely defined, particularly in patients with unresectable disease. Moreover, few studies from low-resource countries, including Tunisia, have specifically examined prognostic determinants, resulting in limited locally applicable evidence.

In our cohort of 104 patients treated between 2012 and 2022, the median OS was 18 months (95% CI: 14.0621.93) after a median follow-up of 16 months. This survival estimate is lower than those reported in contemporary randomised trials and large observational cohorts from high-income countries, where median OS now ranges between 24 and 36 months [5]. These discrepancies likely reflect inequalities in access to molecular testing, targeted therapies, and multidisciplinary management, particularly during the earlier years of our study period. Comparable survival outcomes have been reported in other low-resource settings, including an Indian cohort reported by Sharma et al. [6], as well as in previous Tunisian studies demonstrating median OS values between 18 and 23 months [7,8], supporting the external validity of our findings.

Diabetes was associated with worse survival in our study (HR=1.78; p=0.02), which is consistent with previous findings. Brown et al. showed in a large randomised trial that diabetes was an independent adverse prognostic factor in mCRC [9]. Similar associations were reported in retrospective series by Ottaiano et al. and Miranda Baleiras et al. [10,11]. This negative impact may be due to multiple mechanisms, including chronic inflammation, hyperinsulinemia, insulin resistance, and increased susceptibility to treatment-related complications [11]. Although hypertension was associated with survival in the univariate analysis, it was not retained as an independent prognostic factor. This may be explained by its strong correlation with diabetes. After adjustment for other factors, hypertension was not correlated with survival. This finding is consistent with several studies showing diabetes rather than hypertension alone as the metabolic comorbidity associated with cancer mortality [12].

Lung-only metastases were associated with better survival in univariate analysis, but were not retained in the multivariate model. While this finding appears inconsistent with studies such as the large cohort by Rumpold et al. [13], who reported lung-only metastases as an independent favourable prognostic factor, the discrepancy can be explained by the fact that our multivariate model already included metastatic site resection (HR=0.43; p=0.045), which presents the major driver of the favourable prognosis associated with lung-only disease.

Tumour sidedness and RAS status did not reach statistical significance in our study (p=0.08 and p=0.15, respectively), despite accumulating evidence supporting their prognostic relevance in metastatic colorectal cancer. Recent meta-analyses, including that of Yin et al. [14], support this relevance. Similarly, RAS mutations, while primarily recognised as a predictive biomarker for anti-EGFR therapy resistance, have also been shown to have independent prognostic significance in multiple studies, including those by Ianniello et al. and Boland et al. [15,16]. Several factors may explain these results. First, our sample size(n=104) may be insufficient to detect some prognostic effects after adjustment. Second, RAS status was not available for 53.8% of patients, which may have influenced the assessment of this biomarker as a prognostic factor. Third, the limited use of targeted therapies in our cohort (19.2%) could explain the lack of prognostic significance of RAS status, as its impact is most pronounced in patients receiving anti-EGFR treatments. Furthermore, other key molecular markers with established prognostic value, such as BRAF mutational status and Microsatellite Instability (MSI), were not routinely tested in our cohort due to limited availability in our setting. This lack of data prevents a comprehensive molecular stratification of our patients. These results highlight the challenges of retrospective studies with incomplete molecular data and underscore the importance of systematic biomarker testing in the management of metastatic colorectal cancer.

Baseline carcinoembryonic antigen emerged as a strong independent prognostic factor in our analysis. CEA is a glycoprotein frequently overexpressed in colorectal cancer and is involved in tumour cell adhesion and metastatic dissemination. Elevated CEA levels have consistently been associated with poorer survival outcomes. A systematic review by Petrelli et al. demonstrated an increased risk of mortality associated with elevated CEA in patients undergoing resection of colorectal liver metastases [17]. In the unresectable metastatic setting, similar prognostic implications have been reported by Liu et al. and Thomsen et al. [18,19]. Notably, CEA cut-off values varied widely across studies, ranging from 5 to 200ng/mL [20,21], reflecting the absence of a universally accepted threshold.

