Cervical Cancer Survival in Mali: A 21‐Year Retrospective Cohort Study of Trends, Determinants, and Outcomes at Gabriel Touré Teaching Hospital (2003–2024)

2.1. Study Design and Setting

This retrospective cohort study was conducted at Gabriel Touré Teaching Hospital in Bamako, Mali, and included all women diagnosed with histologically confirmed invasive CC between January 1, 2003, and December 31, 2024. The hospital’s CC management unit, located within the Department of Gynecology and Obstetrics, plays a central role in the national response to CC and serves as a major referral center for diagnosis, staging, and initial management. Because the cohort was hospital‐based, it remains subject to referral and selection bias and should not be considered nationally representative.

According to national guidelines, CC screening is included in the essential healthcare package and should be routinely offered across all health system levels [16]. As illustrated in Figure 1, CC screening activities were introduced in 2001 and have expanded significantly over time. The introduction of radiotherapy in 2014 marked a major milestone in the management of advanced‐stage CC and contributed to improved prognosis [17].

2.2. CC Registry

A hospital‐based CC registry was established in 2003 within the Gynecologic and Breast Cancer Screening and Management Unit at Gabriel Touré Hospital. Data were abstracted from medical records using a standardized form by a team of four or five trained senior medical students each year, all nearing completion of their medical studies, under the supervision of the senior investigator and in accordance with IARC guidance. Every 2 weeks, the senior investigator reviewed the progress of data abstraction with the abstractors. During these review meetings, entries were examined for completeness, internal consistency, duplicate records, and implausible values; identified discrepancies were discussed, and appropriate corrections were made in the database on the basis of the available source documentation. When an inconsistency could not be resolved from the available records, the value was treated as missing or excluded from the relevant analysis. Formal duplicate abstraction and inter‐rater reliability statistics were not performed, which is a limitation of the retrospective registry.

2.3. Study Population

The registry was reviewed to identify all women diagnosed with histologically confirmed invasive CC between 2003 and 2024. Inclusion criteria were admission to the Department of Obstetrics and Gynecology with a new diagnosis of primary CC. Exclusion criteria included recurrent disease and malignancies originating outside the cervix.

2.4. Data Collection

Data were extracted using a standardized collection form. Sociodemographic variables included age, marital status, education level, occupation, region of origin, height, and weight. Gynecologic and obstetric history included menopausal status, parity, gravidity, and contraceptive use.

Clinical presentation variables included symptoms such as hydrorrhea, metrorrhagia, pelvic mass, urinary or rectal signs, and symptom duration. Tumor characteristics comprised FIGO stage, histological type, grade of differentiation, and presence of complications or metastases (urinary tract infection, hydronephrosis, vesicovaginal fistula, cerebral/liver/lung metastases, etc.). Ancillary investigations included complete blood count, creatinine levels, Rhesus typing, pelvic ultrasound, CT scan, and intravenous urography. Treatment‐related variables included surgery, chemotherapy, radiotherapy, and associated complications.

2.5. CC Management

CC treatment at Gabriel Touré follows institutional protocols aligned with international oncology standards, adapted to local resource constraints. Treatment is determined based on FIGO stage. Early‐stage disease (FIGO I‐IIA) is typically managed with radical surgery, often followed by adjuvant chemotherapy and/or radiotherapy. Advanced‐stage disease (FIGO IIB‐IV) is treated primarily with concurrent chemoradiotherapy.

Due to limited infrastructure, Mali has only one radiotherapy center. Treatment delays of 3–9 months are common, necessitating the use of alternative regimens (surgery or chemotherapy alone) when radiotherapy is not immediately available.

Standardized preradiotherapy assessments include clinical examination, complete blood count, renal function, chest X‐ray, and urography. Since 2020, thoracoabdominalpelvic CT has been integrated into the pretreatment workup. External beam radiotherapy is delivered in two phases totaling 70 Gy. Phase I (46 Gy over 23 sessions) targets the uterus and pelvic lymph nodes using four‐field techniques. Phase II (24 Gy over 12 sessions) targets the residual tumor volume. A tumor‐directed boost of 30 Gy (3 Gy × 10 fractions) is administered. Concurrent weekly cisplatin (40 mg/m²) is used as a radiosensitizer.

