The cost-effectiveness of human papillomavirus screening for cervical cancer

Original Papers

Jeremy Holmes1 · Lindsay Hemmett1 · Susan Garfield2

1 PMSI Healthcare, London, UK 2 Digene Corporation, Boston, Mass., USA

The cost-effectiveness of human papillomavirus screening for cervical cancerA review of recent modelling studies

Eur J Health Econom 2005 · 50:30–37DOI 10.1007/s10198-004-0254-1Published online: 29. January 2005© Springer Medizin Verlag 2005

Introduction

The recent publication of the American College of Obstetrics and Gynaecologists (ACOG) guidelines on cervical screening [] has focussed renewed attention on the role of human papillomavirus (HPV) test-ing in screening programmes for cervical cancer. The guidelines acknowledge the val-ue of HPV testing in primary cervical can-cer screening in combination with a Pap smear test for women aged 30 years or over. This reinforces the cervical cancer screen-ing guidelines of the American Cancer So-ciety [2] and the American Reproductive Health Partnership [3] which also support HPV testing for use in combination with the Pap smear test for women aged 30 ye-ars or over.

These newer guidelines are in addition to the view of the American Society for Col-poscopy and Cervical Pathology [4] which supports HPV use as the preferred method to triage abnormal or atypical squamous cells of undetermined significance (ASC-US) cervical cytology test results. In their guidelines the ACOG acknowledges the continuing concern over the high false-negative rate of cytology screening.

These newly published guidelines pro-vide a clinical rationale for using HPV test-ing with a Pap smear test for cervical can-cer screening but raise questions regard-ing the cost of implementing the proposed approach. Importantly, the guidelines ci-te a recent cost-effectiveness study which showed that additional costs associated with introducing HPV testing in conjunc-

tion with cytology (hence improving sensi-tivity) could be offset by an increase in the screening interval [5].

The latter finding strongly echoes pre-vious work in the United Kingdom which showed that considerable net savings would result from the combination of HPV and cytology testing based on screen-ing intervals of 5 years rather than 3 years, whilst at the same time avoiding more cas-es of cervical cancer [6]. In the United Sta-tes for certain low-risk or HPV negative women with otherwise normal cytological results an increase from annual or 2- to 3-yearly screening may be possible. Indeed Goldie et al. [7] have shown that 3-yearly screening using HPV testing and cytology in combination for women aged 30 or old-er provides equivalent or greater benefits than those provided by annual convention-al cytology.

There is now a growing body of evi-dence from modelling studies that exam-ine the cost-effectiveness of HPV testing from a variety of standpoints. However, the large number of screening strategies modelled varies in each of the studies, and different underlying assumptions are used. In addition, currencies and years used for cost estimates are often different, making comparison between studies problematic.

The aim of the present review was there-fore to provide a coherent overview of the findings of the key cost-effectiveness mod-els. Use of simple standardization tech-niques can facilitate an outline compari-son between the cost-effectiveness ratios derived from each model. This review

uses such techniques in order to inform assessments of the cost implications of in-corporating HPV DNA testing into cervi-cal cancer screening programmes.

Method

Based on a search of the PubMed electron-ic library from 995 to date, eight studies modelling the cost-effectiveness of HPV screening were identified in the published literature [5, 7, 8, 9, 0, , 2, 3]. Key search words were: HPV screening, HPV testing, cost-effectiveness, cost per life year, and cost per quality-adjusted life-year (QALY). Papers not reporting cost-effectiveness ra-tios, and papers reporting only clinical tri-al or vaccine studies were excluded.

Models were included in the review if they considered a consistent screening strategy regardless of patient age. Gol-die et al. [7] and Sherlaw-Johnson & Phi-lips [7] consider screening strategies that change at age 30 with the introduction of HPV testing. This two-stage approach was different to that used for the other mod-els identified, and it was therefore not pos-sible to incorporate these recent findings in the overall review of cost-effectiveness ratios; reference to these findings is, how-ever, made elsewhere in the present con-tribution.

