Primary HPV screening compared with other cervical cancer screening strategies in women with HIV: a cost-effectiveness study

Abstract

Objective:

To compare the model-predicted benefits, harms, and cost-effectiveness of cytology, cotesting, and primary HPV screening in U.S. women with HIV (WWH).

Design:

We adapted a previously published Markov decision model to simulate a cohort of U.S. WWH.

Setting:

United States.

Subjects, participants:

A hypothetical inception cohort of WWH.

Intervention:

We simulated five screening strategies all assumed the same strategy of cytology with HPV triage for ASCUS for women aged 21 to 29 years. The different strategies noted are for women aged 30 and older as the following: continue cytology with HPV triage, cotesting with repeat cotesting triage, cotesting with HPV16/18 genotyping triage, primary hrHPV testing with cytology triage, and primary hrHPV testing with HPV16/18 genotyping triage.

Main outcome measure(s):

The outcomes include colposcopies, false-positive results, treatments, cancers, cancer deaths, life-years and costs, and lifetime quality-adjusted life-years.

Results:

Compared with no screening, screening was cost-saving, and > 96% of cervical cancers and deaths could be prevented. Cytology with HPV triage dominated primary HPV screening and cotesting. At willingness-to-pay thresholds under $250,000, probabilistic sensitivity analyses indicated that primary HPV testing was more cost-effective than cotesting in over 98% of the iterations.

Conclusions:

Our study suggests the current cytology-based screening recommendation is cost-effective, but that primary HPV screening could be a cost-effective alternative to cotesting. To improve the cost-effectiveness of HPV-based screening, increased acceptance of the HPV test among targeted women is needed, as are alternative follow-up recommendations to limit the harms of high false-positive testing.

Introduction

In 2019, an estimated 263,900 women were living with human immunodeficiency virus (HIV) in the U.S.[1] Due to immunosuppression, women with HIV (WWH) are less likely to clear a human papillomavirus (HPV) infection and more likely to progress from persistent HPV infection to cervical cancer than women without HIV.[2] Meta-analyses have shown that cervical cancer incidence in WWH is significantly higher than in women without HIV.[3,4]

Evidence of declines in cervical cancer incidence and mortality have been observed in women without HIV because of effective screening and timely treatment.[5] Women without HIV in the U.S. who are aged 21 to 65 years are currently recommended to undergo screening at 3-year intervals with cytology alone or at 5-year intervals at ages 25 or 30 and older with an HPV-based testing strategy, either cytology plus high-risk HPV (hrHPV) testing (known as cotesting) or stand-alone primary HPV-testing. Since 2012, multiple U.S. guideline groups have recommended these primary screening methods in combination with various triage strategies for cervical cancer screening.[6–8]

In the U.S., cervical cancer screening guidelines are tailored to WWH to accommodate their higher risk.[9,10] In particular, the U.S. Centers for Disease Control and Prevention recommends that WWH should be screened for cervical cancer throughout their life. Current recommendations are for annual cytology-based screening starting at age 21 years with a switch to a three-year interval after three consecutive normal results. At age 30 years, cotesting can be conducted every 3 years.[9] The current guidelines do not recommend primary HPV testing. Given a growing body of evidence demonstrating a higher sensitivity and negative predictive value for detection of precancer lesions for HPV testing compared with cytology[11], ongoing discussions are focused on whether primary HPV screening for WWH should be recommended as an alternative to cytology-based screening or cotesting.[12] However, the relatively high prevalence of HPV among U.S. WWH compared with women without HIV[13] raises concerns about false-positive testing, excessive invasive procedures, and extended surveillance in this population. Given these concerns, it remains unclear whether primary HPV screening for WWH constitutes high-value care. To address this gap, the current study aims to (1) estimate the benefits of cervical cancer screening and (2) compare the cost-effectiveness of cytology-based screening, primary HPV screening, and cotesting screening strategies for WWH in the U.S.

Methods

Model Overview

We adapted a previously constructed HPV type-specific Markov decision model to examine the effect of various screening strategies for WWH in the U.S.[14] The natural history model simulates HPV infection to cervical carcinogenesis in a cohort of women from age 10 to 85 years. The simulated cohort of girls aged ten years are assumed to be unexposed to HPV and at risk of becoming infected, which could clear or persist, as a function of age. Persistent HPV infections can progress to cervical intraepithelial neoplasia (CIN) grade 1–3, and CIN 3 can progress to cervical cancer. HPV infection and subsequent natural history is further stratified into HPV-16/18 or other hrHPV types.