Treatment categories were based on treatment plans documented in the medical records and cancer registry. Information on treatment completion, adherence, or treatment interruptions was not consistently available throughout the study period and therefore could not be assessed.

2.6. Palliative Care

Palliative care constitutes a critical component of CC management in Mali, especially for women with advanced or nonoperable disease. However, access to formal palliative care remains limited and largely confined to tertiary centers in Bamako.

At Gabriel Touré, palliative care is delivered by gynecologists and nurses with limited specialized infrastructure. Management focuses on symptomatic relief of pain, bleeding, and urinary/rectal complications using oral morphine, antibiotics, catheterization, wound care, and other supportive measures. Access to morphine is inconsistent due to procurement and regulatory barriers. Psychosocial and home‐based palliative services are underdeveloped. Despite the adoption of a national palliative care policy in 2017, its operationalization remains limited due to funding and training gaps. These contextual factors are important for interpreting survival outcomes in this study.

2.7. Posttreatment Follow‐Up

Patients completing curative treatment were scheduled for follow‐up visits at 3, 6, 12, 18, and 24 months, then annually. Follow‐up evaluations included clinical assessment, pelvic examination, and investigations as needed (ultrasound, CT, X‐ray, MRI, or biopsy). To minimize loss to follow‐up, the Chief Nursing Officer conducted monthly phone calls. If the patient could not be reached, contact was made with family members to ascertain vital and clinical status.

2.8. Outcome Definition

The primary outcome was OS, defined as the time (in months) from histopathological diagnosis to death from any cause or last known contact (through a clinic visit, telephone call, or notification). Vital status was obtained from medical records or verified through telephone contact with the patient or family members. Women alive at the last known contact were censored at that date. Cancer‐specific survival could not be assessed because reliable cause‐of‐death information was not consistently available. This outcome reflects real‐world survivorship in a resource‐limited setting, capturing the cumulative impact of clinical stage at diagnosis, treatment availability, and systemic care constraints, including access to palliative services.

2.9. Statistical Analysis

All statistical analyses were conducted using SAS Version 9.4 (SAS Institute Inc., Cary, NC, USA). Categorical variables were summarized using frequencies and percentages, while continuous variables were described using means with standard deviations (±SDs) or medians with interquartile ranges (IQR), as appropriate based on the distribution.

OS was defined as the time (in months) from histopathological diagnosis to the date of death or last follow‐up. Patients who were alive at last contact were right‐censored at that date. The Kaplan–Meier method was used to estimate survival probabilities, and survival curves were compared between groups using the log‐rank test. Median survival time and cumulative survival at 6, 12, and 24 months were reported.

To identify factors independently associated with mortality, a multivariable Cox proportional hazards regression model was constructed to estimate adjusted hazard ratios (AHRs) and their 95% confidence intervals (95% CI). The proportional hazards assumption was tested using Schoenfeld residuals and visual inspection of log–log survival plots. Candidate variables were selected on the basis of clinical relevance and prior evidence and included age, FIGO stage, histologic type, anemia at admission, treatment modalities (surgery, chemotherapy, and radiotherapy), metastases, and delay between symptom onset and admission. HIV status was not included in the Cox regression analysis because women with a confirmed negative test could not be reliably distinguished from women who had not been tested. Variables with p values < 0.20 in univariate analysis were initially retained. Manual backward elimination was then applied; covariates were sequentially excluded when p > 0.05 and their removal altered the coefficients of remaining variables by less than 10%.

Additionally, stratified analyses were conducted to examine survival differences across distinct calendar periods (2003–2009, 2010–2015, and 2016–2024), corresponding to major milestones in national CC control (screening scale‐up, radiotherapy introduction, and the Weekend70 initiative). All tests were two‐sided, and a p‐value < 0.05 was considered statistically significant.

2.10. Follow‐Up and Sensitivity Analysis

Vital status was ascertained through review of medical records, scheduled clinical follow‐up visits, and active telephone contact with patients or their relatives when necessary. Prior to survival analyses, records with implausible negative survival times resulting from inconsistent date information were identified during data validation. Twenty‐nine records (0.9%) were excluded because reliable correction was not possible.