Each model was reviewed in terms of the population characteristics, screening frequencies, cost per life year saved, cost per QALY, and assumptions as they ap-pear in the original papers regarding prev-alence and incidence of HPV infection

30 | Eur J Health Econom 1 · 2005

able comparison between papers based on this benchmark). Whilst the benchmark of “no screening” is hypothetical for most of the industrialized world, it still has a rele-vance for countries in the developing world. Importantly, it provides a more consistent and more easily understood benchmark than the conventional Pap smear, which is subject to widely different screening inter-vals even within Europe [4]. We are grate-ful to several authors for their provision of original modelling data to facilitate or val-idate this part of the standardization pro-cess.

Many of the cost-effectiveness ratios reported in the papers are based on com-parisons between a given screening strat-egy the next most cost-effective strate-gy thus giving progressively incremental cost-effectiveness ratios. In the present re-view the absolute change in costs and the change in life years or QALYs were identi-fied for each model and standardized to a common comparison with no screening.

This enabled the establishment of a common benchmark for comparison (i.e. results from papers reporting incremental cost-effectiveness ratios were adjusted to en-

and stages of dysplasia or carcinoma, sen-sitivity and specificity of screening tests, and screening and treatment costs.

Models used include Markov mod-elling, Monte Carlo stochastic simulation and decision-tree modelling. Three mod-els used discounting for both costs and out-comes at 3% per year [5, 0, ] and one model used discounting for costs only at 5% per year [2]. For technical reasons, the two models which did not use discounting [8, 3] were not adjusted to incorporate dis-counting. However, we do not believe this would materially change our findings.

Fig. 1 9 Cost per QALY versus no screening

Fig. 2 9 Cost per life-year versus no screening

31Eur J Health Econom 1 · 2005 |

Some of the models reviewed include Pap smear testing in conjunction with HPV testing, and therefore the difference in absolute cost-effectiveness between the-se strategies and either Pap smear or HPV testing on their own (strategies which are also modelled) can be seen based on the cost-effectiveness of each vs. the “no screening” benchmark. This does not of course reveal the incremental cost-effec-tiveness associated with introducing HPV testing on top of an existing Pap smear screening programme. However, the pres-ent authors take the view that it is the over-all cost-effectiveness of a single test or du-al test strategy that is the most valuable de-cision aid for policy makers, and that this premise is in line with the ACOG and oth-er guidelines advocating use of comple-mentary screening tests. Finally, for each model the cost per life year and cost per QALY gained were standardized to 2000 values [8] and converted to United Sta-tes dollars [9].

Results

Populations, screening strategies and outcomes

The populations, screening parameters and outcomes in terms of cost-effective-ness ratios as reported in the published papers are summarized in . Table 1. It is clear from this that most of the cost-effec-tiveness models reviewed have been based on United States populations. Most mod-els assume screening starts at 8 or 20 ye-ars of age, but Mittendorf et al. [2] use 25 as the starting age and van Ballegooi-jen et al. [3] use 30 years of age. Mitten-dorf et al. adopt a different approach to the population definition than the other studies considered because theirs is a co-hort-based study rather than one based on cross-sectional estimates. This mod-els 400,000 women all born in the sa-me year and screened for 20 years (to age 45), with prevalence of HPV, incidence of low-grade squamous intraepithelial lesion (LSIL) and cervical intraepithelial neopla-sia (CIN) II, and progression to CIN III in-put as probabilities for each 5-year period. Screening also stops earlier than in the oth-er models. Only Mittendorf et al. [2] and Cuzick et al. [8] explicitly compare the al-

Abstract

Eur J Health Econom 2005 · 50:30–37DOI 10.1007/s10198-004-0254-1© Springer Medizin Verlag 2005

Jeremy Holmes · Lindsay Hemmett · Susan Garfield

The cost-effectiveness of human papillomavirus screening for cervical cancer. A review of recent modelling studies