Natural history parameters calibration for WWH

We re-parameterized the model to simulate the natural history of cervical cancer in a cohort of WWH that would experience increased risks of HPV and CIN2+ incidence compared with women without HIV (hereafter referred to as “average-risk women”). Considering the risks of HPV and CIN2+ in average-risk women as the baseline, the re-calibrated natural history model reflects the higher HPV prevalence and CIN2+ incidence in WWH than that in average-risk women.[3,15] The age- and type-specific baseline HPV prevalence among average-risk women were obtained from a population-based study of HPV in New Mexico.[16] The age-specific baseline CIN2+ incidence was simulated from our previously calibrated and validated model for women at average risk of cervical cancer.[14] We used Bayesian methods to calibrate the model and obtain a posterior distribution of the natural history parameters, as well as the best-fitting parameter set.[17]

Competing risks

The age-specific background incidence of hysterectomy for benign reasons and all-cause mortality were modeled as competing risks. To account for the higher background mortality among WWH, we applied age-specific relative risks (RRs) to the background mortality in average-risk women.[18] We assumed that WWH experience the same incidence of hysterectomy for benign reasons as average-risk women based on current evidence.[14,19]

Cancer mortality

WWH have a 1.85 times the hazard of cervical cancer mortality (hazard ratio [HR]=1.85, 95% CI 1.51–2.27) compared with average-risk women.[20] Our model incorporated this HR to account for the increased cervical cancer mortality in WWH. The baseline cancer mortality was derived from stage-specific cervical cancer survival from the Surveillance, Epidemiology, and End Results (SEER) data.[21]

Test accuracy

Cervical cancer screening test accuracies for cytology and HPV tests were obtained from studies assessing cervical cancer screening diagnostic accuracy among WWH and in average-risk women and are presented in Table 1.[22,23] We used separate test accuracy estimates in WWH, because test performance, especially specificity, can vary substantially in the setting of surveillance[24] and posttreatment[25] from average-risk women (see Table 1 footnote).

Table 1.

Model Inputs for Test Accuracy Estimates and Costs.

Variable	Valuea	Source
Primary screening accuracy
Cytologic testing	Sensitivity: 0.767 (0.755–0.779) Specificity: 0.816 (0.776–0.851)	Kelly et al [22] Koliopoulos et al [23]
hrHPV testing	Sensitivity: 0.929 (0.926–0.932) Specificity: 0.554 (0.522–0.597)	Kelly et al [22] Koliopoulos et al [23]
Cytologic plus hrHPV test (cotesting) b	Sensitivity: 0.983 (0.982–0.985) Specificity: 0.452 (0.40–0.51)	Calculatedb
Surveillance accuracy c
Cytologic testing	Sensitivity: 0.828 Specificity: 0.347	Calculatedc
hrHPV testing	Sensitivity: 0.950 Specificity: 0.129	Calculatedc
Posttreatment follow-up accuracy d
Cytologic testing	Sensitivity: 0.801 Specificity: 0.625	Calculatedd
hrHPV testing	Sensitivity: 0.941 Specificity: 0.318	Calculatedd
Service, Test, and Treatmente Cost (Range) $a
Office visit	82 (56–101)	CMS [26]
Cytologic testing	29 (18–32)	CMS [26]
hrHPV testing	51 (33–54)	CMS [26]
Colposcopy and biopsy	236 (176–294)	CMS [26]
Cryosurgery	160 (135–245)	CMS [26]
Loop excision	575 (438–1,687)	CMS [26]
Cold knife cone biopsy	1,723 (1,603–1,855)	CMS [26]
First year cancer care, ages <65 y	72,818	Mariotto et al [27]
First year cancer care, ages ≥65 y	60,681	Mariotto et al [27]
Ongoing cancer care	1,914	Mariotto et al [27]
Last year of life cancer care, ages <65 y	158,813	Mariotto et al [27]
Last year of life cancer care, ages ≥65 y	105,875	Mariotto et al [27]

Abbreviations: hrHPV, high-risk human papillomavirus; CMS, Centers for Medicare & Medicaid Services.

