Ecologic Analysis of Correlates of Cervical Cancer Morbidity and Mortality in Sub-Saharan Africa

Nearly 570,000 women were diagnosed in 2018 with cervical cancer worldwide, some 112,000 of whom lived in Sub-Saharan Africa (1). Cervical cancer is the fourth most common cancer in women worldwide and the primary cause of cancer-related deaths among women in Sub-Saharan Africa (2, 3). Infection by the sexually transmitted human papillomavirus (HPV) has long been established as a necessary cause of cervical cancer (4, 5).

In 2020, The World Health Organization (WHO) launched a campaign to eliminate cervical cancer based on a triple strategy of attaining 90% coverage of HPV vaccination among girls by age 15, 70% coverage of screening with a high precision test at ages 35 and 45 years, and 90% treatment coverage for women with cervical disease (6). Although these targets are ambitious even for high-income countries, there is evidence from modelling studies that elimination can be achieved if these strategies were to be jointly deployed in low- and middle- income countries (7, 8). At current levels of screening, vaccination, and treatment strategies, it was estimated that the age-standardized cervical cancer incidence in South Africa will be reduced from 49.4/100,000 women in 2020 to 12/100,000 women in 2120, whereas adopting new screening technologies could reduce incidence to 4.4/100,000 women in 2120, translating to a 44% probability of cervical cancer elimination based on the WHO threshold of 4/100,000 women (9). To address the high cervical cancer rate in Sub-Saharan Africa, countries need to implement sustainable, effective, and accessible cervical cancer screening and HPV vaccination programs, as well as proper therapeutic follow-up.

The incidence of cervical cancer is relatively high for all sub-Saharan regions. Using data of cancer estimates from the Global Observatory 2018 database, Southern and Eastern Africa had the greatest burden at an age-standardized incidence rate of 43.1/100,000 women and 40.1/100,000 women, respectively, compared with Western Africa (29.6/100,000 women), Middle Africa (26.8/100,000 women), and Northern Africa (7.2/100,000 women; ref. 3). Cervical cancer was also the leading cause of cancer-related deaths in women in Eastern, Western, Middle, and Southern Africa with corresponding age-standardized mortality rates of 30.0, 21.1, 20.0, and 23.0 per 100,000 women (3). As of 2020, of the 20 countries with the heaviest burden of cervical cancer, both in incidence and mortality, 19 are sub-Saharan countries (including Comoros; ref. 1). Given the major challenges of eliminating cervical cancer in the sub-Saharan region, where the burden of this disease is the greatest worldwide and resources to deploy vaccination, screening, and treatment are severely limited, we set out to identify contextual factors associated with the cervical cancer burden in this region. To our knowledge, this is the first ecological study of correlates of cervical cancer morbidity and mortality in Sub-Saharan Africa.

We obtained relevant data on country-specific age-standardized incidence and mortality rates for cervical cancer from the Global Cancer Observatory (1) for 37 Sub-Saharan African countries.

We identified several a priori potential correlates of cervical cancer that are well covered in the scientific literature, as referenced below, and extracted corresponding country-specific data from various publicly available sources, as listed in Table 1. Incidence of cervical cancer is strongly correlated with that of cancer of the penis (10, 11). We obtained data on age-standardized penile cancer rates from the Global Cancer Observatory (1). Literacy rate, number of doctors, and screening coverage are negatively associated with cervical cancer incidence and mortality (12, 13). Higher fertility rate is also associated with risk of cervical cancer (14). Some of the shared behavioral characteristics of women with cervical cancer are early sexual debut, tobacco smoking, and childbearing at a young age (15). We obtained data on literacy rate, median age at first sex, median age at first child, fertility rate, number of medical doctors per 1,000 population, screening coverage (reported via Pap smear testing), and tobacco smoking prevalence from the HPV Information Centre (16). Contraceptive use is associated with a four-fold increase in the risk of cervical cancer in HPV-positive women (17). As an immunosuppressive condition, infection with the human immunodeficiency virus (HIV) is also a predictor of HPV-related malignancies, including cervical cancer (17). A 0.2 unit increase in Human Development Index (HDI) is associated with a 20% and 33% decrease in cervical cancer incidence and mortality, respectively (12). We used the United Nations Development Programme 2018 Human Development report to obtain data on contraceptive use, HIV prevalence, life expectancy, HDI, and income index (18). We collected data on rural and Muslim populations from the World Data Bank (19) and Nation Master, respectively (20). We included these variables to explore if urbanization and religion had an impact on cervical cancer rates, under the assumption that they could be correlates of behavioral risk factors for the disease.