AbstractWe compared findings from recent studies modelling the cost-effectiveness of screen-ing for cervical cancer using human papillo-mavirus (HPV) testing and alternative stra-tegies. Data were standardized to facilitate comparison of costs per life year or costs per QALY gained in six studies. Absolute chang-es in costs, life years and QALYs for each strat-egy were normalized to a comparison with no screening. Costs were standardized to US$ in 2000 values. Most models assume screening starts at age 18 or 20 years. As-sumed prevalence of HPV ranges from 10% for those aged 18 years to 20% for those aged 20–25 years and drops substantially af-ter age 30. All except one model assume sen-sitivity to LSIL of 83% or higher. Two mod-els distinguish the increasing specificity of HPV testing in older age groups (up to 95% for LSIL in women aged 55 years or older). All the models include consultation costs as well as screening and treatment costs, but costs for follow-up diagnosis and treatment vary considerably. Two models also include patient time costs. Despite these differences all strategies involving HPV testing have cost per quality-adjusted life-year (QALY) ra-tios in the range of $12,400–$16,600. Costs

per life year vary more widely, the highest being $19,246 (annual screening with liq-uid cytology and HPV). However, excluding strategies using liquid cytology, the highest costs per life year for a strategy including HPV testing are under $14,000 (simultane-ous conventional cytology and HPV every two years). The cost per life year for HPV test-ing alone triennially is lower than for Pap smear testing alone biennially. Costs per QALY are generally lower than costs per life year (given the reported modelling assump-tions and settings). Even with inclusion of patient costs, no strategies involving HPV testing cost more than $16,600 per QALY. Adoption of the ACOG guidelines to include HPV testing with cytology as a screening op-tion for women aged 30 years or older there-fore appears to be cost-effective.

KeywordsScreening strategies · Human papillomavirus testing · Cost-effectiveness ratio · Cost per quality-adjusted life-year · Life years saved


Table 1

Modelled populations, screening parameters and outcomes (original values)

Reference Country Population Screening tests Screening frequency

Cost per life year saved

Cost per QALY saved

Mandelblatt et al. [5]

US Women aged ≥20 years without HIV infection

Pap smear

HPV testing

Pap smear and HPV testing

Biennal to age 65Triennial to age 65Biennal to age 75Triennial to age 75Biennial to age 100Triennial to age 100Biennal to age 65Triennial to age 65Biennal to age 75Triennial to age 75Biennial to age 100Triennial to age 100Biennal to age 65Triennial to age 65Biennal to age 75Triennial to age 75Biennial to age 100Triennial to age 100

––––––––––––––––––

$ 34,529 $ 11,835 $ 29,781 $ 11,830 $56,440$45,250(−$19,615)(−$41,529)(−$288,780)$119,644$1,810,900$100,869$103,504(−$11,871)$70,347(−$106,525)$76,183($381,467)

Goldie et al. [9]

(based on [10])US Lifetime cohort of

women aged 18–24Liquid cytology +/− reflex HPV (annual, biennial, triennial or quinquennial)

Triennial $59,600 (vs. quinquennial liquid based cytology only)

–

Kim et al. [10] US Women aged ≥18 years Conventional cytology + reflex HPV testing (specimen co-collected)Liquid cytology + reflex HPV testing (specimen co-collected)

Biennial

Biennial

$24,700–$33,600

$44,400

$20,400−$28,200

$36,100

Maxwell et al. [11]

US Cohort of women aged 18–85 (military beneficiaries)

Liquid cytology + HPV

Triennial $14,263 (vs. Pap smear triennially)

–

Mittendorf et al. [12]

Germany Cohort of 400,000 women screened aged >25 for 20 years

Smear only

HPV only

Smear + HPV

Annual

Quinquennial (biennial if +ve)Decennial

$261.95 (vs. no screening)$47.81 (compared to no screening)$54.37 (compared to no screening)

–

–

–

Van Ballegooijen et al. [13]

Nether-lands

Women aged 30–60 years; Model A assumes additional duration (vs Pap smear) of pre-clinical phase detectable from HPV DNA is 10 years (to develop into CIN)Model B assumes de-tectable pre-clinical phase of 1 year over Pap smear and lower sensitivity for high-risk HPV types

Pap smear and HPV testing

HPV testing alone

HPV DNA by signal

Triennial QuinquennialDecennialTriennial cytology plus HPV testQuinquennialDecennial−

–Model B: 18,300 DflModel A: 6,800 DflModel B: 20,100 Dfl

–Model A: 3,500 Dfl–

––––

–––


ternative screening strategies modelled with a policy of no screening; the other studies either report incremental cost-ef-fectiveness ratios or use a different com-parator benchmark.

