Age‐specific burden of cervical cancer associated with HIV: A global analysis with a focus on sub‐Saharan Africa

Abstract HIV substantially worsens human papillomavirus (HPV) carcinogenicity and contributes to an important population excess of cervical cancer, particularly in sub‐Saharan Africa (SSA). We estimated HIV‐ and age‐stratified cervical cancer burden at a country, regional and global level in 2020. Proportions of cervical cancer (a) diagnosed in women living with HIV (WLHIV), and (b) attributable to HIV, were calculated using age‐specific estimates of HIV prevalence (UNAIDS) and relative risk. These proportions were validated against empirical data and applied to age‐specific cervical cancer incidence (GLOBOCAN 2020). HIV was most important in SSA, where 24.9% of cervical cancers were diagnosed in WLHIV, and 20.4% were attributable to HIV (vs 1.3% and 1.1%, respectively, in the rest of the world). In all world regions, contribution of HIV to cervical cancer was far higher in younger women (as seen also in empirical series). For example, in Southern Africa, where more than half of cervical cancers were diagnosed in WLHIV, the HIV‐attributable fraction decreased from 86% in women ≤34 years to only 12% in women ≥55 years. The absolute burden of HIV‐attributable cervical cancer (approximately 28 000 cases globally) also shifted toward younger women: in Southern Africa, 63% of 5341 HIV‐attributable cervical cancer occurred in women <45 years old, compared to only 17% of 6901 non‐HIV‐attributable cervical cancer. Improved quantification of cervical cancer burden by age and HIV status can inform cervical cancer prevention efforts in SSA, including prediction of the impact of WLHIV‐targeted vs general population approaches to cervical screening, and impact of HIV prevention.


| INTRODUCTION
Cervical cancer is a major public health problem, representing the fourth most common cancer in women worldwide and accounting for more than 600 000 new cases and 340 000 global deaths in 2020 (GLOBOCAN 2020 database presented in Global Cancer Observatory https://gco.iarc.fr/today/home). 1 However, this burden is unequally distributed, with 9 out of 10 deaths from cervical cancer occurring in low-and middle-income countries (LMIC), and 6 of those in sub-Saharan Africa (SSA) alone. 1 This inequality is partly a product of lack of access to cervical cancer screening and cancer treatment, but also of disparate prevalence of risk factors, including high-risk human papillomavirus (HR HPV) and HIV infection.
Persistent infection with HR HPV types is the underlying cause of all cervical cancer. However, natural history of HR HPV infection is substantially worsened by HIV-related immunodeficiency, such that women living with HIV (WLHIV) are at elevated cervical cancer risk. In a recent systematic literature review and meta-analysis, HIV was estimated to increase cervical cancer risk 6-fold and, in a subsequent modeling exercise, to account for approximately 5% of the global cervical cancer burden. 2 HIV-attributable cervical cancer burden is particularly unequally spread, with 85% of cases diagnosed in SSA alone. In Southern Africa, the SSA region most impacted by the HIV epidemic, more than half of all cervical cancer cases in 2018 were estimated to be attributable to HIV. 2 In 2020, the World Health Organization (WHO) launched a global call to eliminate cervical cancer as a public health problem, for which the main prevention components are HPV vaccination, cervical cancer screening and management of detected disease. 3 However, in settings doubly hit by both HPV and HIV epidemics, most notably SSA, progress toward the cervical cancer elimination goal will also be influenced by HIV-focused prevention measures, such as reducing HIV prevalence, early diagnosis of HIV and timely initiation of combination antiretroviral therapy (cART). In these settings, age-specific estimates of cervical cancer by HIV status are key to informing the design of appropriate cervical cancer control programs (including the extent to which cervical screening should be prioritized and adapted for WLHIV), as well as to predict their impact.
There has been no previous description of the interaction of HIV infection and age on cervical cancer at a population level. Such an effort has been complicated by the changing epidemiology of the HIV epidemic, first in terms of changing HIV prevalence, but even more so by huge fluctuations in life expectancy due to severe co-mortality from opportunistic infections (ie, decrease in life expectancy in the early phases of the epidemic, followed by a subsequent increase in the era of wider and earlier access to cART). Indeed, age-specific estimates of HIV-attributable cervical cancer were considered beyond the scope of the above-mentioned meta-analysis and global modeling exercise, 2 in recognition of the need for a more targeted approach.
To this end, our aim had two parts: first, to develop a methodology to describe the relative contribution of HIV to age-specific cervical cancer burden globally, with a particular focus on SSA. This approach was based on using most recent age-specific estimates of HIV prevalence (UNAIDS) and relative risk (RR), accompanied by a widespread collection of empirical data from cervical cancer series of known HIV status, in order to inform and validate the methodology. Then, as a second step, we applied this methodology to worldwide estimates of cervical cancer incidence (GLOBOCAN 2020), 1 to estimate HIV-and age-stratified cervical cancer burden at a country, regional and global level.

