Skip Navigation



Journal of Public Health Advance Access published online on June 17, 2008
Journal of Public Health, doi:10.1093/pubmed/fdn045
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
30/3/293    most recent
fdn045v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Karnon, J.
Right arrow Articles by Smith, K. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Karnon, J.
Right arrow Articles by Smith, K. J.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2008, Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved

Cost-utility analysis of screening high-risk groups for anal cancer


Jonathan Karnon, Associate Professor in Health Economics1
Roy Jones, Research Fellow2
Carolyn Czoski-Murray, Research Fellow2
Kevin J. Smith, Honorary Senior Clinical Lecturer2

1 School of Population Health and Clinical Practice, University of Adelaide, Adelaide, SA 5005, Australia
2 School of Health and Related Research, University of Sheffield, Sheffield S1 4DA, UK


Address correspondence to Jonathan Karnon, E-mail: jonathan.karnon{at}adelaide.edu.au

Objectives Cost-utility analysis of screening for anal cancer in high-risk groups from a UK perspective.

Methods Criteria for the assessment of screening programmes were combined in a Markov model representing the natural history of anal cancer and HIV infection in the UK population of men who have sex with men (MSM). Alternative screening programmes were overlaid on the natural history model to evaluate their impact. The model was populated using data derived from a systematic review of the literature, and calibrated probabilistically to represent joint uncertainty in the input parameters.

Results Reference case results showed screening is unlikely to be cost-effective. Sensitivity analyses identified two important parameters: regression from low-grade anal intra-epithelial neoplasia (AIN) and utility effects. Increased AIN regression rates resulted in a minimum incremental cost per QALY gained of £39 405, whereas a best case scenario reduced the ratio to £20 996.

Conclusions There are major areas of uncertainty. New analyses of existing primary data, undertaken specifically to inform regression rates may usefully update key parameters at little additional cost. If these analyses increase the likelihood that screening is cost-effective, further studies of the utility effects of treatment for high-grade AIN, and potential screening attendance rates may be justified.

Keywords: cancer, economics, screening


   Introduction
TOP
 Introduction
Methods
 Results
 Discussion
 Limitations of this study
 Conclusions
 Funding
 Appendix
 References
 
Anal cancer is a rare neoplasm that has been etiologically related to a sexually transmitted agent, the human papillomavirus.1,2 In 2003, there were 722 registrations of newly diagnosed malignant neoplasm of the anus and anal canal in England.3 The low-incidence rate means that mass population screening is not indicated for anal cancer. However, if identifiable high-risk groups could be defined, then it may be effective and cost-effective to offer a targeted screening programme. Men who have sex with men (MSM) are a high-risk group, due to the increased risk of exposure to the HPV virus.4 Immunosuppression is also an observed risk factor, when combined with the elevated risk in MSM, an obvious high-risk sub-group is HIV positive MSM.4 The latter group may be at particular risk since the advent of highly active antiretroviral therapy (HAART) has reduced deaths from competing causes.5

Goldie et al.6,7 have published separate cost-effectiveness analyses of screening programmes for HIV positive and negative MSM. They found that screening either of these high-risk groups was a cost-effective strategy. However, issues remain around the applied modelling approach, such as the separate analysis of HIV positive and negative MSM, and the estimation of key parameter values. The previous models were also developed from a US perspective, and so the aim of this study is to assess an alternative modelling approach and to estimate the lifetime costs and consequences of screening for anal cancer in high-risk MSM groups from a UK NHS perspective.


   Methods
TOP
 Introduction
 Methods
Results
 Discussion
 Limitations of this study
 Conclusions
 Funding
 Appendix
 References
 
The cost-effectiveness of a screening programme is a function of well-rehearsed criteria for defining whether a condition is suitable for screening in a defined population: the importance of the health problem, the understanding of the natural history and epidemiology, and the safety, simplicity, accuracy and acceptability of the screening test to the general population.8 These issues were combined using a cohort-based Markov model to represent the natural history of anal cancer, alongside the natural history of HIV in a UK MSM population.

The model was populated using data derived from a review of the literature, and then calibrated probabilistically to represent the joint uncertainty in the input parameters. Alternative screening programmes are overlaid on the natural history model to predict their impact on health service costs and quality-adjusted life years (QALYs) gained over the lifetime of a cohort of MSM. A full probabilistic sensitivity analysis was undertaken, and costs and QALYs were discounted at 3.5% per annum. Cost estimates were uprated (where necessary) to 2007 values.

Model structure
The model structure, shown in Fig. 1, describes the simultaneous incidence and progression of HIV infection and anal cancer HIV negative MSM with no anal intra-epithelial neoplasia (AIN) may develop HIV infection with a CD4 count >500, which may then progress to a CD4 count of 200–500, and then to <200. AIDS was not explicitly modelled as relevant data sources expressed the linkage between HIV infection and mortality, and anal cancer, as functions of CD4 count. In any of the HIV states, individuals may be diagnosed and start to receive appropriate treatment that reduces their transition probabilities to subsequent CD4 count states and death, as well as to the anal cancer states. The choice of CD4 count categories was informed by data available to populate the model, as described in the model population section.