. Table 1 also shows the large number of different screening strategies that are modelled. These include strategies based on the conventional Pap smear only (ev-ery 5 years, 3 years, 2 years or annually), in the case of Mandelblatt et al. [5] includ-ing women up to three different age lim-its (65 years, 75 years and 00 years). Kim et al. [0] and Maxwell et al. [] also mod-el strategies using liquid cytology. Strate-gies employing a combination of Pap sme-ar and HPV tests are modelled by van Bal-legooijen et al. [3], Cuzick et al. [8] and Mittendorf et al. [2] every 0 years, 5 ye-ars, 3 years and 2 years using conventional cytology, and by Kim et al. [0] and Max-well et al. [] using liquid cytology annu-ally every 2 years and every 3 years. Finally, HPV-only testing is modelled by Mandel-blatt et al. [5] (using the three different age limits), van Ballegooijen et al. [3], Cuzick et al. [8] and Mittendorf et al. [2] every 0 years, 5 years, 3 years, 2 years and annu-ally. The resulting cost-effectiveness ratios are expressed in terms of costs per life ye-ar gained, except for Mandelblatt et al. [5] who report cost per QALY and Kim et al. [0] who report both cost per life year gai-ned and cost per QALY.

HPV prevalence and disease progression

Each model assumes a certain prevalence of HPV infection, sometimes but not always broken down by age group. For the youn-gest age groups these assumptions range from 0% for 8 year olds [] to 20% for 20- to 25-year-olds [2]. Where the assumed prevalence is broken down by age group, it drops substantially after age 30. Mitten-dorf et al. [2] assume persistent HPV infec-tion to be present in 50% of women aged 25–30 years but in only 5% of women aged over 30 years. The annual probability of disease progression ranges from 0.0666 for HPV to LSIL (based on 0.2/36 months []) to 0.0046 for HPV to CIN I [9].

Screening sensitivity and specificity

The assumptions regarding sensitivity and specificity of both HPV testing and the Pap smear test (whether using conventional or liquid cytology) also vary. The lowest sensi-tivity for HPV testing is 83% in relation to LSIL [9], but for high-grade squamous in-traepithelial lesion (HSIL) this is 93% [9]. Mandelblatt et al. [5] use different sensitiv-ity rates for women aged under 55 years vs. ≥55 years both for LSIL (43% and 70%, respectively) and for HSIL (63% and 95%). Mittendorf et al. [2] use an overall sensi-tivity rate of 90% and van Ballegooijen et al. [3] 00% sensitivity for HPV+CIN in their model A. The most conservative sensitivity rates for HPV detection of dys-

plasia are therefore those used by Mandel-blatt et al. [5] in relation to LSIL (overall 55% across both age groups). All the sen-sitivity rates used in the other models are 83% or higher (Mandelblatt et al. [5] use an overall figure of 89% for HSIL). It should be noted that Mandelblatt et al. [5] also use a conservative estimate of the sensitiv-ity of the Pap smear, if ASC-US is consid-ered a negative result (63% for LSIL, 78% for HSIL or invasive cancer).

In relation to specificity it is well estab-lished that the specificity of HPV testing increases with the age of the population be-ing screened [4, 8]. The age breaks used for differential specificity rates vary across models. Mandelblatt et al. [5] use an age break of 55 years, with HPV specificity for LSIL of 80% for under 55 years and 95% for 55 years or over. Goldie [9] uses an age break of 30 years with HPV specificities for LSIL under and over this age of 75% and 85%, respectively. Other studies do not re-port HPV specificity to LSIL. However, Mandelblatt et al. [5] assume that in the old-er age group only it is superior to the Pap smear. Liquid cytology is assumed by Max-well et al. [] to improve the sensitivity but not the specificity of the Pap smear test.

