Norwegian Cervical Cancer Prevention Supplementary Appendix
SUPPLEMENTARY APPENDIX
Accompanying the manuscript:
Cost-effectiveness of cervical cancer screening with primary human papillomavirus (HPV) testing in Norway
Emily A. Burger, MPhil
Jesse D. Ortendahl, BS
Stephen Sy, BS
Ivar Sonbo Kristiansen, MD, MPH, PhD
Jane J. Kim, PhD
Part I: Model and parameterization
Part II: Cost assumptions
Part III: Strategies and assumptions
Part IV: Additional Results
Part V: References
Part I: Model and parameterization
The first-order Monte Carlo simulation model (i.e., stochastic)of cervical cancer has been previously described (1-3). The model, comprising of mutually exclusive, collectively exhaustive health states, follows individual women throughout their lives, calculates lifetime cervical cancer risk, cancer incidence and mortality, and life expectancy. The model also tracks costs associated with events such as vaccination, screening, diagnostic follow-up, treatment of precancer, and cancer treatment and care. By simulating a large number of individual women, expected health benefits and costs of alternative prevention policies that may include screening, vaccination or both, can be assessed. Baseline transition parameter values describing the natural history of disease were based on the best available empirical data, have been previously published and assume that the underlying mechanism of cervical carcinogenesis does not vary across epidemiological settings(1-3). Risk factors, such as sexual behaviour, and cervical cancer incidence rates differ between countries; therefore, country-specific data are needed to adjust baseline inputs to account for variations in progression and regression rates. We leveraged empirical data from Norway and used a likelihood-based algorithm to identify candidate sets of parameter values that achieve good-fit to epidemiological outcomes observed in the Norwegian population.
Defining calibration targets
In total, 37 calibration targets were defined. The Norway-specific targets included age-specific prevalence of HPV-16, -18 in women, age-specific prevalence of CIN23, HPV-16, -18 and other high-risk HPV distributions in high-grade CIN, HPV-16 and -18 distributions in cervical cancer and age-specific cancer incidence. For each calibration target, we determined a point estimate and confidence interval, using population-based sources.
Calibration target data was used to inform multipliers of the initial model inputs. All prevalence and HPV type distribution targets were calculated using 95% confidence intervals of the binomial distribution in STAT/SE 11.0, and cancer incidence bounds were informed by taking the minimum and maximum age-specific annual incidence from 1953-1969.
Calibration target data sources and model fitting
Age-specific prevalence of HPV-16,-18
There are limited number of HPV prevalence studies which have been published in Norway and even fewer which inform the prevalence of HPV among younger women. Three published studies (4-7) were identified by a literature search; however, we were not able to extract the pertinent data to inform our model. For one study (6;7), the reported estimates pooled the prevalence from only five high-risk HPV types or did not separate high-risk types from low-risk types, and the second study did not separate HPV-16, -18, -6,-11 from each other. The third study (4), reported the prevalence of HPV-16 and -18 among younger women, though we could not simultaneously stratify the age groups and attribute multiple HPV infections hierarchically from the published manuscript. The best available data came from a study affiliated with the Norwegian Cancer Registry (Personal communication: Mari Nygaard, MD, PhD, March 2011). They gave us preliminary insight to results from a study they conducted in a large city in Norway (Appendix Table and Figure 1). The study selected a random sample of Norwegian women in 18-45 yrs of age, who attended to screening in 2007 in S. Olav hospital, in Trondheim, Norway. HPV DNA positivity was detected by PCR G5+/6+ for 30+ HPV types. We assumed participants to be sexually active and used a weighted average from two sexual behavior studies conducted in Norway (8;9) to adjust for sexual debut in the younger age groups. The adjustment factor for women aged 18-19 and 20-24 years was 80% and 95%, respectively.
Appendix Table 1: Age-specific prevalence of HPV-16,-18 in women (adjusted for non-sexually active women) with 95% confidence intervals
AGE GROUP / N / X / PREV HPV-16/18 / LB / UB15-17
18-19 / 77 / 16 / 0.2085 / 0.1254 / 0.3192
20-24 / 689 / 151 / 0.2190 / 0.1826 / 0.2440
25-29 / 296 / 46 / 0.1554 / 0.1183 / 0.2013
30-34 / 213 / 11 / 0.0516 / 0.0281 / 0.0911
35-39 / 235 / 10 / 0.0426 / 0.0223 / 0.0775
40-44 / 234 / 1 / 0.0043 / 0.0001 / 0.0263
45-49 / 57 / 2 / 0.0351 / 0.0027 / 0.1261
Appendix Figure 1: Model output from 50 good-fitting sets and the upper and lower bound from the empirical data (bold). 5 best fitting sets in red.