Figure 1
View larger version (21K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 1 Cost-effectiveness acceptability frontier for increased AIN regression sensitivity analysis.

 
At the same time as individuals may be progressing through the HIV states, they may also be progressing through the AIN states, which are the precursors to anal cancer. Individuals without AIN in the seven HIV-related states are assigned alternative probabilities of developing either low- or high-grade AIN. They may then progress through the high-grade AIN state to develop invasive anal cancer. Only a single anal cancer state is represented primarily due to data limitations that are discussed in Model Population section.

The model was developed as a cohort-Markov model using an annual cycle (chosen because an annual screening programme was the shortest interval tested). The time horizon of the model followed MSM to a maximum age of 100 years.

Model population
Systematic literature reviews were undertaken of the epidemiology and natural history of anal cancer, and screening technologies for anal cancer. The following sections briefly describe the data sources and methods of analysis informing the model's input parameters. Probability distributions were specified for each parameter, which are presented in Tables 1–3. Further details on the methods of analysis are presented in Appendix.


View this table:
[in this window]
[in a new window]

  		Table 1 Age-specific input parameters

 


View this table:
[in this window]
[in a new window]

  		Table 2 CD4 count-specific input parameters

 


View this table:
[in this window]
[in a new window]

  		Table 3 Other input parameters

 
MSM and HIV parameters
Age-specific prevalence rates for MSMs were informed by questions about whether male respondents had had sex with a man in the last year (upper estimate) and in the last three months (lower estimate) from the National Survey of Sexual Attitudes and Lifestyles (NATSAL) II.9 Age-specific data on the number of diagnosed HIV positive MSM were informed by data provided by the Survey Of Prevalent HIV Infections Diagnosed (SOPHID) team at the Health Protection Agency (HPA), with only minor adjustments required to assign those cases for which CD4 count was not reported. Only an aggregate estimate of the number of undiagnosed HIV-infected MSM in the UK was available, from which age- and stage-specific rates were modelled on the basis of relative likelihoods of remaining undiagnosed informed by clinical opinion.

Age-specific HIV infection incidence rates were fitted to estimated aggregate incidence rates,10 using a sub-Markov model that described transitions from the ‘CD4 count >500’ state to another CD4 state or death.11

Separate sets of HIV transition probabilities were required for diagnosed and undiagnosed HIV positive individuals. The former were informed by modelling using cross-sectional marker surveys,10,12 while Goldie et al.6 report transition data reflecting the impact of HAART.

Sweeting et al.10 use Bayesian back-calculation to estimate median estimates of the quarterly probabilities of HIV diagnosis by CD4 count category states for homosexual men. Age-specific clinical presentation rates were estimated by applying subjectively defined weights that increased the likelihood of older MSM presenting clinically.

Risks of AIDS-related mortality and other cause deaths in HIV positive MSM, by CD4 count, were informed by a separate modelling exercise that combined: the overall relative risk (RR) of death in HIV patients (4–7),13,14 the proportion of AIDS-related deaths in HIV patients post-HAART (0.07),14 the RR of developing AIDS by CD4 count (0.57 for a 100 point increase in CD4 counts), and the RR of developing a non-AIDS illness (0.84 for a 100 point increase in CD4 counts).15 RRs of death for non-diagnosed HIV patients used the aggregate RR for death observed in 1997 (17.5–27.5) in combination with the proportion of deaths due to AIDS in 1997 (0.84).13

Anal cancer natural history parameters
The natural history of anal cancer includes progression through the pre-invasive stages of the disease (low- and high-grade AIN), and from high-grade AIN to invasive anal cancer.

Transition probabilities between health states describing no AIN, low-grade AIN and high-grade AIN were informed by two US studies, published in the mid to late 1990s.16–18 UK studies were excluded on account of the absence of any histological follow-up.19

Alternative ranges describing progression from high-grade AIN to anal cancer were tested during the full model calibration process. Progression rates in HIV-negative and -positive MSM were assumed to be similar, though age-specific transition probabilities were specified as a proxy for time-dependent transition probabilities.

AIN and anal cancer treatment parameters
Treatment of low-grade AIN was not modelled because treatment effectiveness is not established. All detected high-grade AIN lesions were assumed to receive treatment: 20–30% of detected high-grade AIN was assumed to be treated non-surgically, with the remainder receiving radical surgery. HIV-positive patients were assumed to receive treatment for recurrence every 2 years, and HIV-negative patients every 10 years.20–22

The anal cancer model initially differentiated between early stage (primary) and late stage (metastatic) anal cancer. Data informing progression from early stage disease to late stage disease were of limited value as the absolute rate of progression was unclear and no information on the timing of progression was presented.23,24 As most studies presented survival rates after the treatment of early stage anal cancer, the model structure was adapted to describe only a transition to death from a single anal cancer state.