^aRange used in probabilistic sensitivity analysis.

^bCotesting sensitivity and specificity estimated assuming parallel testing of cytology plus hrHPV testing. sensitivity_cotest = sensitivity_cytology + sensitivity_{hrHPV test} - sensitivity_cytology × sensitivity_{hrHPV test}; specificity_cotest = specificity_cytology × specificity_{hrHPV test}

^cSurveillance sensitivity/specificity estimated using logit transformation of summary cytology accuracy estimates [23]: Ln[(a)/(1-a)] + {Ln[(b)/(1-b)] - Ln[(c)/(1-c)]}=d; 1/(1+e^d)=estimated accuracy in the surveillance setting, where a=test accuracy in the HIV setting [22], b=cytology summary accuracy [23], c=cytology summary surveillance estimate [24]

^dPosttreatment follow-up sensitivity/specificity estimated using logit transformation of summary cytology accuracy estimates [23]: Ln[(a)/(1-a)] + {Ln[(b)/(1-b)] - Ln[(c)/(1-c)]}=d; 1/(1+e^d)=estimated accuracy in the posttreatment setting, where a=test accuracy in the HIV setting [22], b=cytology summary accuracy [23], c=cytology summary post-treatment follow-up estimate [25]

^eTreatment choice depended on cervical intraepithelial neoplasia (CIN) grade and patient age. For CIN2, we assumed immediate treatment of 15% of women aged 21–39 years (cryosurgery 5%; loop excision 95%) and 100% of women aged 40+ years (loop excision 50%; cone biopsy 50% based on a prior study [37]). The remaining women aged 21–39 years had surveillance for up to 2 years. Surveillance was repeat colposcopy and cytology every 6 months for up to 2 years. Persistent CIN2 was treated (cryosurgery 5%; loop excision 95%). All women with CIN3 had immediate treatment as follows: aged <40 years: cryosurgery 5% and loop excision 95%; aged 40+ years: loop excision 50% and cone biopsy 50%. We assumed loop excision after 2 years of CIN1 persistence in this high-risk population as endorsed by 2012 ASCCP guidelines. [38]

Utilities and costs

The utilities associated with a cervical cancer diagnosis, receiving a screening test, and undergoing surveillance and treatment were obtained from a previously described cross-sectional study in average-risk women.[14] Direct medical costs associated with screening, diagnosis, precancerous lesion treatment, cancer, and cancer death were assumed to be independent of the HIV infection status of a woman and have been previously described.[14] Direct medical costs were based on Medicare reimbursement rates and SEER-Medicare claim data, measured in 2016 US dollars and inflated using a 3% annual rate to 2020 dollars.[26,27]

Screening strategies

An inception cohort of WWH was simulated from age 10 to 85 years under a no screening scenario and five screening strategies described below:

• cytology for women aged 21 and older with HPV triage for women with atypical squamous cells of undetermined significance (ASC-US) cytology results (hereafter referred to as “cytology with HPV triage of ASC-US”);

• cytology for women aged 21 to 29 years, followed by cotesting for women 30 years and older, with repeat cotesting in 1 year triage for women with normal cytology results and positive hrHPV test results (hereafter referred to as “cotesting with repeat cotesting triage”);

• cytology for women aged 21 to 29 years, followed by cotesting for women 30 years and older, with immediate HPV16/18 genotyping triage for women with normal cytology results and positive hrHPV test results (hereafter referred to as “cotesting with genotyping triage”);

• cytology for women aged 21 to 29 years, followed by primary hrHPV testing for women 30 years and older, with immediate cytology triage for women with positive hrHPV test results (hereafter referred to as “primary HPV with cytology triage”);

• cytology for women aged 21 to 29 years, followed by primary hrHPV testing for women 30 years and older, with immediate HPV16/18 genotyping triage for women with positive hrHPV test results (hereafter referred to as “primary HPV with genotyping triage”).

Clinical algorithms and utility maps for the five screening strategies have been previously described in detail.[14] Abnormal screening follow-up and post-treatment surveillance correspond to the recommended algorithms for average-risk women.