Prevalence of high-grade cervical lesions by age
We used a published study (6;7) in which Pap smears were taken for a cytological analysis from 4419 women (Appendix Table and Figure 2). These women visited selected specialist gynecological clinics in Oslo, and the samples were taken consecutively in the period from February to June 2001. The gynecologists performed the Pap smears, which were then screened by experienced cyto-technologists at two different laboratories. The cytology was evaluated independently of the HPV testing. Distribution of lesions was reported by for women ≤29 and by 10-year age groups, thereafter.
Appendix Table 2: Age-specific prevalence of CIN23 in women
AGE GROUP / N / X / Prev / PREV CIN2-3 (X*1.4)(1) / LB(X) / UB(X*1.4)≤29 / 283 / 0 / 0 / 0.0000 / 0.0000 / 0.0130
30-39 / 1023 / 14 / 0.0140 / 0.0196 / 0.0075 / 0.0300
40-49 / 1211 / 6 / 0.0050 / 0.0070 / 0.0018 / 0.0130
50-59 / 1208 / 4 / 0.0030 / 0.0042 / 0.0009 / 0.0096
60+ / 694 / 1 / 0.0010 / 0.0014 / 0.0000 / 0.0080
(1) Prevalence was corrected by 40% false negatives in all age-groups.
Appendix Figure 2: Model output from 50 good-fitting sets and the upper and lower bound from the empirical data (bold). 5 best fitting sets in red.
Distribution of HPV-16, -18 and other high-risk types among CIN and distribution of HPV-16, -18 among cervical cancer
We reported the proportion of HPV-16,-18 and other high-risk HPV in CIN23 from a Norwegian epidemiological study (working paper) of HPV type distribution in high-grade cervical pre-cancer using standardized HPV DNA detection and typing on archived, formalin-fixed, paraffin-embedded cervical and excision specimens (personal communication: Steinar Thoresen) (Appendix Table and Figure 3). The upper and lower bounds from this study were consistent with a study identified in the literature search(10), but was not utilized because we could not attribute other high-risk types hierarchically to estimate the distribution of other high-risk HPV types among CIN. We identified two publications during our literature search which reported HPV type distributions within cervical cancer, however, one (11) used invasive cervical cancer specimens from the late 80’s which may no longer be representative of current HPV distributions and we could not separate CIN3 from invasive cancer from the other (12). We estimated HPV-16, -18 type distribution among invasive cancer from the same working paper as above (personal communication: Steinar Thoresen, PhD, February 2011). Women were aged 18 or above at the time of collection of cervical/excision specimens, and had been diagnosed with invasive cervical cancer from 2001 onwards. HPV type distribution targets were calculated using 95% confidence intervals of the binomial distribution in STAT/SE 11.0.
Appendix Table 3: Distribution of HPV-16,-18 and other high-risk types in high-grade CIN and HPV-16,-18 types in cervical cancer with 95% confidence intervals
Lesion & HPV-type / N / X / PREV TYPE / LB / UBCIN 23 HR_16 / 255 / 121 / 0.4745 / 0.4119 / 0.5377
CIN 23 HR_18 / 255 / 17 / 0.0667 / 0.0393 / 0.1046
CIN 23 HR_Other / 255 / 103 / 0.4039 / 0.3432 / 0.4669
CANCER HR_16 / 342 / 163 / 0.4766 / 0.4226 / 0.5310
CANCER HR_18 / 342 / 66 / 0.1930 / 0.1525 / 0.2389
Appendix Figure 3: Model output from 50 good-fitting sets and the upper and lower bound from the empirical data (bold). 5 best fitting sets in red.
Cervical cancer incidence
To evaluate model outcomes on the natural history of disease in the absence of screening, targets on the age-specific incidence of invasive cervical cancer were defined based on the minimum and maximum annual incidence from Norwegian Cancer Registry data, 1953-1969 (Contact: Gry Skare, Cancer Registry of Norway) (Appendix Table and Figure 4). The registry is based on a modified version of International Classification of Disease, version 7 or version O (ICD-7/ICD-O). Staging is done according to Federation Internationale Gynecologie et d’Obstetrique (FIGO). The registration of invasive cervical cancer is nearly 100% complete in Norway (13). The 50 good-fitting sets in relation to the empirical cancer incidence bounds are shown in Appendix Figure 5.