Study populations included few HIV-positive patients, and so survival post-anal cancer was estimated in two parts. First, non-anal cancer-related mortality rates were estimated separately for HIV-positive and -negative patients.13–15 Secondly, treated anal cancer-specific mortality was informed by the radiotherapy with chemotherapy arm of the largest and only UK study identified.25 An assumption of equal anal cancer mortality in HIV positive MSM was informed by a matched comparison of 10 HIV positive patients treated with HAART with 10 HIV negative patients.26 Untreated anal cancer mortality was subjectively informed by survival in the radiotherapy arm of the UKCCCR trial (which was defined as the upper 95% CI interval).25

Anal-screening test and clinical presentation parameters
Clinical presentation rates are not observable (at least, the denominator is not observable), and so subjective ranges were specified as inputs to the calibration analysis.

The review of studies investigating the characteristics of screening tests for AIN and anal cancer revealed significant variation in the estimated sensitivity and specificity rates,27–30 which is probably due to differences in the definition of a positive test and in the study populations. The studies included very few cases of high-grade AIN and so similar test characteristics were assumed for all grades. The model does differentiate between test characteristics in HIV-negative and -positive MSM, and assumes that an ‘Atypical Squamous Cells of Undetermined Significance’ (ASCUS) test result would be considered a positive test.

The second category of screening parameters describes the likelihood that individuals eligible for screening will attend for screening. Cranston et al.31 reported that ~10% of 125 MSM who were invited to take part in a screening study refused. However, as Knight32 suggests, the more limiting factor is likely to be the identification of MSM to whom screening can be offered. MSMs indicate, their preferred sources of safer sex support are predominantly friends, the gay press and sexual peers rather than formal HIV services.32 Findings from these studies were qualitatively considered to inform subjective parameter estimates.

Cost and utility weight parameters
Screening cost estimates were based on detailed costs estimated for the provision of cervical cancer screening in the UK.33 Costs of diagnostic follow-up (including false positive cases) and treatment of pre-cancerous AIN lesions were estimated by applying UK unit costs to clinical care algorithms developed by the University of California at San Francisco Faculty Practice.6 In the absence of unit costs for procedures specific to anal cancer, unit costs reported for similar procedures for the diagnosis and treatment of cervical cancer were applied.

Goldie et al.6 based treatment costs for anal cancer on published costs for colorectal cancer. A similar process was used in the current model, using a UK cost study.34 As a proxy for the cost effects of screen-detected cases presenting at an earlier stage than clinically presenting cases, lifetime costs associated with diagnosis at Dukes stage A (£8895) and Dukes stage B (£13 334) were assumed to represent the cost of screen detected and clinically presenting anal cancers, respectively.

Relevant UK costs for treating HIV infection by CD4 count were derived from the National Prospective Monitoring System ± HIV Health Economics Consortium (NPMS ± HHC) study.35

Utility weights for individuals with anal cancer were derived from a rating scale valuation undertaken by patients.6 The utility decrements were applied to both diagnosed and undiagnosed individuals (undiagnosed patients may have slightly higher utility than diagnosed patients, but later diagnosed patients may have lower utility during their treatment period). Goldie et al. used the same process to estimate a utility value for patients receiving treatment for high-grade AIN lesions of 0.98,6 which was specified as a utility multiplier. The utility effects of HIV infection were sourced from a number of studies that reported patient-based elicitation of utility values.36–38

Model calibration
The anal cancer screening model was calibrated to age-specific estimates of the incidence of anal cancer in 2003 as reported by the Cancer Registries.3 As anal cancer is a rare and relatively slow progressing cancer, it was assumed that the cancer registry estimates for anal cancer had 77% completeness, at the low end of observed completeness rates.39,40

Age-specific incidence rates in the MSM populations were fitted to sampled values for aggregate incidence, and RRs for anal cancer in HIV-positive5 and -negative men,41 as well as to age-specific RRs.42 This process was repeated for 5000 iterations in order to represent the joint uncertainty in these parameters.

For each of 5000 sampled sets of input parameter values, the calibration model collected the age-specific incident rates of anal cancer in HIV-negative and -positive MSMs in 2003. A two-stage calibration process was adopted. The first stage identified 4515 eligible input parameter sets that predicted aggregate incidence rates within the observed 95% confidence intervals (CIs). The second stage estimated the sum of differences between the predicted outputs and the observed 95% CIs for the incidence rates by age group and HIV status. The relative accuracy of the predicted outputs informed the assignment of sampling probabilities to each of the eligible input parameter sets.43

Model analysis
Ten programmes were analysed, included five programmes for HIV-positive MSM only, and five programmes for all MSM. Screening intervals of 1, 2, 3, 4 and 5 years were analysed for both eligible populations. The reference case analysis involved the random sampling of 5000 sets of input parameter sets based on the ranking probabilities estimated as part of the calibration process.