Validation targets

Internal validation was achieved by comparing the model-predicted natural history outcomes with the calibration targets (i.e., HPV prevalence and CIN2+ incidence in WWH).[28] External validation was achieved by comparing the model-predicted outcomes to data not used to inform the model, in particular, cervical cancer incidence in WWH and the proportion of abnormal cytology results in WWH. Among WWH, studies have shown that the cervical cancer incidence was 4.1 times that of women without HIV (95% CI 2.3–6.6)[3] and 16%−38% of the cytology test results were abnormal.[29,30]

Analysis

The primary outcomes for each scenario were: colposcopies, false-positive test results, treatments, early- and late-stage cancers, cancer deaths per 1000 WWH as well as life-years (to age 85, undiscounted and discounted at an annual rate of 3%), lifetime costs (to age 85, discounted at an annual rate of 3%), and lifetime quality-adjusted life-years (QALYs) (to age 85, discounted at an annual rate of 3%) per WWH. The benefits of screening were quantified in terms of the proportion of cancers and cancer deaths prevented and prolonged life-years compared with no screening. The harms of screening were quantified in terms of the number of false-positive test results, decreased QALYs, and increased lifetime costs compared with no screening. We also evaluated the discounted (at an annual rate of 3%) incremental cost-effectiveness ratio (ICER) by dividing the additional cost by the additional health benefit of a specific strategy compared with a less costly, nondominated strategy.

We performed both deterministic and probabilistic sensitivity analyses. The deterministic sensitivity analyses examined the independent effect of utility losses associated with various aspects of screening, follow-up, and treatment on the ICERs. We excluded the utility losses in a stepwise manner as follows: primary screening, surveillance, false-positive testing, and lastly, treatment and cancer.

Probabilistic sensitivity analyses were conducted to account for uncertainty in model input parameters. Uncertainty of model parameters considered in the probabilistic sensitivity analyses included the natural history parameters, obtained from the Bayesian calibration and literature, costs, utilities, and screening test accuracy estimates from published studies. We randomly sampled 1000 parameter sets from their corresponding distributions and ran the model at each of these sets. For each strategy, we calculated its probability of being cost-effective and the probability that the optimal strategy (i.e., the strategy with the highest expected net benefit) is cost-effective, both across a wide range of cost-effectiveness thresholds ($1000-$250 000 per QALY).

Model development and reporting were performed under the Consolidated Health Economic Evaluation Reporting Standards reporting guideline (Supplementary Table 1). The study is exempt from institutional review board review, because it uses publicly available and published data.

Results

Model validation

Supplementary Figure 1–3 show that the calibrated parameter sets propagate a reasonable range that fits to both the calibration targets of the HPV prevalence and CIN2+ incidence targets in WWH as well as the validation target of cancer incidence among WWH. In addition, the model was validated against the proportion of abnormal cytology results in WWH screened by cytology tests. The calibrated model predicted a proportion of abnormal cytology results in WWH ranging from 14% to 31% depending on the screening history, which is within the reported range (16%−38%).[29,30]

Screening strategies

All screening strategies evaluated were effective in reducing cervical cancers and cancer deaths among WWH. Compared with no screening, the model predicted that > 96% of cervical cancers and cancer deaths could be prevented by screening, with an increase in life-years ranging from 0.86 to 0.87. Table 2 presents the screening, surveillance, treatment, and cervical cancer outcomes per 1000 WWH. In general, screening tests with higher sensitivity and lower specificity resulted in more colposcopies, treatments, and false-positive results but were more effective in reducing cancers and cancer deaths and prolonging life expectancy (measured in life-years). Cotesting with genotyping triage resulted in numbers of colposcopies and treatments that were, respectively, 2.4 and 2.6 times the numbers resulting from cytology with HPV triage of ASC-US and was also associated with approximately 40% fewer cancer cases and cancer deaths. Of the five screening strategies, the most effective strategy in reducing cervical cancer and deaths was cotesting with genotyping triage. Primary HPV-based screening prevented more cancers and cancer deaths than cytology with HPV triage of ASC-US while resulting in fewer colposcopies than cotesting-based screening.

Table 2.