Page 1 of 18
Norwegian Cervical Cancer Prevention Supplementary Appendix
Appendix Table 4: Incidence per 100,000 womann-years, by year and age (Cumulative from stages I-IV and unknown).
AGE GROUP / 1953 / 1954 / 1955 / 1956 / 1957 / 1958 / 1959 / 1960 / 1961 / 1962 / 1963 / 1964 / 1965 / 1966 / 1967 / 1968 / 1969 / LB (min) / UB (max)10-14 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00
15-19 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.00 / 0.68 / 0.00 / 0.00 / 0.68
20-24 / 0.96 / 0.00 / 0.00 / 2.06 / 1.04 / 1.02 / 0.00 / 0.97 / 0.95 / 0.92 / 0.00 / 0.00 / 0.78 / 0.00 / 0.00 / 1.98 / 1.31 / 0.00 / 2.06
25-29 / 6.88 / 11.53 / 3.60 / 6.49 / 6.59 / 2.92 / 6.99 / 13.38 / 0.00 / 3.15 / 9.27 / 7.05 / 4.90 / 8.71 / 4.67 / 8.04 / 6.72 / 0.00 / 13.38
30-34 / 21.78 / 23.99 / 17.94 / 20.24 / 21.82 / 23.46 / 23.30 / 22.86 / 17.81 / 10.46 / 20.60 / 19.08 / 22.69 / 15.75 / 17.95 / 22.78 / 23.22 / 10.46 / 23.99
35-39 / 26.81 / 22.26 / 33.91 / 36.46 / 39.44 / 46.35 / 31.14 / 38.89 / 43.27 / 32.12 / 35.82 / 40.53 / 40.36 / 34.85 / 31.53 / 42.35 / 34.23 / 22.26 / 46.35
40-44 / 39.36 / 31.78 / 36.46 / 30.70 / 41.78 / 41.44 / 39.29 / 34.91 / 48.24 / 38.24 / 43.63 / 55.02 / 40.71 / 50.88 / 58.65 / 36.91 / 57.93 / 30.70 / 58.65
45-49 / 37.16 / 44.49 / 37.05 / 37.50 / 36.88 / 40.64 / 43.77 / 28.79 / 40.88 / 40.64 / 39.46 / 48.56 / 43.07 / 44.86 / 49.37 / 47.84 / 50.05 / 28.79 / 50.05
50-54 / 36.51 / 44.47 / 34.96 / 38.39 / 32.66 / 34.25 / 37.27 / 44.75 / 24.27 / 34.93 / 39.62 / 33.55 / 48.40 / 46.45 / 42.02 / 39.21 / 50.12 / 24.27 / 50.12
55-59 / 47.19 / 49.10 / 41.97 / 38.43 / 38.82 / 28.86 / 37.12 / 42.49 / 43.10 / 34.42 / 35.08 / 41.80 / 35.91 / 30.99 / 33.98 / 33.58 / 31.72 / 28.86 / 49.10
60-64 / 33.25 / 33.47 / 32.48 / 32.59 / 22.57 / 47.03 / 26.68 / 32.39 / 44.20 / 26.28 / 38.08 / 29.71 / 32.41 / 34.01 / 41.42 / 29.64 / 29.18 / 22.57 / 47.03
65-69 / 18.47 / 33.32 / 26.64 / 27.40 / 35.27 / 30.13 / 38.52 / 28.35 / 11.22 / 43.50 / 23.42 / 29.73 / 37.99 / 25.11 / 43.07 / 30.81 / 26.24 / 11.22 / 43.50
70-74 / 33.93 / 9.76 / 26.42 / 25.74 / 30.50 / 29.73 / 27.56 / 23.48 / 29.42 / 28.85 / 24.92 / 22.62 / 33.59 / 26.49 / 28.35 / 31.41 / 34.43 / 9.76 / 34.43
75+ / 23.78 / 27.38 / 21.61 / 11.87 / 15.52 / 29.18 / 12.37 / 27.80 / 26.04 / 11.56 / 19.44 / 20.20 / 22.88 / 21.94 / 15.09 / 17.52 / 25.50 / 11.56 / 29.18
Page 1 of 18
Norwegian Cervical Cancer Prevention Supplementary Appendix
Appendix Figure 5:Model output from 50 good-fitting sets (light grey) and the upper and lower bound from the empirical data (bold).
