Background
The successful implementation of cervical cancer (CC) screening and the introduction of human papillomavirus (HPV) vaccine as a preventative strategy to reduce cervical cancer burden has had a great impact especially in the high-income countries (HIC) [
1]. However, CC is still estimated to be the fourth most common cancer worldwide among women, with an increasing number of new cases: from 493,000 new cases in 2002 to 530,000 in 2008, and the number of deaths increasing from 274,000 in 2002 to 275,000 deaths in 2008 [
2‐
4]. About 85% of the world’s CC cases occur in the low-income countries (LICs) [
5]. CC is the most frequent cancer type among women in Africa, and highly prevalent among women ages 15 to 44 [
6] and in the most disadvantaged population [
7]. Over 90% of all cervical cancer cases are caused by persistent infection with high-risk types Human Papillomaviruses (HPV) [
1,
6], which can lead to pre-cancerous lesions that may progress to invasive cervical carcinoma if left untreated [
6].
The lack of data and poor quality data on CC are likely to result in an underestimated number of CC cases, since women often die of other competing causes, e.g. other AIDS defining illnesses, prior to cervical cancer diagnosis, and since the poor health infrastructure in many LICs results in under-reporting of CC [
6]. However, quantifying the CC rate is a critical first step towards prevention as it provides vital information to policy and decision-makers when ascertaining all resources needed to tackle the disease [
7,
8]. It has been established that the most accurate measure of CC incidence can be attained from population-based registries, which provide estimates of disease occurrence in a well-defined population [
9]. Research has demonstrated that the quality and completeness of data collection, as well as accurate and reliable measures of population denominators are very crucial components for cancer registries [
7]. Unfortunately, for LICs like Swaziland, the lack of proper resources and infrastructure for case findings and reporting prevent the establishment and maintenance of accurate cancer registries. Furthermore, such challenges in LICs have contributed to the fact that many cases of CC go undiagnosed and unreported [
1,
2]. About 80% of cervical cancer patients in developing countries like Swaziland present with late-stage tumors when they are diagnosed, resulting in poor prognosis [
2]. As means of cervical cancer screening, Pap smear was introduced in national cervical cancer prevention programme in 1983 [
10]. However, in 2009, the government of Swaziland incorporated the “See and Treat” approach to quicken the early detection of cervical lesions and facilitate the extension of cervical cancer prevention services across four political regions [
11].Currently, HPV vaccine is not part of Extended Programme on Immunization (EPI) in the country.
The understanding of the epidemiology and natural history of cervical cancer at population level and to prevent the escalating burden of the disease in LICs is essential. Scarcity of country data on the burden of cervical cancer remains a huge challenge in some LICs such as Swaziland, yet such data are critical to informing decisions about resource allocation to combat the disease. The lack of cancer registries to provide these data in LICs is the major limitation to establish cancer incidence.
The aim of this study is to develop a prediction model to estimate cervical cancer incidence without a population-based cancer registry, but using currently country detected hr-HPV prevalence and other continental prevalence. Measuring the CC burden is of paramount importance to better inform policy guidance on cervical cancer screening, as well as developing strategies on HPV vaccine implementation.
Discussion
Cervical cancer remains a significant public health concern worldwide especially in the low-income countries [
43,
44]. Continental reports or studies on the incidence of cervical cancer have demonstrated the severity of the HPV related condition [
45]. It has been established that population-based cancer registries are a source for quantifying the disease burden in a defined population. However, the most regrettable situation is that cancer registries are either non-existent or not fully operational in most LCIs such as Swaziland, thus preventing the estimation of the actual disease burden [
43,
44]. Therefore, the use of available HPV prevalence and other HPV natural history parameters data to predict cervical cancer incidence become of paramount importance to cover such a gap. Hence, our study used the local and other African countries’ HPV prevalences to predict cervical cancer incidence for Swaziland. Our study demonstrated, as anticipated, a significant linear correlation between population prevalence of hr-HPV infection and cervical cancer incidence. Our study established that HPV among women with normal cytology is a strong predictor of cervical cancer incidence. Based on the three models triangulation approach employed in this study, the predicted average annual age-standardized CC incidence was 58.6 per 100 00 in Swaziland. However, after factoring current HIV prevalence into the model, a higher CC incidence rate estimate of 65.0 per 100000was estimated.
Strengths of the study
This is the first study in Swaziland to estimate the incidence of cervical cancer utilizing local hr-HPV prevalence data and other African countries’ data. In addition, we used 3 accepted methods to triangulate a “best guess” estimate. Furthermore, we sourced multiple estimates for the natural history model to try getting the best-weighted estimates for the progression/regression parameter values and also performed a sensitivity type analysis. The further novelty of our study is that we factored HIV in the model to estimate the impact of HIV on the incidence rate of cervical cancer in Swaziland.