OS was defined as the time from diagnosis to death from any cause or last known contact and was expressed in months.

To evaluate the potential impact of loss to follow‐up on survival estimates, a sensitivity analysis was performed under a worst‐case scenario in which all patients lost to follow‐up were assumed to have died at their last recorded contact. Survival estimates obtained from this analysis were compared with those from the primary Kaplan–Meier analysis.

2.11. Ethical Consideration

This study was approved by the Ethics Committee of the Faculty of Medicine and Pharmacy, University of Bamako (Approval No. 2020/233/CE/FMOS/FAPH). The study was conducted using secondary data extracted from the hospital‐based cancer registry and medical records. Because of the retrospective nature of the study and the use of routinely collected anonymized data, individual informed consent for participation in this research was waived by the Ethics Committee. All data were deidentified before analysis to ensure patient confidentiality and privacy.

3.1. Sociodemographic Characteristics

Among the 3213 women diagnosed with CC, 29 in situ carcinomas and 72 cases with unknown histological type were excluded, leaving 3112 patients with confirmed invasive CC for analysis (Table 1).

Most women were from Bamako (62.2%), followed by Koulikoro (10.8%), Kayes (7.7%), Sikasso (6.9%), Ségou (4.6%), Mopti (3.0%), Tombouctou (0.7%), Gao (0.3%), and Kidal (0.1%). The mean age at diagnosis was 51.3 years, with 56.4% aged 50 or older. The majority were uneducated (79.8%), housewives (85.6%), and married (79.1%), including 60.3% in polygamous unions. Only 17.0% had ever used modern contraceptives. More than half (53.4%) were grand multiparous, and 58.6% were postmenopausal. Most women (90.5%) had a BMI < 30 kg/m², whereas 9.5% had a BMI ≥ 30 kg/m².

3.2. Clinical Presentation and Treatment

Among the 3112 women with invasive CC, metrorrhagia was the most common symptom (65.4%), followed by hydrorrhea (31.6%), urinary symptoms (12.1%), and rectal symptoms (9.0%) (Table 2). Anemia was observed in 48.4%. HIV positivity was recorded for 4.2% of the overall cohort. Pelvic mass was recorded in 2.4%. Squamous cell carcinoma predominated (91.2%), with most patients diagnosed at advanced stages; 60.1% had FIGO Stage IV disease, and 11.4% had distant metastases. Hydronephrosis occurred in 13.9% of cases.

3.3. Kaplan–Meier Survival Analysis

During data validation, 29 records (0.9%) with implausible negative survival times were excluded from survival analyses, leaving 3083 women for Kaplan–Meier and Cox regression analyses (Figure 2).

Follow‐up status was successfully ascertained for most patients. Among the 3083 women included in the survival analyses, 1814 (58.8%) were confirmed dead during follow‐up, 1072 (34.8%) were alive at the last known contact, and 197 (6.4%) were lost to follow‐up.

Kaplan–Meier analysis was performed among 3083 women included in the final survival analyses. Median OS was 15.5 months (95% CI: 14.3–17.1) (Figure 3A‐D). Survival probabilities declined progressively over time, with the steepest decrease observed during the first year following diagnosis.

Survival differed significantly according to age group, with women younger than 40 years exhibiting higher survival probabilities than women aged 60 years and older (log‐rank p < 0.001). Survival also varied across calendar periods of diagnosis, with women diagnosed between 2016 and 2024 demonstrating significantly better survival than those diagnosed in earlier periods (log‐rank p < 0.01).

As expected, survival decreased with advancing FIGO stage (log‐rank p < 0.001). Twelve‐month survival exceeded 70% among women diagnosed at Stage I but was below 20% among those diagnosed at Stage IV.

In the sensitivity analysis assuming that all patients lost to follow‐up had died at their last recorded contact, median OS was 14.1 months (95% CI: 13.1–15.3), and the overall interpretation of the findings remained unchanged.