   Results
TOP
 Introduction
 Methods
 Results
Discussion
 Limitations of this study
 Conclusions
 Funding
 Appendix
 References
 
The mean results from the reference case analysis of the model are presented in Table 4, which show that screening MSM is actually more costly and less effective than providing no screening, i.e. no screening dominates screening MSM for anal cancer. The reduced number of QALYs gained is due to the utility decrements associated with false positive screening results, and particularly the adverse side effects from treatment of HG-AIN. The probabilistic analysis shows that in all iterations no screening has the highest net benefits and is 100% certain to be the most cost-effective option (given the model, inputs and assumptions) up to a QALY value of £50 000.


View this table:
[in this window]
[in a new window]

  		Table 4 Mean results from the reference case analysis

 
Sensitivity analyses were defined in the context of the reference case results that showed screening to have little chance of being cost-effective. The estimates of anal cancer incidence used to calibrate the model were not excessively uncertain. A more uncertain parameter was the rate of regression of low-grade AIN, for which Goldie et al. had estimated higher regression rates.6 The calibration and main models were re-run incorporating upwardly revised annual regression rates (30% in HIV-negative MSM, 2–5% in HIV-positive MSM). Table 5 presents the results that show that the minimum incremental cost per QALY (ICQ) for screening is £39 405. Screening all MSM is more effective than screening HIV-positive MSM due to gains from the early detection of undiagnosed HIV-positive MSM. The cost-effectiveness acceptability frontier in Fig. 2 shows that at higher QALY values the screening programmes for all MSM have higher mean net benefits (though no screening retains the highest probability of cost-effectiveness).


View this table:
[in this window]
[in a new window]

  		Table 5 Mean results from analyses incorporating increased regression rates from low-grade AIN

 


Figure 2
View larger version (7K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 2 Anal cancer screening model structure.

 
Other sensitivity analyses tested the assumption that false positive screen results and the treatment for HG-AIN incurred no utility decrements, which reduced the minimum ICQ to just over £60 000. Reducing the costs of screening by 50% did not alter the reference case results that all screening programmes are dominated by no screening.

Combining the effects of the three sensitivity analyses (all of which tested alternative assumptions that favoured screening), the minimum ICQ for screening is £20 996 (for 4-yearly screening of all MSM). Looking only at HIV positive MSM, the minimum ICQ is predicted to be £29 442 (for annual screening of HIV positive MSM).


   Discussion
TOP
 Introduction
 Methods
 Results
 Discussion
Limitations of this study
 Conclusions
 Funding
 Appendix
 References
 
Main finding of this study
The reference case cost-effectiveness model found that screening for anal cancer is very unlikely to be cost-effective. A key determinant of this finding was the low observed incidence of anal cancer in the UK population to which the model was calibrated. Estimated anal cancer incidence in the MSM population (by HIV status) was not conservative, incorporating upper ranges of the RRs of anal cancer in HIV-negative and -positive MSM compared with the general population of 45 and 272, respectively, and a low estimate of completeness in the observation of cases of anal cancer by the Cancer registries.

Sensitivity analyses showed that the reference case results were not overly robust to alternative assumptions around two parameters in particular: the probability of regression from low-grade AIN, and the utility effects of false positive cases and adverse effects of treatment of high-grade AIN. Other key areas of investigation include the modelled consequence that screening for anal cancer leads to the detection and the treatment of additional HIV positive MSM. The model assumes undiagnosed HIV-positive MSM are equally likely to be screened as HIV-negative MSM, though a specific mode of operation for a screening programme was not defined due to the unique issues around identifying and inviting different categories of MSM to a screening programme.

New analyses of existing primary data, undertaken specifically to inform regression rates may usefully update some key parameters at little additional cost. If these analyses increase the likelihood that anal cancer screening is cost-effective, further studies of the utility effects of treatment for high-grade AIN, and potential screening attendance rates may be justified.

What is already known on this topic
The only other identified cost-effectiveness analyses of screening for anal cancer found screening to be extremely cost-effective in both HIV-negative MSM (3-yearly) and HIV-positive MSM (annually), from a US perspective.6,7

What this study adds
Compared with previous models,6,7 the main structural development of the current study is the combined representation of HIV negative, undiagnosed HIV positive and diagnosed HIV-positive MSM in the same model, so allowing for the transition from HIV-negative to HIV-positive status. The model analysis also included a probabilistic sensitivity analysis and a process of recalibration to test the impact of specific input parameters (e.g. regression rates).

In the directly comparable analyses, the total cost differences in the US and UK analyses are similar, but the estimated QALY gain per HIV-positive MSM from an annual screening programme compared with no screening was 0.2 and 0.01, respectively. The US study appears to use similar incidence rates of anal cancer to calibrate the models, but the calibrated transition probabilities from high-grade AIN to cancer are significantly higher than those estimated in the current study (3.6–5%).