Screening and treatment costs

Assumptions regarding screening and treat-ment costs also exert an influence on the cost-effectiveness ratios reported. Some studies (e.g. [9, 3]) distinguish between the costs for a normal smear result and

Table 1

Modelled populations, screening parameters and outcomes (original values)

Reference Country Population Screening tests Screening frequency


Cost per QALY saved

Cuzick et al. [8] UK Women aged 20–64 years; same assumptions for models A and B as van Ballegooijen et al. [13]

Cytology

Cytology and HPV testing

HPV testing only

TriennialQuinquennialTriennial

Quinquennial

Triennial

Quinquennial

Biennial

£390£155Model A: £900Model B: £1,050Model A: £420Model B: £570Model A: £400Model B: £715Model A: £100–$65,484(vs. same strategy triennially)

–––––––––Model B: £425–

(continued)


Original Papers

Table 2

Summary of cost-effectiveness ratios vs. no screening (US$ 2000 Values)

Mandelblatt et al. [5] Cost per QALY

Kim et al. [10]

Maxwell et al. [11] Cost per life year saved

Mittendorf et al. [12] Cost per life year saved

Van Ballegooijen et al. [13] Cost per life year saved

Cuzick et al. [8] Cost per life year saved


Cost per QALY

Pap smear every 5 years – – – – – – $369j

Pap smear every 3 years<65 years<75 years<100 years

–$11,835$11,830$11,917

––––

––––

–$4,023f

––––

–$5,233

–$929j

Pap smear every 2 years<65 years<75 years100 years

–$13,413$13,429$13,557

–$11,893a, $14,610b

–$10,755a, $13,131b

–$5,868f

––

––––

––––

––––

Pap smear every year – – – $13,535f $236g – –

Pap smear, HPV testing every 10 years – – – – $49g $ 3,121h –

Pap smear, HPV testing every 5 years – – – – – $8,400i $1,001h,J, $1,358i,J

Pap smear, HPV testing every 3 years<65 years<75 years<100 years

–$14,079$14,128$14,259

––––

––––

––––

––––

––––

–$2,145h,J, $2,502i,J

Pap smear, HPV testing every 2 years<65 years<75 years<100 years

–$16,293$16,399$16,600

–$13,827c, $13,902d

–$12,417c, $12,485d

––––

––––

––––

–––

Liquid-based cytology every 3 years – – – $6,046f – – –

Liquid-based cytology every 2 years – $13,670a, $16,951b

$12,303a, $15,144b

$8,998f – – –

Liquid-based cytology, HPV testing every 3 years

– – – $5,603f – – –

Liquid-based cytology, HPV testing every 2 years

– $15,774e $14,095e $8,636f – – –

Liquid-based cytology, HPV testing every year

– – – $19,246f – – –

HPV testing every 10 years – – – – – $1,607h –

HPV testing every 5 years – – – – $43g – $238h,J, $1,013i,J

HPV testing every 3 years<65 years<75 years<100 years

–$12,428$12,459$12,579

––––

––––

––––

––––

–$9,227i

–$953h,J, $1,704i,J

HPV testing every 2 years<65 years<75 years<100 years

$14,382$14,454$14,620

––––

––––

––––

––––

––––

––––

a Ignoring ASC-US; b Repeat cytology for ASC-US; c Simultaneous screening and HPV testing; d Women with ASC-US return within 1 month for a second visit for HPV DNA testing (two-visit HPV DNA); e Reflex HPV DNA testing-a sample co-collected at the time of the initial screening; f Includes (army) screening, diagnosis, treatment costs; g Includes out-patient and in-patient costs plus test and laboratory costs only; does not include medication, chemotherapy and indirect costsh Model A: assumes the additional duration (vs pap smear) of the pre-clinical phase detectable from HPV DNA is 10 years (to develop into CIN); i Model B: assumes detectable pre-clinical phase of only 1 year over pap smear and lower sensitivity for high-risk HPV types; j Inflated from 1988 as screening costs were taken from this year. However, treatment costs were taken from 1997 so the standardized cost effectiveness ratios are considered an overestimate. It not possible to disaggre-gate screening and treatment costs within the cost-effectiveness ratio.


the costs for an abnormal result. Others (e.g. [5]) use a fixed test cost that is simply repeated for a repeat smear. The costs for undertaking the conventional Pap smear itself range from 24 [8] to 57 [5] and up to 64 for a ThinPrep smear with an ab-normal result [0], excluding the cost of of-fice visits. Kim et al. [0] and Maxwell et al. [] include screening strategies using liq-uid as opposed to conventional cytology. Liquid cytology is assumed to cost more. Kim et al. [0] assume a marginal cost of 3 over conventional cytology; Maxwell et al. [] assume a marginal cost of 5.