3.4. Factors Associated With Survival

Multivariable Cox regression identified several independent predictors of mortality (Table 3). Compared with women who had secondary or higher education, mortality risk was significantly higher among women with no formal education (AHR = 1.47; 95% CI: 1.20–1.80) and those with primary education (AHR = 1.57; 95% CI: 1.23–2.00). Menopausal status (AHR = 1.17; 95% CI: 1.04–1.32) and anemia at diagnosis (AHR = 1.60; 95% CI: 1.45–1.76) were also independently associated with increased mortality.

Tumor stage had the strongest association with survival. Compared with women diagnosed at FIGO Stage I, mortality risk increased progressively for Stage II (AHR = 1.76; 95% CI: 1.22–2.56), Stage III (AHR = 2.48; 95% CI: 1.75–3.52), and Stage IV disease (AHR = 4.95; 95% CI: 3.53–6.94).

Compared with palliative care, receipt of radiotherapy alone (AHR = 0.75; 95% CI: 0.63–0.90), chemotherapy alone (AHR = 0.31; 95% CI: 0.23–0.42), radiotherapy plus chemotherapy (AHR = 0.55; 95% CI: 0.48–0.62), and multimodal treatment (AHR = 0.54; 95% CI: 0.37–0.80) were associated with lower mortality.

4.1. Limitations

First, this was a retrospective hospital‐based cohort rather than a population‐based study. Nevertheless, CC diagnosis and treatment services were largely centralized in Bamako during the study period, and Gabriel Touré Teaching Hospital functioned as a major national referral center. Because many women from regional areas temporarily relocated to Bamako during treatment, recorded residence may not accurately reflect geographic origin. Therefore, although the findings should not be interpreted as population‐based estimates, they provide important evidence regarding survival patterns among women who accessed specialized CC care in Mali.

Second, although active follow‐up was conducted through scheduled visits and telephone contact with patients or their relatives, outcome ascertainment partly relied on indirect reporting. Consequently, some degree of outcome misclassification may have occurred. However, only 197 women (6.4%) were lost to follow‐up among the 3083 included in the survival analyses. Furthermore, a sensitivity analysis assuming that all patients lost to follow‐up had died at their last recorded contact yielded similar survival estimates, suggesting that the principal findings were robust to potential follow‐up bias.

Third, OS was based on all‐cause mortality rather than CC‐specific mortality because reliable cause‐of‐death information was not consistently available. Consequently, mortality attributable specifically to CC may have been overestimated, particularly among older patients with competing health conditions.

Fourth, treatment categories were based on assigned treatment plans rather than confirmed treatment completion. Information on treatment adherence, interruptions, and completion was not consistently available, which may have introduced some degree of treatment misclassification and influenced treatment‐survival associations. In addition, treatment allocation was determined by clinical presentation, disease stage, and treatment availability rather than random assignment. Therefore, treatment‐related findings should be interpreted as associations and not as evidence of causal treatment effects.

Finally, several potentially important variables, including HIV status, detailed tumor characteristics, HPV genotype, treatment delays, and patient‐reported outcomes, were incompletely recorded or unavailable in the registry. HIV status was therefore reported descriptively but excluded from the multivariable Cox model. In addition, formal inter‐rater reliability among the annual teams of abstractors was not assessed. Although the models adjusted for available prognostic factors, residual confounding and measurement error cannot be excluded.

4.2. Study Implications

This study highlights several priorities for improving CC outcomes in Mali and similar low‐resource settings. The poor survival observed in this cohort, together with the high proportion of women diagnosed with advanced‐stage disease, underscores the importance of expanding early detection through screening programs, primary healthcare services, community outreach, and public awareness initiatives. The strong associations observed between anemia, educational attainment, disease stage, and mortality suggest that both clinical and social determinants influence survival outcomes and should be addressed through comprehensive cancer control strategies.

The findings also highlight persistent challenges related to access to specialized cancer care. During most of the study period, radiotherapy services were available only in Bamako, creating important geographic and financial barriers for women living outside the capital. Efforts to reduce treatment delays, improve referral pathways, strengthen supportive care, and expand access to comprehensive oncology services may contribute to better outcomes. Finally, continued investment in cancer registration systems and long‐term follow‐up mechanisms is essential to improve surveillance, evaluate interventions, and support evidence‐based cancer control policies.