The regression rates assumed in the current model led to higher rates of high-grade AIN, with extremely low rates of transition to anal cancer. The model predicted that a high proportion of population prevalent high-grade AIN would either not progress to anal cancer in the lifetime of the patient or would not present clinically if it did progress (i.e. patients die of other causes). Thus, screening is predicted to detect many latent cases. Increased regression rates reduce the proportion of latent screen-detected cases.

In the current model, the high incidence of high-grade AIN leads to greater utility decrements due to the side effects of treatment (including treatment of recurrent lesions). The UK model also incorporates higher mortality effects for HIV infection, such that the potential for benefit from screening for anal cancer is reduced. Finally, the UK model assessed the cost-effectiveness of screening in the current population of MSM in the UK, not in a cohort of 30-year-old MSM, which reduces the scope for benefit as older MSM have fewer QALYs to gain.


   Limitations of this study
TOP
 Introduction
 Methods
 Results
 Discussion
 Limitations of this study
Conclusions
 Funding
 Appendix
 References
 
The anal cancer screening model was complicated, jointly modelling HIV and AIN progression. More flexibility to represent this complexity would have been afforded by an individual sampling model (ISM), but previous experience in the development of complex screening models with limited data informed our efforts to apply the model as a cohort-based model.43,44 To apply the probabilistic calibration methods described in this report, ISMs require model running times several orders of magnitude higher than cohort models.

The main limiting assumptions of the cohort model include the use of non-time dependent transition probabilities, which meant that proxies for time dependency were incorporated, for example, transition probabilities were assumed to increase with age. The model also adopted a simplistic representation of the post-diagnosis phase of the disease, though this was partly due to data inadequacies. The model analysis indicates that these structural limitations had only a minor impact on the cost-effectiveness results relative to the impact of uncertainties around key input parameter values.

Evidence from the Seattle cohort study showed an increased RR of high-grade AIN for men engaging in receptive anal intercourse more than three times per month (mean 2.2, 95% CI 0.9–5.5).16 This indicates that screening is likely to be more cost-effective if an eligible population of high-frequency MSM could de defined. Such information might usefully inform an awareness campaign for an anal cancer screening programme, but restricting screening eligibility on the basis of frequency of intercourse was assumed to be neither practical nor appropriate.

A screening programme might offer the opportunity for additional benefits that were not represented, such as education regarding safer sexual practices that could be promoted during screening attendance. Screening might also impact on disease progression more indirectly, for example, transmission of HPV may be reduced following the detection of AIN as a result of modified sexual activity. The model did not consider these potential benefits because there is no evidence of effect in the reviewed literature, and the link between sexual activity and anal cancer and HIV risk would add significantly to the complexity of the model.


   Conclusions
TOP
 Introduction
 Methods
 Results
 Discussion
 Limitations of this study
 Conclusions
Funding
 Appendix
 References
 
The current evidence base does not provide sufficient evidence to justify a screening programme for anal cancer, even for identifiable high-risk groups. Analyses of existing primary data to specifically inform the anal cancer screening model might reduce the uncertainty to a significant degree.


   Funding
TOP
 Introduction
 Methods
 Results
 Discussion
 Limitations of this study
 Conclusions
 Funding
Appendix
 References
 
This project was funded by a project grant from the UK National Institute for Health Research, Health Technology Assessment programme.


   Appendix
TOP
 Introduction
 Methods
 Results
 Discussion
 Limitations of this study
 Conclusions
 Funding
 Appendix
References
 
The Appendix provides additional details on the estimation of the screening model input parameters.

MSM and HIV parameters
Prevalence rates for MSMs were informed by the National Survey of Sexual Attitudes and Lifestyles (NATSAL) II.8 Rates in 46–55, 56–65, 66–75 and 76+ age ranges were not included in the NATSAL and so were assumed to be 95, 75, 5 and 0% of the rates estimated for the 35–44 age range.

To allocate the estimated aggregate number of undiagnosed HIV-positive MSM by age and stage, relative likelihoods of remaining undiagnosed were assigned to the individuals of different ages based on clinical opinion. Individuals in the youngest age groups (15–25 and 26–35 year olds) were assigned a relative value of 1, 36–45 and 46–55 year olds a value of 0.5 and 56–65 year olds 0.25. For 5000 iterations, an aggregate undiagnosed HIV-positive MSM population and a set of age-specific diagnosed HIV-positive MSM populations were sampled from relevant Poisson distributions. For each iteration, the non-adjusted numbers of undiagnosed HIV-positive MSM (aggregate undiagnosed cases multiplied by the proportion of diagnosed cases) in each age group were multiplied by the relative likelihood weights, which were then adjusted proportionately so that the aggregate number of undiagnosed individuals equalled the sampled value. The stage distribution was estimated by applying clinical opinion-based relative likelihoods that individuals in each of the three CD4 stages would remain undiagnosed (<200 CD4–0.05; 200–500 CD4–0.1; >500 CD4–1), and then adjusting to equal the estimated aggregate number of undiagnosed HIV cases. Finally, ranges for the prevalence rates were estimated by dividing the lower and upper estimates of the numbers by the upper and lower estimates of the relevant age-specific numbers of MSM in the general population, respectively.