The costs of undertaking an HPV test (in current prices before standardization to 2000 values) vary from 26 for the assay only [8] to 65 for storage, processing and interpretation as well as the test itself [] (all costs converted using 2000 exchange rates). In addition, consultation costs are included in all the models considered. The costs of follow-up diagnosis and treat-ment, as opposed to screening, are driven firstly by the disease progression and sen-sitivity rates used for each model and sec-ondly by the unit costs used for both ini-tial treatment and the care of advanced dis-ease. The assumed cost of initial manage-ment and treatment of CIN ranges from 469 for LSIL [5] to ,797 for LSIL []. The cost of care for advanced disease ran-ges from 2,840 [3] to 4,89 for CIN IV []. Cuzick et al. [8] assume reduced costs for the treatment of invasive and ad-vanced cancer as the level of detection th-rough screening rises, leading to cost off-sets ranging from a mean of 20 per pa-tient (5-yearly HPV testing alone in mod-el B) to 200 per patient (3-yearly cytol-ogy plus HPV testing in model A). Man-delblatt et al. [5] and Kim et al. [0] also include patient time costs in their mod-els. Mandelblatt et al. [5] assume a 5 ti-me cost for the screening visit, 9 and 50 for diagnostic evaluation of LSIL and HSIL, respectively, and 7 and 282 for initial treatment of LSIL and HSIL, rising to 2,355 for treatment of distant metastas-es. Patient costs associated with continu-ing care range from 62 (local disease) to 92 (regional or distant disease) with ter-minal care time costed at 878. Kim et al. [0] use base case estimates of 2 for pa-tient time costs associated with cytologi-cal screening. This is not increased if an

HPV DNA sample is co-collected at the same time, but it is repeated if a two-visit strategy is used.

Standardized cost-effectiveness ratios

. Table 2 summarizes the cost-effective-ness ratios reported, following standard-ization to a common comparison with no screening and conversion of costs to US 2000. Strategies involving HPV test-ing have cost per QALY ratios in the range of 2,400 (HPV testing every 3 years [5]; Pap smear plus HPV testing every 2 years [0]) to 6,600 (Pap smear plus HPV test-ing every 2 years [5]). It should be noted that Mandelblatt et al. [5] include patient time costs, and therefore the lower end of this range is conservative. The cost per QALY would be lower if only direct costs (excluding patient time) were used.

The cost per QALY for strategies involv-ing only the conventional Pap smear ran-ges from 0,755 (ignoring ASC-US [0]) to 3,557 [5]. . Fig. 1 shows the cost per QALY ratios from these two studies in rela-tion to each other. There is wider variation in the cost per life year saved, as shown in . Fig. 2. For strategies involving the con-ventional Pap smear only, the cost-effec-tiveness ratio for screening every 3 years ranges from 929 per life year saved [8] to 5,233 [3]. In fact Goldie et al. [7] have es-timated that the cost-effectiveness of life-time screening every 3 years using the con-ventional Pap smear only is dominated by the alternative strategy of using the same test every 4 years (which has a cost per life year saved of 9,400). The wide variation in these estimates indicates underlying dif-ferences in the construction of the models with regard to the use of the Pap smear.

Costs per life year saved for strategies involving the combined use of convention-al Pap smear and HPV testing are report-ed in four of the models. The highest is that reported by Kim et al. [0] (3,902), assuming women with ASC-US Pap sme-ar results return for an HPV test at a sec-ond visit within month. Patient time costs are also included in this estimate. As with the cost-effectiveness ratios report-ed for the Pap smear alone, the lowest es-timates for the combined test strategy are those by Cuzick et al. [8] (,00–,358

for every 5 years; 2,45–2,502 for every 3 years) and Mittendorf et al. [5] (43, al-though this does not include any medica-tion costs). The highest cost per life year for a strategy involving both cytology and HPV testing is 9,246, assuming annual screening with liquid cytology and HPV []. In general the costs per life year using liquid cytology are higher than any of the other strategies modelled.