| Empirical evidence on HIV prevalence in cervical cancer, by age
The same literature review described above 2 also identified 19 epidemiological studies reporting HIV status in consecutively diagnosed series of cervical cancer. These studies were all conducted in SSA, and authors were invited to share data on HIV prevalence according to the age groups ≤34, 35 to 44, 45 to 54 and ≥55 years. Relevant data were obtained for 17 series from 13 countries: Botswana, 6 Côte d'Ivoire, 7,8 Kenya, 9,10 Malawi, 11 Mozambique, 12 Nigeria, 13 Rwanda, 4 Senegal, 14 South Africa, 10,15,16 United Republic of Tanzania, 17 Uganda, 18 Zambia 19 and Zimbabwe. 20 We were unsuccessful in obtaining age-stratified data from two additional eligible studies from Malawi. 21,22   For each country, and by age group, we calculated the number of new cervical cancer cases (a) diagnosed among WLHIV, and (b) attributable to HIV, respectively, by multiplying these two respective fractions by the number of new cervical cancer cases estimated from GLOBOCAN 2020. Country-specific estimates were aggregated worldwide, according to WHO region and for the WHO African Region (WHO/AFRO), and additionally according to UN sub-region, that is, Eastern, Western, Central and Southern Africa, referred to collectively here as SSA. Overall (ie, all ages combined), HIV prevalence and PAFs were derived from the aggregate of age-specific numerators. Finally, these estimates were applied to age-standardized incidence rates (ASIRs) of cervical cancer, by HIV attribution status, per 100 000 person years for four SSA sub-regions, as available in GLOBOCAN 2020.

| Statistical analysis
All statistical analyses were conducted using Stata software (Version 14.2) and world maps drawn using QGIS3 software.

| Fractions of invasive cervical cancer diagnosed among WLHIV and attributable to HIV
In 2020, 5.6% of global cervical cancer cases were estimated to be diagnosed among WLHIV. However, this fraction varied substantially according to age, being 15.5% for cervical cancer diagnosed at ≤34 years, 9.4% at 35 to 44, 5.6% at 45 to 54 and 1.7% at ≥55 years ( Geographic and age-specific patterns of the fraction of cervical cancer among WLHIV and the PAF are mapped in Figures 1 and 2, respectively. Of note, for women diagnosed with cervical cancer at ≤34 years, HIV prevalence was estimated to be higher than 20% in a number of individual countries outside SSA.