Age-specific HIV population incidence rates by CD4 count were estimated using a sub-Markov model, which assumed:

  • All individuals developing an HIV infection start in the ‘>500’ CD4 count category.
  • All individuals who developed HIV prior to 1990 died in 2005.
  • Age-specific incidence rates of HIV have remained constant since 1990.
The estimation process involves the following stages:

Individuals may leave the ‘>500’ state due to transition to another CD4 state (using the main model transition probabilities) or death (informed by UK male life tables as a CD4 count >500 has been shown to have little effect on mortality11). Transition probabilities were not assumed to be affected by treatment so the model does not distinguish between diagnosed and undiagnosed individuals.

The model estimates the numbers of individuals remaining in the ‘ > 500’ state at each age in 2005 as the sum of the product of the incidence rate at each age and the probability that patients remain in the ‘>500’ state in 2005. This informed the absolute numbers of individuals in the ‘>500’ state in 2005 in each of seven age categories (15–25, 26–35, 36–45, 46–55, 56–65, 66–75 and >75 years). A simulation exercise was then undertaken to identify the combinations of input parameter values that best predicted the observed prevalence of diagnosed and undiagnosed HIV-infected MSM with a CD4 count of >500.

The number of MSM at risk of HIV infection within each age range was estimated as the prevalence estimate of all MSM (in each age range) minus prevalent HIV infected MSMs at the start of each age range and half of all incident cases in the age range being analysed. The 10-year rates were converted to annual rates using the established equation.45 The estimates for age groups >66 years were unstable due to small numbers, as the incident numbers were so small for these age groups, incidence rates of zero were assumed.

Risks of AIDS-related mortality and other cause deaths for HIV-positive MSM combined the RRs for death compared with patients with a CD4 count of <200, and the specified distribution of CD4 counts, the Solver function in Excel was used to fit the RR of death in patients with a CD4 count of <200 to a sampled value for the overall RR of death for HIV patients. This process was repeated 1000 times to obtain a joint distribution of the RRs for the three CD4 count categories.

Anal cancer natural history parameters
Two sets of analyses using published data from the two main cohort studies were undertaken.16–18 The complex analysis used literature-based sources to control for false negative and false positive results in cytology-alone detected cases, the unobserved development of low-grade disease prior to high-grade disease, the timing of events and incomplete follow-up. These analyses were found to be sensitive to the assumptions around follow-up times and the proportions of cases receiving a histological diagnosis. As a result, the calibration inputs were informed by simpler analyses that did not adjust for test sensitivity and specificity, and estimated person years at risk assuming individuals experiencing an AIN event had half the follow-up of those not experiencing an event. It was also assumed that half of the observed cases of high-grade AIN (in individuals with no AIN at baseline) experienced low-grade AIN prior to high-grade AIN.

In addition to increased progression in HIV-positive MSM, the San Francisco study showed that CD4 count is a significant predictor of AIN progression (P-test for trend, <0.0005),17 and the presented RRs were used to adjust AIN-transition probabilities by CD4 count.

Model calibration
The calibration model commences in 1990. The probability of having high-grade AIN in 1990 was estimated using a simple sub-model that estimated the progression of HIV negative MSM from normal to high-grade AIN, assuming a mean age of becoming a MSM of 26 years.

The following stages were involved in the assignment of sampling probabilities to each of the eligible input parameter sets:

  1. The sum of differences between each age- and HIV status-specific incidence estimate and the bounds of the respective observed 95% confidence intervals were calculated for each iteration.
  2. The proportions of the sum of differences contributed by each input parameter set were estimated.
  3. The reciprocals of the proportions of the sum of differences contributed by each iteration were estimated (no iterations had a sum of differences of zero).
  4. The reciprocals were divided by the sum of the reciprocals to estimate the probability that each of the 4515 parameter sets was the most relevant set.


   References
TOP
 Introduction
 Methods
 Results
 Discussion
 Limitations of this study
 Conclusions
 Funding
 Appendix
 References
 