In strategies using HPV alone the-re is less variation, and the estimates are lower. The most favourable modelling as-sumptions used by Cuzick et al. [8] (the additional duration vs. Pap smear of the pre-clinical phase for CIN detectable from HPV being 0 years) suggest a cost per life year of only 238 for 5-yearly screening. However, net treatment costs in this mod-el are treated as negative (due to cost off-sets associated with reduced invasive and advanced cancer treatment). The highest estimate is that by van Ballegooijen et al. [3] for 3-yearly HPV screening (9,227). . Fig. 2 shows the wide variation in these cost per life year ratios, especially for those strategies involving liquid cytology. Howev-er, the upper bound for each strategy and the clustering of estimates from the differ-ent studies is informative because no one model comprehensively compares every screening strategy.

Conclusions

Some of the difference in cost-effective-ness ratios between different models is in-evitably due to differences in the assump-tions used, including the characteriza-tion of disease progression, unit costs of screening, inclusion or exclusion of treat-ment (and its unit costs), and the practice of cervical cancer screening in the differ-ent market settings (particularly triage strategies). Some of the difference is also likely to be due to underlying modelling uncertainty, including the decision as to which specific screening strategies to mod-el. These represent significant limitations to the present review. Nevertheless, whilst there is considerable variation in the costs per life year, the highest estimate report-ed for a strategy including HPV testing with the exception of strategies using an-nual liquid cytology is under $14,000 (Pap smear and HPV every two years compared


Original Papers

to no screening). A consensus appears to be emerging that cost-effectiveness ratios in this field are substantially below the generally accepted threshold of $50,000 per life year saved [16].Cost-effectiveness ratios for screening strategies specifically involving HPV test-ing are well within this range of accept-ability across a number of different stud-ies. In particular, Mandelblatt et al. [5] ha-ve shown that the cost per QALY for HPV testing alone every 3 years is lower than that for Pap smear testing every 2 years. This is consistent with the findings of Kim et al. [10] if ASC-US results are not ignored, and the findings of Mittendorf et al. [12] as well as the previous work by Cuzick and Sasieni [6]. Kim et al. [10] argue that a pol-icy of reflex HPV testing (where the sam-ple is co-collected at the same time as the Pap smear) costs less and is more effective than repeat cytology or delayed HPV test-ing. Kim et al. [10] also show that the costs per QALY are generally lower than the costs per life year saved (given the same modelling assumptions), and their stu-dy reports markedly lower costs per QA-LY than Mandelblatt et al. [5]. This may be taken to further reinforce the findings of Mandelblatt et al. with regard to optimal screening strategies. The inclusion of pa-tient time costs in both these two mod-els would in fact be expected to raise the cost-effectiveness ratio, but even with in-clusion of these costs none of the strate-gies involving HPV testing cost more than $16,600 per QALY saved.The issue of quality adjustment to esti-mate QALYs as opposed to life years is still controversial. Outcomes expressed only in life years saved do not take account of reduced morbidity (for example, the mor-bidity associated with incidence of cancer that does not prove fatal, or the anxiety as-sociated with repeat testing), but quality weights to estimate QALYs are often diffi-cult to establish. Nevertheless, policy mak-ers are increasingly recognizing the im-portance of quality adjustment, and the National Institute for Clinical Excellence in the United Kingdom specifically recom-mends the use of health status measures to generate QALYs in the economic evalua-tion of health technologies [20].Unit costs clearly exert an influence on the modelling results. Cuzick et al. [8] use

the lowest unit costs for both Pap smear and HPV testing (as well as including nega-tive costs for invasive and advanced treat-ment), Kim et al. [10] and Mandelblatt et al. [5] use the highest unit cost for a con-ventional Pap smear, and Maxwell et al. [11] use the highest unit cost for an HPV test. Maxwell et al. [11] also use the high-est unit costs for management and treat-ment of LSIL. This is consistent with the fact that Cuzick et al. [8] report the lowest cost per life year saved (although this re-sult is driven primarily by the disease pro-gression assumptions used in their model A), and the highest costs per life year are reported by Kim et al. [10] and Maxwell et al. [11] (although the latter reports the highest overall cost per life year based on a strategy using annual liquid cytology as well as HPV testing).Overall, the models reviewed indicate that adoption of the ACOG guidelines to include HPV testing with cytology as a screening option for women aged 30 ye-ars or over would likely be cost-effective. This is supported by Sherlaw-Johnson & Philips [17] who estimate that 3-yearly HPV testing for women aged 30 and over (plus liquid cytology for positive high risk HPV types) is more cost-effective that re-peat cytology, either conventional or liq-uid, every 3 years or 5 years. This cost-ef-fectiveness is driven by the increased sen-sitivity offered by the combined test, lead-ing to a decrease in unnecessary follow-up testing, and the additional potential safely to lengthen the screening interval. Further research is needed into the practi-calities of implementing such a policy and expanding the role of HPV testing, taking into account the morbidity (as well as mor-tality) aspects of both screening and dis-ease, and the consequent impacts on qual-ity of life.