We computed descriptive statistics and generated a correlation matrix with the Pearson correlation coefficients between pairs of correlates. The correlation coefficient for each pair of variables varies between −1 (perfect negative correlation) and 1 (perfect positive correlation).

We used univariate linear regression to examine the independent association between each variable and three dependent variables: cervical cancer incidence rate, cervical cancer mortality rate, and case-fatality. The latter, used as a measure of clinical prognosis, was calculated as the ratio between mortality and incidence rates. We also conducted separate stepwise multiple linear regression analyses using a combination of forward and backward selection techniques to determine the best subset of correlates of cervical cancer incidence, mortality, and case-fatality based on the Akaike's Information criterion. The model with the lowest value was considered the one with the best balance of goodness of fit and parsimony. Based on the best model, predicted values for incidence, mortality, and case-fatality were calculated using the predict function in R. The predictive value of each model was evaluated using a goodness-of-fit test, P values and R-squared measures. Scatter plots for the observed versus predicted values were constructed to assess linearity for the assumed model for each of the dependent variables. The linear regression analyses were done (i) restricting to countries with data; and (ii) including all countries by imputing the variable with missing data using a weighted mean based on data from neighboring countries. Statistical analyses were performed using R version 3.6.2.

The data generated and/or analyzed are available from the corresponding author upon reasonable request. The raw data used in the analyses are publicly available as per the sources described in the paper.

Figure 1 shows country-specific cervical cancer incidence and mortality in 2018. Incidence ranged from a low of 13.8 to a high of 100.1 cases per 100,000 women for Niger and Eswatini, respectively. Similarly, Niger had the lowest mortality rate at 12.1 deaths per 100,000 women. Malawi had the highest at 71 deaths per 100,000 women while Eswatini was the second highest at 67.6 deaths per 100,000 women. Case-fatality ranged from 0.39 for South Africa to 0.88 for Niger.

Table 2 presents the statistical distribution for candidate correlates of cervical cancer. On average, the life expectancy of a woman was around 63 years. The average literacy rate was 71.5%, ranging from 17% to 99.5%. The median age of bearing a child was low at 19.8 years, consistent with the high fertility rate at 5 children per woman and the low average prevalence of contraceptive use (31%). There was less than 1 doctor per 1,000 population. Screening coverage was drastically low at an average of 7.55% and reached a maximum of 23.6%. As shown, data for some countries were missing for HPV prevalence (51%), cervical cancer screening coverage (38%), prevalence of tobacco smoking (35%), median age at first child (11%), and median age at first sex debut (5%). We present in Supplementary Table S1 the prevalence of these correlates by country and indicate whether data were measured or imputed.

Supplementary Fig. S1 displays pairwise correlations for all the possible combinations of correlates. The relationship between correlates was as expected. There was a strong positive correlation between income index and HDI (r = 0.79) and between literacy rate and HDI (r = 0.75). HPV prevalence and fertility had the strongest negative correlation (r = −0.71).

Results from the univariate linear regression analyses are presented in Table 3. Literacy rate, contraceptive use, HIV positivity, and penile cancer incidence were correlated with the incidence of cervical cancer. For every unit increase in literacy rate and contraceptive use, there was a 0.40 (95% CI, 0.08–0.73) and 0.73 (95% CI, 0.34–1.06) estimated increase in incidence, respectively. Similarly, contraceptive use, HIV positivity, rural population, and penile cancer incidence were positively correlated with mortality, whereas screening coverage was negatively correlated. Neither fertility rate nor the number of medical doctors per 1,000 population were correlated with either incidence or mortality. All correlates, except for median age at first sex and tobacco smoking prevalence, were significantly correlated with case-fatality.