  1. Spence AR, Franco EL, Ferenczy A. The role of human papillomaviruses in cancer: evidence to date. Am J Cancer (2005) 4(1):49–64.[CrossRef]
  2. Melbye M, Sprogel P. Aetiological parallel between anal cancer and cervical cancer. Lancet (1991) 338(8768):657–9.[CrossRef][Web of Science][Medline]
  3. Office for National Statistics. Cancer Statistics Registrations (2005).
  4. Zeller JL, Lynm C, Glass RM, Anal cancer. JAMA (2008) 299(16):1980.[Free Full Text]
  5. Bower M, Powles T, Newsom-Davis T, et al. HIV-associated anal cancer: has highly active antiretroviral therapy reduced the incidence or improved the outcome? J Acquir Immune Defic Syndr (2004) 37(5):1563–5.[Web of Science][Medline]
  6. Goldie SJ, Kuntz KM, Weinstein MC, et al. Cost-effectiveness of screening for anal squamous intraepithelial lesions and anal cancer in human immunodeficiency virus-negative homosexual and bisexual men. Am J Med (2000) 108(8):634–41.[CrossRef][Web of Science][Medline]
  7. Goldie SJ, Kuntz KM, Weinstein MC, et al. The clinical effectiveness and cost-effectiveness of screening for anal squamous intraepithelial lesions in homosexual and bisexual HIV-positive men. J Am Med Assoc (1999) 281(19):1822–9.[Abstract/Free Full Text]
  8. UK National Screening Committee. (2006) Criteria for Appraising the Viability, Effectiveness and Appropriateness of A Screening Programme. London.
  9. Erens B, McManus S, Field J, et al. National Survey of Sexual Attitudes and Lifestyles II: Technical Report (2001) London, UK: Natcen.
  10. Sweeting MJ, De Angelis D, Aalen OO. Bayesian back-calculation using a multi-state model with application to HIV. Stat Med (2005) 24:3991–4007.[CrossRef][Web of Science][Medline]
  11. Lewden C. Responders to antiretroviral treatment over 500 CD4/mm3 reach same mortality rates as general population: APROCO and Aquitaine Cohorts. (2005) 10th European AIDS Conference: France. abstract PE18.4/8.
  12. Satten GA, Longini IM. Estimation of incidence of HIV infection using cross-sectional marker surveys. Biometrics (1994) 50:675–88.[CrossRef][Web of Science][Medline]
  13. Ormaasen V, Sandvik L, Dudman SG, et al. HIV related and non-HIV related mortality before and after the introduction of highly active antiretroviral therapy (HAART) in Norway compared to the general population. Scand J Infect Dis (2007) 39(1):51–7.[CrossRef][Web of Science][Medline]
  14. Martinez E, Milinkovic A, Buira E. Incidence and causes of death in HIV-infected persons receiving highly active antiretroviral therapy compared with estimates for the general population of similar age and from the same geographical area. HIV Med (2007) 8(40):251–8.[CrossRef][Web of Science][Medline]
  15. Baker J, Peng G, Rapkin J, et al. HIV-related immune suppression after ART predicts risk of non-opportunistic diseases: results from the FIRST study. (2007) 14th Conference on Retroviruses and Opportunistic Infections, February 26: Los Angeles. Abstract 37.
  16. Critchlow CW, Surawicz CM, Holmes KK, et al. Prospective study of high grade anal squamous intraepithelial neoplasia in a cohort of homosexual men: influence of HIV infection, immunosuppression and human papillomavirus infection. AIDS (1995) 9(11):1255–62.[Web of Science][Medline]
  17. Palefsky JM, Holly EA, Hogeboom CJ, et al. Virologic, immunologic, and clinical parameters in the incidence and progression of anal squamous intraepithelial lesions in HIV-positive and HIV-negative homosexual men. J Acquir Immune Defic Syndr Hum Retrovirol (1998) 17(4):314–9.[Web of Science][Medline]
  18. Palefsky JM, Holly EA, Ralston ML, et al. High incidence of anal high-grade squamous intra-epithelial lesions among HIV-positive and HIV-negative homosexual and bisexual men. AIDS (1998) 12(5):495–503.[CrossRef][Web of Science][Medline]
  19. Sayers SJ, McMillan A, McGoogan E. Anal cytological abnormalities in HIV-infected homosexual men. Int J STD AIDS (1998) 9(1):37–40.[Abstract/Free Full Text]
  20. Marchesa P, Fazio VW, Oliart S, et al. Perianal Bowen's disease: a clinicopathologic study of 47 patients. Dis Colon Rectum (1997) 40:1286–93.[CrossRef][Web of Science][Medline]
  21. Chang GJ, Berry JM, Jay N, et al. Surgical treatment of high-grade anal squamous intraepithelial lesions: a prospective study. Dis Colon Rectum (2002) 45(4):453–8.[CrossRef][Web of Science][Medline]
  22. Brown SR, Skinner P, Tidy J, et al. Outcome after surgical resection for high-grade anal intraepithelial neoplasia (Bowen's disease). Br J Surg (1999) 86(8):1063–6.[CrossRef][Web of Science][Medline]
  23. Doniec JM, Schniewind B, Kovacs G, et al. Multimodal therapy of anal cancer added by new endosonographic-guided brachytherapy. Surg Endosc (2006) 20(4):673–8.[CrossRef][Web of Science][Medline]