Surgical resection of the primary tumour was associated with prolonged survival in our study. This issue remains controversial, and several meta-analyses and controlled studies have explored its impact in asymptomatic patients with unresectable mCRC. A meta-analysis by Shu et al. demonstrated improved survival with primary tumour resection in this population [22], whereas Zhang et al. and Huang et al. did not observe a significant OS benefit [23,24]. Furthermore, Liang et al. reported that upfront primary tumour resection was associated with improved OS only in patients who did not receive upfront chemotherapy [25], suggesting that patient selection and treatment sequencing play a critical role.

Resection of metastatic sites was independently associated with improved OS in our cohort. These findings are consistent with a large population-based study by Siebenhüner et al., which reported a significant survival benefit associated with metastasectomy (HR = 0.47) [26]. Current international guidelines, largely based on retrospective evidence, recommend surgical resection of metastatic disease when technically feasible and oncologically appropriate [27,28], reinforcing the importance of a multidisciplinary approach.

The administration of targeted therapies was also associated with improved survival in our study (HR = 0.42). Anti-EGFR agents such as Cetuximab or panitumumab combined with chemotherapy are standard treatments for patients with RAS wild-type, left-sided mCRC [27,28]. A meta-analysis by Zmuc et al. demonstrated a significant OS benefit associated with targeted therapy (HR = 0.88) [29]. In contrast, bevacizumab is commonly used in RAS-mutated tumours; however, a meta-analysis by Baraniskin et al. showed a consistent benefit in progression-free survival but an inconsistent effect on OS [30]. Limited access to molecular profiling and targeted agents in our setting likely influenced treatment patterns and outcomes.

Data on survival predictors in metastatic colorectal cancer from Tunisia remain scarce. In this context, observational real-world studies contribute valuable insights into clinical and therapeutic prognostic factors and help inform decision-making. A strength of our study lies in the inclusion of clinically relevant variables in the multivariate analysis, including performance status, tumour sidedness, haemoglobin level, timing of metastases, and baseline CEA level.

Nevertheless, several limitations must be acknowledged. The retrospective and monocentric design may limit generalizability. Incomplete molecular data, including RAS mutational status as well as the lack of BRAF and MSI assessment, represent a significant limitation that prevented comprehensive evaluation of these established prognostic and predictive factors. Additionally, the long inclusion period (20122022) resulted in a heterogeneous cohort reflecting evolving treatment standards. Despite these limitations, our findings corroborate existing evidence, highlight the disparities in access to molecular testing and targeted therapies, and support the need for personalised treatment strategies in resource-limited settings.

Conclusion     

This retrospective study provides valuable insights into survival outcomes and prognostic factors in metastatic colorectal cancer within a resource-limited setting. Overall survival in our cohort remained lower than that reported in high-income countries, reflecting disparities in access to molecular profiling and modern systemic therapies. Diabetes and elevated baseline carcinoembryonic antigen levels were independently associated with poorer survival, whereas surgical resection of the primary tumour and metastatic sites, administration of targeted therapies, and objective response to first-line treatment were associated with improved outcomes. These findings underscore the importance of optimised patient selection, multidisciplinary management, and improved access to molecular testing and innovative treatments to enhance survival in metastatic colorectal cancer, particularly in low-resource countries.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

All authors participated in the writing of this manuscript; they all read and approved the final version.

Tables and figures     

Table 1: patient, tumour, and treatment characteristics
Table 2: univariate analyses evaluating prognostic factors for overall survival
Table 3: multivariate analyses evaluating prognostic factors for overall survival
Figure 1: flow diagram of patient selection
Figure 2: Kaplan-Meier curve of overall survival in the studied population
Figure 3: Kaplan-Meier analysis of overall survival based on independent prognostic factors

References