Table 4 shows the correlates that were retained as independent explanatory variables from stepwise multiple linear regression analyses. In all three models, tobacco smoking, fertility rate, HIV prevalence in adults, and screening coverage were not retained. Only contraceptive use (P = 0.009) and penile cancer incidence rates (P = 0.004) were positively correlated with cervical cancer incidence. The number of medical doctors/1,000 population–negatively (P = 0.0433)– and penile cancer incidence–positively (P = 9.77 × 10^–5) –were significantly correlated with mortality. Although the goodness of fit of both models were statistically significant, they only explained 49% and 37% of the variability, respectively. In contrast, the case-fatality model was highly predictive, explaining nearly 95% of the variability in cervical cancer case-fatality across countries in Sub-Saharan Africa (adjusted R² = 0.948, P = 6.822 × 10^–16). Figure 2 shows the scattergrams resulting from the above parsimonious models with their respective goodness of fit statistics.

Supplementary Table S2 shows results for the multiple linear regression model restricting to countries with measured data. The resulting dataset included eight countries with data on all 16 correlates. For incidence and mortality, Muslim population, median age at first sex, median age at first child, tobacco prevalence, HIV and HPV prevalence were retained. Muslim population, median age at first sex, median age at first child, and tobacco prevalence were significantly correlated with cervical cancer incidence and mortality. For the case-fatality model, Muslim population, median age at first sex, tobacco prevalence, and HIV prevalence were not retained. HPV prevalence, a key predictor of cervical cancer, which was initially significant in the analysis with the imputed data (P = 0.025) lost significance in the current analysis with measured HPV data (P = 0.112). Nonetheless, case-fatality continued to explain most of the variability in the data.

This ecological study sought to examine the underlying correlates of cervical cancer morbidity and mortality in Sub-Saharan Africa. 16 candidate cervical cancer correlates were explored to determine their correlation with cervical cancer incidence and mortality rates as well as case-fatality ratio. Our findings show that they correlate significantly with multiple contextual factors.

Maintaining an effective national cervical cancer screening program has been the mainstay of cervical cancer control programs (7). Although cervical cancer screening is free (or provided at a low cost) in most of Sub-Saharan Africa and several countries have launched national or demonstration screening programs (21), the question remains as to whether these screening services are easily accessible to their target populations and are of sufficient quality.

There are high levels of willingness to vaccinate but also low levels of awareness and knowledge of cervical cancer in Sub-Saharan Africa (22, 23). A scoping review of 9 studies published between 2005 and 2019 highlighted the utilization of community engagement (i.e., religious organizations, traditional leaders, and educational institutions) in sub-Saharan communities to increase awareness of cervical cancer and address disparities related to cervical cancer incidence and mortality (24). Studies have shown that women in Sub-Saharan Africa are willing to get vaccinated if vaccination services are accessible. Around 90% of 10 to 19 year-old girls in Mozambique were willing to get vaccinated (25). Ninety-four percent of women in Ghana, and 88% of women in Botswana were willing to allow their daughters to receive the HPV vaccine (26, 27). However, adequate screening coverage continues to be the greatest challenge in cervical cancer prevention in Sub-Saharan Africa at a disappointingly low coverage of 7.6%. Our study shows that the weighted case-fatality for Sub-Saharan African is 0.75, which is very high compared with that in the United States at 0.34. This likely reflects ineffective cervical cancer screening and prevention programs, as well as low coverage of specialized treatment.

Sub-Saharan countries have less than 1 doctor per 1,000 population, relative to 3 doctors per 1,000 population in developed countries (18). This low level of healthcare manpower is not conducive with the WHO's bold goal of eliminating cervical cancer in the region. In Sub-Saharan Africa, the average HDI, a measure of a country's human development, is 0.3 compared with 0.9 in developed countries (18).

In Sub-Saharan Africa, countries with higher literacy rates tend to have higher cervical cancer incidence rates in comparison with less literate countries (28). Our univariate analyses showed that literacy rates correlated with cervical cancer incidence rates and case-fatality. Countries with higher levels of education, such as Zimbabwe (93.5%) and Zambia (90.6%), also had respectively high levels of cervical cancer incidence rates of 84.4 and 89.8 cases/100,000 women (19). Education enables individuals to ascend the social class ladder and that in turn brings economic opportunities that largely influence demographic, religious, cultural, and sexual behavior patterns (28, 29). Our study found a correlation between HIV prevalence (a proxy for the risky sexual behavior) and incidence and mortality. HIV-seropositive women have a higher risk of acquiring HPV infection, and most importantly, they experience two to 12 times increased risk of developing precancerous lesions as the HPV infection persists (30). We found that penile cancer rates were correlated with an increase in incidence and case-fatality. A similar association between penile cancer rates and cervical cancer rate has been found in another study with a Pearson correlation of 0.88 (10).