Screening, surveillance, treatment, cervical cancer outcomes across cervical cancer screening strategies.^a

Screening strategy	Screening, Surveillance, and Treatment Outcomes, No. per 1000 WWH			Cervical Cancer Outcomes, No. per 1000 WWH				Average life-years per WWH
	Colposcopies	False-positive test results	Treatments	Stage 1	Stages 2–4	All stages	Cancer deaths	Undiscounted	Discounted
No screening	0	0	0	26.04	27.68	53.72	43.11	54.9311	25.6951
Primary HPV with cytology triage	10086	9145	1065	0.99	0.24	1.23	0.94	55.7975	25.8781
Primary HPV with genotyping triage	14157	13196	1071	0.95	0.23	1.18	0.90	55.7984	25.8783
Cotesting with repeat cotesting triage	16489	15567	2359	0.87	0.22	1.09	0.83	55.7999	25.8786
Cotesting with genotyping triage	19219	18287	2360	0.87	0.22	1.09	0.83	55.8000	25.8786
Cytology with HPV triage of ASC-US	7958	7018	921	1.41	0.35	1.76	1.35	55.7884	25.8761

Abbreviations: HPV, human papillomavirus; ASC-US, atypical squamous cells of undetermined significance.

^aAll outcomes were estimated up to age 85.

Cost-effectiveness

Screening was cost-saving compared with no screening ($2597-$4400 per woman vs $7583 per woman). The base-case cost-effectiveness analysis results are shown in Table 3. Lifetime QALYs ranged from 24.90 to 25.19, and lifetime costs ranged from $2,597 to $4,400 per woman. Cytology with HPV triage of ASC-US yielded the highest QALYs (25.19) and lowest cost ($2,597 per woman). Cotesting and primary HPV test-based strategies yielded fewer QALYs and higher costs and were dominated by cytology with HPV triage of ASC-US. When only considering primary HPV testing and cotesting strategies, primary HPV with genotyping triage was associated with more QALYs and costs than primary HPV with cytology triage (ICER= $30,828 per QALYs) and was cost-effective at a willingness-to-pay (WTP) threshold of $50,000. Both cotesting strategies were dominated by primary HPV with genotyping triage. Comparing the two cotesting strategies, genotyping triage yielded a higher number of QALYs and costs than triage with repeat cotesting in 12 months (ICER=$35,112 per QALY).

Table 3.

Estimated costs, quality adjusted life-years (QALYs), and incremental cost-effectiveness ratios.^a

Screening strategy	Cost	Incremental cost, $	Quality adjusted life-years	Incremental quality adjusted life-years	Incremental cost-effectiveness ratio
No screening	$7583	Not applicable	Not applicable	Not applicable	Not applicable
Comparing all strategies
Cytology with HPV triage of ASC-US	$2597	Not applicable	25.1930	Not applicable	Not applicable
Primary HPV with cytology triage	$3307	710	24.9425	−0.2505	Dominated
Primary HPV with genotyping triage	$3537	940	24.9500	−0.2430	Dominated
Cotesting with repeat cotesting triage	$4195	1597	24.9042	−0.2888	Dominated
Cotesting with genotyping triage	$4400	1803	24.9100	−0.2830	Dominated
Comparing triage strategies for primary HPV
Primary HPV with cytology triage	$3307	Not applicable	24.9425	Not applicable	Not applicable
Primary HPV with genotyping triage	$3537	230	24.9500	0.0075	$30,828/QALY
Comparing triage strategies for cotesting
Cotesting with repeat cotesting triage	$4195	Not applicable	24.9042	Not applicable	Not applicable
Cotesting with genotyping triage	$4400	205	24.9100	0.0058	$35,112/QALY
Comparing all HPV-based strategiesb
Primary HPV with cytology triage	$3307	Not applicable	24.9425	Not applicable	Not applicable
Primary HPV with genotyping triage	$3537	230	24.9500	0.0075	$30,828/QALY
Cotesting with repeat cotesting triage	$4195	658	24.9042	−0.0458	Dominated
Cotesting with genotyping triage	$4400	205	24.9100	−0.0399	Dominated

Abbreviations: HPV, human papillomavirus; ASC-US, atypical squamous cells of undetermined significance

^aAll outcomes estimated up to age 85

^bAll HPV-based strategies assumed cytology-based screening ages 21–29 followed by HPV-based strategies ages 30 and older.