Part II: Cost assumptions
Direct medical and non-medical costs of screening, diagnosis and treatment, were based on a combination of primary data and expert opinion. There are evidence gaps in Norway with respect to detailed costing estimates for cervical cancer screening and treatment; therefore, we present details of our estimation and rationales below. Input values and ranges used in this analysis are presented in the Main Manuscript Table 1.
Estimation of costs
Direct medical costs of screening included the test, supplies, specimen transport, laboratory processing of the screening sample, staff time, and the office visit. Diagnostic costs included colposcopy, biopsy, supplies, equipment, laboratory processing, staff time, and the office visit. Pre-cancer treatment costs included the procedure, which included pharmaceuticals and supplies, complications, and hospitalization, and the facility visit. Direct medical costs of cancer care included staging of cancer severity, work-up, hospitalization, stage-appropriate treatment, and follow-up visits for five years (discounted). Direct non-medical costs and time costs associated with screening, diagnosis, and treatment for precancerous lesions and invasive cancer included all patient time in transport, waiting, receiving treatment, and in hospitalization as well as actual transport costs. Invasive cervical cancer stages Ia-IIa were classified as local cancer, stages IIb-IIIb as regional cancer, and stages IVa-IVb as distant cancer, based on FIGO staging system.
To estimate costs associated screening, diagnosis and treatment, we based costs on official national tariffs (outpatient care) (14;15) and hospital-based DRG reimbursement rates (inpatient care) (16) using official treatment guidelines (17) and expert opinion from Norwegian gynaecologists to quantify resource use. See Appendix Table 5 for descriptive estimates of the direct medical and non-medical costs associated with screening, treatment and vaccination.
Appendix Table 5: Description of estimates used in base case analysis
CATEGORY / DESCRIPTION /2010 ($)
Office costs / Weighted average of GP based visit (80%) and gynecologist-based visit (20%). Cost of staff, facilities for a common "office" visit in a primary clinic -- exam plus a basic clinic room with an exam bed, sending of sample to laboratory and letter/results / $81
Patient time for office visit / Cost of patient time (2 way travel (60min), waiting(15min), receiving care (15min)) for a primary clinic + the cost of 2 way patient transport to and from the primary clinic / $87
Conventional Pap test / Cost of pap collection materials, rubber gloves, disinfectant, glass slide of tube, speculum, other disposable supplies, includes all laboratory transport, equipment, supplies, facilities, staff (Independent/adjusted analysis) (material fee of $3 added to estimated lab costs, explained below). / $49
Liquid-based cytology / Cost of pap collection materials, rubber gloves, disinfectant, tube, speculum, other disposable supplies, includes all laboratory transport, equipment, supplies, facilities, staff (Independent/adjusted analysis) / $50
HPV DNA test / Cost of HPV collection kit, rubber gloves, disinfectant, speculum, other disposable supplies, includes all laboratory transport, equipment, supplies, facilities, staff (Independent/adjusted analysis) Includes co-collection fee for LBC (assumed $8), regardless if LBC is conducted) / $62
Patient time for colposcopy / Cost of patient time for 2-way travel, waiting, and receiving services at a district/regional hospital + cost of patient transport / $138
Colposcopy procedure / Cost of the facilities, staff time to perform colposcopy. Cost of supplies and equipment to do colposcopy (speculum, colposcope, rubber gloves, disinfectant, etc), Including taking and analyzing biopsy. / $199
CIN1 treatment / Cost of treating a person who has true CIN1. This is a weighted average of LEEP, conisation, and simple hysterectomy for people with this true lesion status and includes the treatment specific staff time, supplies, equipment, hospitalization, and follow-up visits and procedures as well as patient time receiving services, hospitalization, and follow-up and patient transport for the same / $1,024
CIN23 treatment / Cost of treating a person who has true CIN23. This is a weighted average of LEEP, conisation, and simple hysterectomy for people with this true lesion status and includes the treatment specific staff time, supplies, equipment, hospitalization, and follow-up visits and procedures as well as patient time receiving services, hospitalization, and follow-up and patient transport for the same / $2,162
Local cancer treatment / Diagnosis, conisation (19%), simple (19%) and radical hysterectomy (41%), radiotherapy and/or adjuvant chemo (19%), fertility preserving (2%), complications (10%), relapse/retreat (20%), recommended follow-up for 5 years conditioned on survival, transport, productivity loss for direct treatment time and f/u procedures
(Stages Ia-IIa) / $25,770
Regional cancer treatment / Diagnosis, radical hysterectomy (6%), 25 visits radiotherapy with adjuvant chemo (92%), 10 visits simplified external radiotherapy (2%), complications (10%), relapse/retreat (20%), recommended follow-up for 5 years conditioned on survival, productivity loss for treatment time and f/u procedures