  24. Wong S, Gibbs P, Chao M, et al. Carcinoma of the anal canal: a local experience and review of the literature. ANZ J Surg (2004) 74(7):541–6.[CrossRef][Web of Science][Medline]
  25. Arnott SJ, Cunningham D, Gallagher J, et al. Epidermoid anal cancer: results from the UKCCCR randomised trial of radiotherapy alone versus radiotherapy, 5-fluorouracil, and mitomycin. Lancet (1996) 348(9034):1049–54.[CrossRef][Web of Science][Medline]
  26. Oehler-Janne C, Seifert B, Lutolf UM, et al. Local tumor control and toxicity in HIV-associated anal carcinoma treated with radiotherapy in the era of antiretroviral therapy. Radiat Oncol (2006) 1:29.[CrossRef][Medline]
  27. Palefsky JM, Holly EA, Hogeboom CJ, et al. Anal cytology as a screening tool for anal squamous intraepithelial lesions. J Acquir Immune Defic Syndr Hum Retrovirol (1997) 14(5):415–22.[Web of Science][Medline]
  28. Cranston RD, Darragh TM, Holly EA, et al. Self-collected versus clinician-collected anal cytology specimens to diagnose anal intraepithelial neoplasia in HIV-positive men. J Acquir Immune Defic Syndr J AIDS (2004) 36(4):915–20.[CrossRef]
  29. Arain S, Walts AE, Thomas P, et al. The anal pap smear: cytomorphology of squamous intrapithelial lesions. Cytojournal (2005) 2(1):4.[CrossRef][Medline]
  30. Chiao EY, Giordano TP, Palefsky JM, et al. Screening HIV-infected individuals for anal cancer precursor lesions: a systematic review. Clin Infect Dis (2006) 43:223–33.[CrossRef][Web of Science][Medline]
  31. Knight D. Health care screening for men who have sex with men. Am Fam Phys (2004) 69(9):2149–56.[Web of Science][Medline]
  32. Scott P, Coxon T, Kirk M, et al. Cottaging and Cruising in Barnet, Brent and Harrow: Final Project Report 2007. http://www.racoon.dircon.co.uk/bbh/cot.htm (2 July 2007, date last accessed).
  33. Moss SM, Gray A, Legood R, et al. Evaluation of HPV/LBC Cervical Screening Pilot Studies, First report to the Department of Health on evaluation of LBC. (2002) December.
  34. Tappenden P, Chilcott J, Eggington S, et al. Option appraisal of population-based colorectal cancer screening programmes in England. Gut (2007) 56:677–84.[Abstract/Free Full Text]
  35. Miners AH, Sabin CA, Trueman P, et al. Assessing the cost-effectiveness of HAART for adults with HIV in England. HIV Med (2001) 2(1):52–8.[CrossRef][Medline]
  36. Mauskopf J, Lacey L, Kempel A, et al. The cost-effectiveness of treatment with lamivudine and zidovudine compared with zidovudine alone: a comparison of Markov model and trial data estimates. Am J Manag Care (1998) 4(7):1004–12.[Web of Science][Medline]
  37. Trueman P, Youle M, Sabin CA, et al. The cost-effectiveness of triple nucleoside analogue therapy antiretroviral regimens in the treatment of HIV in the United Kingdom. HIV Clin Trials (2000) 1(1):27–35.[CrossRef][Medline]
  38. Goldie SJ, Freedberg KA, Weinstein MC, et al. Cost effectiveness of human papillomavirus testing to augment cervical cancer screening in women infected with the human immunodeficiency virus. Am J Med Errors (2001) 111(2):140–9.[CrossRef]
  39. Forman D. Effective Registration of Skin Cancer in Yorkshire. The Research Findings Register Summary number 682. http://www.ReFeR.nhs.uk/ViewRecord.asp?ID=682 (22 June 2007, date last accessed).
  40. Silcocks PB, Waterhouse A. How good are our data? Initial studies of completeness. (2007) www.trentcancer.nhs.uk/publications/presentations/02iacrpbs1.html.
  41. Lampinen TM, Miller ML, Chan K, et al. Randomized clinical evaluation of self-screening for anal cancer precursors in men who have sex with men. Cytojournal (2006) 3:4.[CrossRef][Medline]
  42. Frisch M, Biggar RJ, Goedert JJ. Human papillomavirus-associated cancers in patients with human immunodeficiency virus infection and acquired immunodeficiency syndrome. J Nat Cancer Inst (2000) 92(18):1500–10.[Abstract/Free Full Text]
  43. Karnon J, Czoski-Murray C, Smith K, et al. A preliminary model-based assessment of the cost-utility of a screening programme for early age-related macular degeneration, Health Technol Assess. (2008) 12(27).
  44. Carlton J, Karnon J, Czoski-Murray C, et al. The clinical effectiveness and cost-effectiveness of screening programmes for amblyopia and strabismus in children up to the age of 4–5 years: a systematic review and economic evaluation. Health Technol Assess (2008) 12(25).
  45. Miller DK, Homan SM. Determining transition probabilities: confusion and suggestions. Med Decis Mak (1994) 14(1):52–8.[CrossRef]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
30/3/293    most recent
fdn045v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Karnon, J.
Right arrow Articles by Smith, K. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Karnon, J.
Right arrow Articles by Smith, K. J.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?