Sensitivity Analyses

The impact of the stepwise removal of utility losses on QALYs is shown in Figure 1. Although utilities associated with HPV test-related health states were generally lower than those for cytology testing, excluding the utility losses associated with screening did not change the cost-effectiveness results; cytology with HPV triage of ASC-US dominated all other strategies. When all utility losses were excluded (i.e., the effectiveness was represented as life-years), cytology with HPV triage of ASC-US triage yielded the lowest average life-years and costs and cotesting with genotyping triage was dominated by cotesting with repeat cotesting triage. If cytology with HPV triage of ASC-US was used as the reference, the ICERs for the other four screening strategies ranged from $368,354 per life-year gained (primary HPV with cytology triage) to $736,687 per life-year gained (cotesting with genotyping triage).

Figure 1.

Effect of sequential removal of utility losses on estimated quality adjusted life-years (QALYs)/life-years.

Probabilistic sensitivity analyses showed that cytology with HPV triage of ASC-US dominated all other strategies regardless of the WTP threshold. Figure 2 illustrates the probability of each strategy being cost-effective at various WTP thresholds (ranging from 0 to $250,000) when only considering primary HPV testing and cotesting strategies. At the WTP thresholds of $50,000, $100,000, and $250,000, the two primary HPV testing strategies were cost-effective in 99.1%, 98.5%, and 98% of the iterations, respectively.

Figure 2.

The probability of each strategy being cost-effective and the probability that the optimal strategy is cost-effective at various cost-effectiveness thresholds.

Discussion

Studies evaluating the cost-effectiveness of currently recommended cervical cancer screening strategies and primary HPV screening for WWH in the U.S. are lacking. We compared the effectiveness and cost-effectiveness of two primary HPV screening strategies with recommended cytology-based screening and cotesting screening. Compared with no screening, the model predicted > 96% of cervical cancers and cancer deaths in WWH could be prevented by screening, with an increase in life-years ranging from 0.86–0.87. However, different strategies were associated with a range in the number of tests performed and effectiveness in prolonging life. Specifically, cotesting strategies were predicted to prevent the most cancers and cancer deaths but were also associated with the highest number of screening tests and false-positive test results.

The ranking of strategies based on their effectiveness in prolonging life for WWH is similar to findings from a previous study we conducted for women at average risk of cervical cancer and also consistent with other modeled evaluations.[14,31] In particular, Kim et al. modeled the benefits and harms of various cervical cancer screening strategies for a cohort of average-risk women in the U.S. They noted higher predicted life-years with cotesting and primary HPV screening compared with cytology-based screening; for primary HPV screening, HPV genotyping triage was more effective in increasing life-years compared with cytology triage.

Despite the similar relative rankings, the absolute counts of screening tests performed, false-positive test results, and treatments were much higher in WWH compared with average-risk women. Undergoing screening and being informed of abnormal test results are both associated with a lower utility.[14,32] Screening strategies associated with higher number of screening tests and follow-up procedures highlight the potential harms of over-screening in this population. Therefore, we compared the benefit-to-harm balance of different screening strategies. When we compared the cost-effectiveness of primary HPV testing and cotesting, primary HPV testing was cost-effective in over 98% of simulations at various WTP thresholds (ranging from 0 to $250,000), regardless of the triage option, indicating primary HPV screening could be a cost-effective alternative to cotesting for WWH. Our results are consistent with observed screening outcomes. Strickler et al. noted that among 865 WWH in the Women’s Interagency HIV Study, screening with primary HPV testing resulted in fewer unnecessary colposcopies than cotesting under some conditions of patient management and could be a potential alternative to cotesting in WWH.[12]

Among all strategies evaluated, cytology with HPV triage of ASC-US yielded the most QALYs, dominating primary HPV testing and cotesting strategies, regardless of the WTP threshold or utility losses associated with different screening stages. The current study and our previous cost-effective analysis of various screening strategies among average-risk women suggest that, regardless of HIV infection status, cytology-based screening provided higher QALYs with lower costs, and therefore, dominated primary HPV testing and cotesting strategies.[14] These results suggest that the current cytology-based screening recommendations for WWH are cost-effective and could reduce the harms associated with over-screening.

We explored the reasons for lower QALYs associated with primary HPV and cotesting screening. The most important driver of the differences noted is the preference for cytology tests over HPV tests. [33] Although true positive cytology results reflect HPV infection, HPV testing was associated with lower utility values even when test results were comparable. These findings highlight the need to address the acceptance and awareness of HPV testing through education programs. For example, evidence suggests that women participating in HPV self-sampling studies have consistently reported a higher preference toward self-sampling versus clinician-collected sampling [34], and efforts are being put into promoting HPV self-sampling globally. [35] Self-sampling instruments have recently been approved in the U.S.; with widespread availability, women may become more accustomed to and more accepting of HPV testing.