(Stages IIb-IIIb) / $51,589
Distant cancer treatment / Diagnosis, 25 visit radiotherapy with adjuvant chemo (27%), 30 visits radiotherapy with boost and adjuvant chemo (50%), simplified external radiotherapy (8%), complications (90%), relapse/retreat (50%), recommended follow-up for 5 years conditioned on survival, productivity loss for treatment time and f/u procedures
(Stages IVa-IVb) / $59,635
Vaccine dose 1 / Vaccine costs (excluding VAT), wastage, supplies. Not including patient time and transport, because the vaccine is assumed to be administered through a school-based program. / $163
Vaccine dose 2 / $163
Vaccine dose 3 / $163
Women’s Time and Transport Cost Estimates
Time and transportation cost estimates for two-way travel to clinical services, including follow-up visits and hospitalization, were based on two data sources. The time spent travelling and the costs associated with transport to and from screening were approximated from a prospective study which determined the time and transportation costs for colorectal screening in Norway (18). We applied 100 Kroner for roundtrip transportation costs and 60 minutes travelling time for the screening office visit. The time spent travelling to a hospital to receive cervical cancer treatment was estimated from a health survey conducted by Statistics Norway for the World Health Organization, estimated to be 44 minutes one-way (19). The transportation cost for women seeking cancer treatment was estimated using the published deductible for roundtrip transportation for hospital-treated patients (260 Kroner) (20). We assumed a four-hour production loss for each radiotherapy or chemotherapy treatment. As a proxy for production loss, we used the 2010 average gross monthly income of Norwegian women obtained from Statistics Norway (33,500 Kroner) and adjusted the wage to include social benefits (40%) paid by employers (21).
Estimation of laboratory costs
Published reimbursement rates for the laboratory analysis of cervical screening tests (22) are thought to be underestimated. The Norwegian health care system is funded by taxes and patient co-payments, approximately 80% and 20%, respectively. Screening tests are mainly taken by general practitioners (GPs), but some are taken by private practicing gynaecologists and in hospital out-patient clinics. Hospitals are publicly funded and the great majority of them are also publicly owned. GPs are funded by patient co-payments, per capita payments from the municipalities and service fees from the Norwegian Welfare and Labour Administration (in Norwegian: “NAV”) while private practicing gynaecologists are funded by patient co-payments, block grants from the Regional Health Authorities and service fees from the Norwegian Welfare and Labor Administration. For both these physician groups we used co-payments, service-fees and per-capita payments to estimate the cost of their services. Hospitals’ in-patient and out-patient services are funded in part by the DRG system and in part by block grants. The former is supposed to represent 60% of the total costs. The DRG cost weights are based on costing in a sample of Norwegian hospitals.
Conventional cytology, liquid based cytology tests (LBC) and HPV tests are analyzed in microbiology and pathology laboratories whether they are taken by GPs, private practicing physicians or at out-patient hospital clinics. For in-patients, laboratory costs are included when estimating DRG weights, and the hospitals do not receive any additional reimbursement for laboratory tests. For all the others (i.e. the overwhelming majority of tests in the cervical cancer screening program), the laboratories are funded by means of service fees from The Norwegian Health Economics Administration (in Norwegian: ”HELFO”). In principle these service fees shall represent 40% of the total costs. For various political reasons, (see below) laboratory fees are much lower than the costs would indicate. Since the 1960s, Norway has had several private laboratories (clinical chemistry, pathology, microbiology) run by physicians working in public hospitals (mainly professors in the respective disciplines). The physicians worked in these private laboratories, presumably during their leisure time. The tests performed, however, were paid in full (fee-for-service) by the national health insurance (tax paid). Public hospitals which performed the same tests for outpatients, received the same fees as private laboratories. The profits in these private laboratories increased over the years, and different governments considered the system, which subsidized private business, unreasonable. To squeeze out the private laboratories, fees were gradually reduced – for private as well as public laboratories. Although laboratories in public hospitals disliked this squeeze, they survived because hospitals’ main revenue is block grants paid by the government. The process, however, implied that fees paid to public laboratories far from covered the real costs when taking into account that 40% of the real costs should be covered by the fees and 60% by the block grants. Because the laboratory costs represent a small proportion of the total hospital costs, the situation has been financially viable. The service fees for out-patient laboratory services, however, cannot be used as proxies for the societal costs of laboratory services.