Weakness of the study
The key limitations of our study was that our findings are likely to underestimate the incidence rate since our hr-HPV prevalence was obtained from women of the ages between 15 and 49, yet studies have shown that prevalence at later ages tend to show a better prediction of CC incidence. Another limitation of our study is the effect of ecological fallacy relating to model 2. Furthermore, the age specific CC incidence rates for CC may not be same as in Swaziland (very much biased towards South Africa). However, we have similar burdens for HIV/hr-HPV: most of the countries across the southern African region have experienced high HIV and HPV infection. Another limitation is that we did not factor in HIV. However, future work indicates that we will attempt to refine these estimates including HIV parameter/stratification in all modelling approaches. Finally, the mathematical model: the parameter values may be more biased to more developed settings and hence underestimate CC transition probability.
This current study found a strong correlation between the current population hr-HPV prevalence among women with normal cytology and age standardized cervical cancer incidence. These findings are analogous to those observed from the past epidemiological studies [
7,
46]. However, Sharma et al. demonstrated the age factor in the HPV correlation, where HPV prevalence at later ages was found to be an excellent predictor of cervical cancer incidence compared to that of women below the age of 35 years, with prevalence in women age 55–64 presenting the strongest correlation [
7]. Such high risk could be due to a longer persistence of hr-HPV among old age women. Scientific evidence has been presented that the persistence of hr-HPV acutely increases the risk of developing cervical cancer [
7,
47‐
49].
Our study presented, as expected, a predicted high age-standardized cervical cancer incidence (69.4 per 100,000) among the population in Swaziland. Our results were slightly higher than the ASR estimates provided by the GOBOCAN 2012 (53.1 per 100,000) [
50]. These discrepancies might be due to the fact that our study used actual data as compared to the use of standard population or the rates of from neighboring countries or registries in the same area. In addition, the GLOBOCAN data is not stratified by HIV. Comparing our findings with the GBD 2015 (58.1 per 100,000, 95%CI: 17.3–159.1) [
51] our study triangulation estimate without HIV (58.6 per 100,000) were almost identical to GBD estimates. The further novelty of our study is after factoring current HIV prevalence in the model to estimate the impact of HIV on the incidence rate of cervical cancer in Swaziland, a huge increase of ASR CC incidence rate of 101.1 per 100,000 (95%CI: 90.3–112.2) was observed in the ecological model and could suggest that approaches that do not account for high co-infection of hr-HPV and HIV could potentially underestimate cervical cancer incidence in HIV hyper endemic settings, particularly in Southern Africa. The high ASR in the country may be due to the fact that the country is facing a high epidemic of HIV infection as well as an HIV link with high hr-HPV infection both of which are more likely to be persistent. Studies have established that due to the lack of access to relevant prevention approaches and the association with the HIV epidemic, cervical cancer incidence is expected to rise in the next two decades [
52]. Women infected with HIV have an elevated risk of developing certain malignancies and those malignancies are found to be HPV-related, which reflects the high rate of co-infection with HPV in women with HIV [
53].
When comparing our estimated number of incident cases for Swazi female population age 15+, our current study estimated 221 incident cases. Our estimates were in line with annual number of new cervical cancer (223) reported by the GLOBCAN 2012 [
50,
54] and the average prevalent annual number of 220 reported by Swaziland National Cancer Registry in 2015 [
55].
This current study reinforces the affirmation that a well conducted population-based HPV survey may possibly offer crucial information to estimate the risk of cervical cancer, more especially in the absence of or an inaccurate national registry data. Up-to-date and authentic cancer data are crucial to identify most the important considerations for cancer control strategies at the country level, therefore establishing a quality reporting system and legalizing cancer reporting at national level (in private and public health settings) and creating data linkage procedures with the newly established cancer registry will increase the coverage and quality registry in the country. Finally, the biggest implication of such high incidence is the large cost that will occur for public health care resources utilized for the management and treatment of cervical cancer in Swaziland. The higher the incidence of cervical cancer, the higher the economic burden of cervical cancer in the country.
Conclusions
In conclusion, the observation of this study raises a concern over the burden of cervical cancer where reliable cervical cancer statistics are limited despite the current study showing the high prevalence of hr-HPV and HPV/HIV-coinfection among the Swazi reproductive age women. Our model provided an overall estimate of cervical cancer incidence that can be functional to inform health policy decisions and decision-makers on the allocation of limited resources to prevent and treat cervical cancer effectively. Finally, our study significantly showing the need for future research to modify the natural history model of cervical cancer to factor in HIV co-infection in hyper-endemic settings.
Acknowledgements
We thank the Kingdom of Swaziland Ministry of Health for technical, logistic, financial supporting and allowing us to implement the study, and the support from the International Agency for Research on Cancer (IARC), WHO-Swaziland local office, MTN-Swaziland, Ministry of Health Epidemiology Unit and Sexually Reproductive Health Unit (SRH). We are so thankful for the facilities and support of the HPV/Cervical cancer research team.