Structured Literature Review to Identify Human Papillomavirus's Natural History Parameters for Dynamic Population Models of Vaccine Impacts

Abstract

Human papillomavirus (HPV) is a common sexually transmitted virus that can cause cervical cancer and other diseases. Dynamic transmission models (DTMs) have been developed to evaluate the health and economic impacts of HPV vaccination. These models typically include many parameters, such as natural history of the disease, transmission, demographic, behavioral, and screening. To ensure the accuracy of DTM projections, it is important to parameterize them with the best available evidence. This study aimed to identify and synthesize data needed to parametrize DTMs on the natural history of HPV infection and related diseases. Parameters describing data of interest were grouped by their anatomical location (genital warts, recurrent respiratory papillomatosis, and cervical, anal, vaginal, vulvar, head and neck, and penile cancers), and natural history (progression, regression, death, cure, recurrence, detection), and were identified through a systematic literature review (SLR) and complementary targeted literature reviews (TLRs). The extracted data were then synthesized by pooling parameter values across publications, and summarized using the range of values across studies reporting each parameter and the median value from the most relevant study. Data were extracted and synthesized from 223 studies identified in the SLR and TLRs. Parameters frequently reported pertained to cervical cancer outcomes, while data for other anatomical locations were less available. The synthesis of the data provides a large volume of parameter values to inform HPV DTMs, such as annual progression rates from cervical intraepithelial neoplasia (CIN) 1 to CIN 2+ (median of highest quality estimate 0.0836), CIN 2 to CIN 3+ (0.0418), carcinoma in situ (CIS) 2 to local cancer+ (0.0396), and regional to distant cancer (0.0474). Our findings suggest that while there is a large body of evidence on cervical cancer, parameter values featured substantial heterogeneity across studies, and further studies are needed to better parametrize the non-cervical components of HPV DTMs.

Keywords: Anal cancer; Cervical cancer; Disease transmission models; Genital warts; HPV; Head and neck cancer; Penile cancer; RRP; Vaginal cancer; Vulvar cancer.

Fig. 1

PRISMA diagram of the selection of publications in the systematic literature review (SLR). The SLR search strategy yielded a total of 4551 records considered for screening, and 239 studies were included after full-text (level 2) screening. Data were extracted from 150 studies according to the prioritization (study published in 2008 or later, reported original data, and included at least 30 human papillomavirus [HPV]-positive individuals). In addition, supplemental manual searches identified 43 studies reporting parameters of interest in non-cervical anatomical locations (e.g., anal, vaginal, vulvar, head and neck, penile, and warts), leading to a total of 193 studies identified for data extraction. HPV human papillomavirus, PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses, SLR systematic literature review

Fig. 2

Description of the selection of publications in the targeted literature review (TLR). The TLR search strategy yielded 709 records. Of those, 297 studies were included for title and abstract screening, and 26 studies were selected for data extraction after full-text screening. Four additional articles from manual searches on Google Scholar were selected for data extraction, leading to a total of 30 articles for data extraction and analysis. TLR targeted literature review

Fig. 3

Summary of data synthesis for parameter annual rates in cervical diseases. The parameters extracted as rates for cervical diseases (e.g., progression rates) are shown using a logarithmically transformed scale (using log₁₀ where Inf. is abbreviated for parameter values with an infinite upper bound) to enhance the visual differentiation of estimates clustered around low values, and for plotting a wide range of values. For each parameter, the value, or the median of values, reported in the highest quality study is shown in blue, and all values extracted for the parameter are in orange. Each study’s population sample size is denoted using the size of the data point. A circle was used to denote data from clinical trials, and a diamond was used to denote data from real-world studies. Data is presented for all studies combined, as well as stratified on the basis of age group. The “Other” age group includes parameters defined for populations whose age range spanned < 26 years through 26+ years as well as studies reporting unspecified age ranges. The parameters from the highest quality study shown in the age stratification belong to the highest quality study from all data extracted (“all studies”) for each parameter. CIN cervical intraepithelial neoplasia, CIS carcinoma in situ, HPV human papillomavirus

Fig. 4

Summary of data synthesis for parameter annual rates in head and neck (H&N), anal, penile, vaginal, and vulvar diseases. The parameters extracted as rates for H&N, anal, penile, vaginal, and vulvar disease are shown using a logarithmically transformed scale (using log₁₀ where Inf. is abbreviated for parameter values with an infinite upper bound) to enhance the visual differentiation of estimates clustered around low values, and for plotting a wide range of values. For each parameter, the value, or the median of values, reported in the highest quality study is shown in blue, and all values extracted for the parameter are in orange. Each study’s population sample size is denoted using the size of the data point. A circle was used to denote data from clinical trials, and a diamond was used to denote data from real-world studies. CIS carcinoma in situ, HPV human papillomavirus, H&N head and neck, IEN intraepithelial neoplasia

Fig. 5

Summary of data synthesis for parameter annual rates in warts and recurrent respiratory papillomatosis (RRP). The parameters extracted as rates for warts and RRP are shown using a logarithmically transformed scale (using log₁₀ where Inf. is abbreviated for parameter values with an infinite upper bound) to enhance the visual differentiation of estimates clustered around low values, and for plotting a wide range of values. For each parameter, the value, or the median of values, reported in the highest quality study is shown in blue, and all values extracted for the parameter are in orange. Each study’s population sample size is denoted using the size of the data point. A circle was used to denote data from clinical trials, and a diamond was used to denote data from real-world studies. HPV human papillomavirus, RRP recurrent respiratory papillomatosis

Fig. 6

Summary of data synthesis for performance of detection tests of cervical disease parameters. The performance of detection tests (i.e., positive predictive value [PPV], negative predictive value [NPV], sensitivity, and specificity) related to cervical complications parameters are reported using percentages (0–100%). For each parameter, the value, or the median of values, reported in the highest quality study is shown in blue, and all values extracted for the parameter are in orange. Each study’s population sample size is denoted using the size of the data point. A circle was used to denote data from clinical trials, and a diamond was used to denote data from real-world studies. Data is presented for all studies combined, as well as stratified on the basis of the available age groups. The “Other” age group includes parameters defined for populations whose age range spanned < 26 years through 26+ years as well as studies reporting unspecified age ranges. The parameters from the highest quality study shown in the age stratification belong to the highest quality study from all data extracted (“all studies”) for each parameter. CIN cervical intraepithelial neoplasia, CIS carcinoma in situ, NPV negative predictive value, PPV positive predictive value