Corresponding authorJeremy Holmes

PMSI Healthcare, 64 Highgate High Street, London, N6 5HX, UK e-mail: [email protected]

Acknowledgements

The preparation of this contribution was funded by an educational grant from Digene Corporation.

Conflict of interest: No information supplied

References

1. Anonymous (2003) Clinical management guide-lines for obstetrician-gynecologists. ACOG practice bulletin no. 45. Obstet Gynaecol 102:417–427

2. Saslow D et al. (2002) American Cancer Society guideline for the early detection of cervical neopla-sia and cancer. CA Cancer J Clin 52:342–362

3. Association of Reproductive Health Professionals (2004) Advances in cervical cancer protection. Clin Proc, Sept 2003

4. Wright TC et al. (1998) Reflex human papillomavi-rus deoxyribonucleic acid testing in women with abnormal Papanicolaou smears. Am J Obstet Gy-naecol 178:962–966

5. Mandelblatt JS et al. (2002) Benefits and costs of using HPV testing to screen for cervical cancer. JA-MA 287:2372–2381

6. Cuzick J, Sasieni P (1997) Estimates of the cost im-pact of introducing human papillomavirus testing into a cervical screening programme. In: Franco, Monsonego (eds) New developments in cervical cancer screening & prevention. Oxford: Blackwell

7. Goldie S, Kim J, Wright T (2004) Cost-effectiveness of human papillomavirus DNA testing for cervical cancer screening in women aged 30 years or more. Obstet Gynaecol 103:619–631

8. Cuzick J et al. (1999) Assessing effectiveness, costs and cost-effectiveness of cervical cancer screening and HPV testing. Health Technol Assess 3:95–109

9. Goldie S (2002) Health economics and cervical cancer prevention: a global perspective. Virus Res 89:301–309

10. Kim J, Wright T, Goldie S (2002) Cost-effectiveness and alternative triage strategies for atypical squa-mous cells of undetermined significance. JAMA 287:2382–2390

11. Maxwell GL et al. (2002) Costs and effectiveness of al-ternative strategies for cervical cancer screening in military beneficiaries. Obstet Gynaecol 100:740–748

12. Mittendorf T, Schulenburg JMG von der, Petry K, Ift-ner T (2003) Economic evaluation of human papil-lomavirus screening in Germany. Eur J Health Econ 4:209–215

13. Ballegooijen M van et al. (1997) Present evidence on the value of HPV testing for cervical cancer screening: a model-based exploration of the (cost-)effectiveness. Br J Cancer 76:651–657

14. Linos A, Riza E (2000) Comparisons of cervical can-cer screening programmes in the European Union. Eur J Cancer 36:2260–2265

15. Clavel C et al. (2001) Human papillomavirus test-ing in primary screening for the detection of high-grade cervical lesions: a study of 7932 women. Br J Cancer 89:1616–1623

16. Salzmann P, Kerlikowske K, Phillips K (1997) Cost effectiveness of extending screening mammogra-phy guidelines to include women 40–49 years of age. Ann Intern Med 127:955–965

17. Sherlaw-Johnson C & Philips Z (2004) An evalua-tion of liquid-based cytology and human papil-lomavirus testing within the UK cervical cancer screening programme. BR J Cancer 91:84–91

18. Annual country-specific inflation rates taken from OECD: Wages, Costs, Unemployment & Inflation, E 073

19. Average annual exchange rates taken from http://www.oaanda.com

20. Guidance for Manufacturers & Sponsors, National Institute for Clinical Excellence, June 2001. http://www.nice.org.uk


Documents

The cost-effectiveness of human papillomavirus screening for cervical cancer