| Empirical data of HIV prevalence in cervical cancer by age
As a validation exercise, Figure 3 compares the above-described IARC age-specific estimates of HIV prevalence in cervical cancer in 2020 F I G U R E 1 HIV prevalence in cervical cancer in 2020, by age group. The designations used and the presentation of the material in this article do not imply the expression of any opinion whatsoever on the part of WHO and the IARC about the legal status of any country, territory, city, or area, or of its authorities, or concerning the delimitation of its frontiers or boundaries F I G U R E 2 HIV-attributable fraction in cervical cancer in 2020, by age group. PAF, population-attributable fraction. The designations used and the presentation of the material in this article do not imply the expression of any opinion whatsoever on the part of WHO and the IARC about the legal status of any country, territory, city, or area, or of its authorities, or concerning the delimitation of its frontiers or boundaries

| Absolute burden of invasive cervical cancer cases diagnosed in WLHIV and attributable to HIV
By applying the fractions described in Table 1 Table S1.
3.4 | Incidence rates of cervical cancer, by HIV attribution status  The validity of using age-specific RRs in our model was confirmed by reproducing the HIV prevalence observed in contemporary cervical cancer cases series, in which HIV prevalence was also observed to decrease strongly by age. Indeed, a model applying the single overall RR (ie, 6) to age-specific HIV prevalence did not reproduce age-specific HIV prevalence in these recent cervical cancer series as precisely as an age-specific RR approach ( Figure S1). Interestingly, the age-specific RR approach was able to reproduce contemporary empirical evidence in settings with broadly different HIV prevalence (eg, Botswana and Côte d'Ivoire), clearly illustrating that RRs are independent of HIV prevalence. Rather, the reason RRs decrease so strongly by age is likely driven by the strong underly- In recognition of the increased risk of cervical cancer in WLHIV, it is widely recommended for cervical screening of WLHIV to start at an earlier age, and for subsequent screening intervals to be shorter for WLHIV than for the general female population. 35 In SSA, and some other settings without population-level cervical cancer screening programs, this has led to several cervical cancer screening initiatives that primarily target WLHIV, most notably those funded by PEPFAR. 36 This approach can be facilitated by WLHIV undergoing regular followup in the healthcare system. Furthermore, investment in new screening infrastructures for WLHIV can be a catalyst for subsequent expansion of these services to the wider HIV-uninfected population. 37 Thus, age-specific burden of cervical cancer in WLHIV vs that in HIVuninfected women at a population level can help inform the appropriate lower age limit of WLHIV-targeted cervical cancer screening, as well as assess the relative impact of a targeted approach for WLHIV vs that of a more general population program. 32,38 The WHO global strategy to eliminate cervical cancer is ultimately underpinned by widespread implementation of HPV vaccination programs. Vaccination prior to sexual activity is expected to prevent cervical cancer in vaccinated cohorts, but this impact will not occur for a number of decades and will first be seen on cervical cancer burden at youngest ages. 39 Vaccine impact should be witnessed irrespective of HIV status, given that HIV is not a direct carcinogen, but rather acts via impaired immunity to worsen the carcinogenic effect of HR HPV.
Hence, increasing HPV vaccination coverage in countries with a high burden of cervical cancer, particularly those with a double burden of HPV and HIV, remains a critical priority. This is especially the case as these same high burden countries are often those where populationwide cervical screening coverage is low. 40 Of note, all ages combined, overall burden estimates of HIVassociated cervical cancer at a country, regional and global level for 2020 are materially unchanged from those of 2018. 2  In conclusion, the burden of cervical cancer associated with HIV is strongly shifted toward women at younger ages and has been changing with the evolution of the HIV epidemic in SSA.
Thus, locally relevant data on the age-specific contribution of HIV to cervical cancer should be used to better design prevention programs, particularly in settings in SSA hit by a double burden of HPV and HIV. To keep evidence up to date, SSA countries should be encouraged to document HIV status and cART use in clinical cervical cancer series. This should be feasible given that cervical cancer is an AIDS-defining condition, and that HIV-testing is becoming increasingly widespread. These data will be highly informative for monitoring progress toward the WHO cervical cancer elimination goal in SSA.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of our study derive from several sources. Publicly available sources are: cancer incidence in five continents volume XI (http://ci5.iar.fr/) and Globocan 2020 (Global Cancer Observatory https://gco.iarc.fr/today/home). For included studies, please refer to cited published references. Further information is available from the corresponding author upon request.