Although HPV-based testing is more sensitive than cytology testing and prevents more cancers and cancer deaths, the specificity is lower compared with cytology.[23] Lower test specificity results in more false-positive test results under the current follow-up recommendations, introducing more follow-up tests and lower QALYs. Such an effect is expected to be more pronounced in WWH due to the higher prevalence of HPV or CIN1 compared with average-risk women.[3] With HPV test-based screening, many would be referred to colposcopies who do not have CIN2+ diagnosed; resulting in increased surveillance. As a result, the model-predicted benefits associated with cancer prevented in this high-risk group may be outweighed by the harms.

It remains unclear whether the current recommendations for follow-up after an abnormal test result are appropriate for WWH. Current follow-up criteria correspond to the estimated risks for women with average risk.[6] Whether a patient should receive immediate treatment or remain under surveillance depends on the estimated immediate CIN3+ risk. Given the higher likelihood of progression, WWH with positive HPV results or CIN1 diagnoses probably have a higher immediate CIN3+ risk than average-risk women with the same test results. Estimates of CIN3+ risks among WWH with different screening results are needed to determine whether expedited treatment is preferred, possibly reducing the number of unnecessary follow-up procedures and increasing the QALYs.

Our study has several strengths. First, to our knowledge, this is the first cost-effectiveness analysis that included currently recommended cytology-based screening and cotesting, as well as primary HPV screening for WWH in the U.S. Our study identified gaps that need to be addressed to provide high-value care to this high-risk population. Second, our model was previously validated in average-risk women. In this study, we re-calibrated and validated the model using both natural history and screening outcome targets to ensure our revised model well reflects WWH in the U.S. Third, to assess the influence of uncertain parameters and evaluate the robustness of our results, we conducted several sensitivity analyses. Our model predictions were robust across these analyses.

Our study has several limitations. In reality, WWH could experience lower QALYs due to HIV-related illness. However, our model structure wouldn’t allow us to distinguish HIV infection status or other factors that influence the severity of immunosuppression such as whether on ART treatment or CD4 counts.[36] Therefore, we assumed that the impact on QALYs would be independent of the HIV status of a woman, and decreases in QALYs were only driven by increased cancers and undergoing a screening, surveillance, or diagnostic test.

Similarly, due to the limitation of the current model structure, we did not specifically model the immunosuppressive status of WWH. We assumed that a general evaluation of various screening for WWH in the US as a high-risk group as a whole is more pressing given the lack of any recent modeling study for this group of women.

We applied the utilities associated with screening and treatment in the model wherever these events occurred. How long to apply these utilities has not been determined, especially when screening is conducted over an extended period of time, as is the case with cervical cancer screening for WWH. Depending on the colposcopy results, women can remain under surveillance for many years. Whether the first and the following few abnormal results are associated with the same loss in QALYs remains unclear. Such limited information highlights the need to evaluate different recommendations following an abnormal screening result for WWH to provide high-quality care for this high-risk population.

Conclusion

Cervical cancer screening for WWH is necessary and effectively prevents cervical cancer cases and cancer deaths in this high-risk population. Our modeling study suggests the current cytology-based screening recommendation is cost-effective, but that primary HPV screening could be an alternative to cotesting. To improve the cost-effectiveness of HPV-based screening, increased acceptance is needed as is an evaluation of alternative follow-up recommendations to limit the harms of over-screening in this high-risk population.

Abbreviations:

ASC-US

atypical squamous cells of undetermined significance

CIN

cervical intraepithelial neoplasia

HIV

human immunodeficiency virus

HPV

human papillomavirus

hrHPV

high-risk HPV

HR

hazard ratio

ICER

incremental cost-effectiveness ratio

QALYs

quality adjusted life-years

RRs

relative risks

SEER

Surveillance Epidemiology and End Results

WWH

women with HIV

WTP

willingness-to-pay

Data Availability Statement

Full details of the model are provided in Sawaya et al. JAMA IM 2018 with additional details provided in the Methods section. We are happy to discuss potential collaborations with interested researchers.