Many Sub-Saharan African countries lack sustainable cervical cancer screening programs and vaccination policies. Sustainability is a major issue; many programs are run by foreign donors and are discontinued when the donor exits the country (29). In a survey of 29 best-resourced centers from 12 sub-Saharan countries (Ethiopia, Uganda, Kenya, Tanzania, Malawi, South Africa, Botswana, Zambia, Zimbabwe, Namibia, Nigeria, and Ghana), around 300 new cases of cervical cancer were reported annually in each center; about half of the surveyed countries had an HPV vaccination program in place, 88% of participating centers had a surgeon with an expertise in performing oncological surgeries, and a third had to delay or substitute treatment because of the lack of consistent medical supply (31). These findings reflect the advancements that have been made in cervical cancer management in Sub-Saharan Africa but also highlight that access to adequate treatment is essential in these regions. A scoping review of 19 studies in 7 countries in Sub-Saharan Africa of interventions used to increase the uptake of cervical cancer screening among women reported that health education (i.e., lecture-based, peer health educators, multimedia lessons) was the most prevalent type of intervention (11 studies, 57.9%), followed by innovative service delivery (i.e., changes to the type of screening provided, screening location, removing environmental barriers, integrating screening with HPV vaccination) interventions (6 studies, 31.6%), and economic incentive (i.e., lottery-style game, a reward program offered by a health insurance plan), interventions (2 studies, 10.5%; ref. 32). The authors found that screening coverage after these interventions ranged from 1.7% to 99.2%, compared with the baseline 0% to 53.6% coverage (32).

A few limitations inherent to an ecologic study need to be acknowledged. Data were missing for important covariates such as HPV prevalence and screening coverage. However, based on results from the sensitivity analysis (dataset restricted to measured covariates only), the model retained the same variables as those based on the model using imputed variables. Also, the case fatality model still explained most of the variation. Another limitation relates to the validity of country-level data which is unknown as some data points may not have been recorded due to limited resources. Despite these limitations, the study had key strengths including the use of multiple sources of data and following a rigorous analytical approach.

For elimination of cervical cancer, Sub-Saharan Africa needs to implement, in addition to cervical cancer screening and treatment as well as HPV vaccination, targeted interventions that focus on improving contextual factors, such as those that we found to be significantly correlated with cervical cancer burden or relative survival. Findings from the current analyses would be of interest to public health practitioners and epidemiologists who seek to understand the aggregate correlates of cervical cancer in Sub-Saharan Africa. They might also guide subsequent stakeholders in developing problem-targeted population-level interventions to reduce the cervical cancer burden in Sub-Saharan Africa.

C.R. Gapare reports other support from Division of Cancer Epidemiology during the conduct of the study. M. El-Zein reports a patent for DNA methylation markers for early detection of cervical cancer pending. E.L. Franco reports grants from Merck; and grants from Canadian Institutes of Health Research outside the submitted work; in addition, E.L. Franco has a patent for Methylation markers pending to self. No disclosures were reported by the other authors.

C.R. Gapare: Conceptualization, data curation, formal analysis, visualization, methodology, writing–original draft, project administration, writing–review and editing. M. El-Zein: Supervision, methodology, writing–original draft, writing–review and editing. H. Patel: Data curation, writing–review and editing. P. Tope: Data curation, writing–review and editing. E.L. Franco: Conceptualization, supervision, funding acquisition, methodology, writing–review and editing.

The authors are grateful to Dr. Talía Malagón for valuable insights into data analyses.

The work was supported by the Canadian Institutes of Health Research (CIHR Foundation Grant; grant no. FDN-143347 to E.L. Franco). C.R. Gapare received an MScPH stipend from the Division of Cancer Epidemiology, McGill University (Montreal, Quebec, Canada).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.