Mathematical modelling of cost-effectiveness of chlamydia testing highlights complexities and uncertainties that should not be overlooked. Sarah C Woodhall, Koh Jun Ong, John Saunders, Kevin Dunbar
Sarah Woodhall, Public Health England
23 February 2015
We welcome the recent article by Looker et al. exploring the impact and cost-effectiveness of chlamydiascreening in Scotland given current state of knowledge about the input parameters to cost-effectiveness models[1]. Their work clearly shows some of the inadequacies in current knowledge that need to be addressed before models of cost-effectiveness of chlamydia screening can be useful tools for policy decision making.
Looker et al. highlight some of the uncertainties which hinder estimating cost-effectiveness. We agree with the authors that improved parameter estimates, especially with regard to management costs and loss of quality-adjusted life years (QALY) associated with chlamydia-related outcomes, would greatly strengthen future models. While management costs for an episode of PID in England have recently been estimated from the POPI trial[2], management costs for other chlamydia-related sequelae are still uncertain, as these costs are dependent on the care pathway provided for the sequelae. With TFI in particular, since in-vitro fertilisation (IVF) is among the major treatment options, the total number ofIVFs per case of TFI may have increased, since the more recent NICE recommendation advocated for 3 cycles of IVF in women under the age of 40 who failed to conceive after 2 years[3]. Cost estimates are also further complicated by the fact that milder tubal blockage may be managed by tubal surgery, and the proportion of women with TFI being treated with this may vary depending on local practice and policy.
In addition, estimates of QALY loss associated with the sequelae of chlamydia vary greatly in the literature, especially for sequelae with longer duration of disutility, such as TFI. For example, currently three utility estimates are available for TFI, measured using health utilities index (HUI) or time-trade-off methods, giving the values of 0.76 to 0.84[4]. However, individual models then apply different duration of disutility, which can range from 1 year[1] to rest of lifetime until the age of 50[5]. Hence, we have two layers of uncertainty with regards to QALY estimates here, the first being that the standard recommended generic utility elicitation tool is the EQ5D[6] and we currently lack estimates elicited using this tool. The seconduncertainty is the duration of disutility experienced by patients. Public Health England plan to conduct data collection to improve these estimates – as a prerequisite for more accurate cost-effectiveness analyses to guide health policies.
The authors rightly point out that their estimate is based on the model assumptions. The influence of several of these assumptions is worth addressing to explore the potential influence on the cost-effectiveness to inform public health policy.
Firstly, the base case of the model assumes that PID and TFI are only prevented through interrupting transmission of chlamydia and not through treating someone who is diagnosed with chlamydia. In sensitivity analysis, the authors demonstrate a tendency towards improved cost-effectiveness estimates iftreatment is assumed to reduce risk of complications. While the natural history of infection with chlamydia is not precisely understood[7;8], there is evidence to suggest that some risk reduction would be expected from treatment of infections detected through screening. The largest prospective randomisedcontrolled trial demonstrated that treating prevalent infections reduced the risk of developing PID at 12 months (relative risk 0.17, 0.03-1.01)[9] and existing models are consistent with the theory that damage occurs throughout the duration of infection[7;10]. Therefore the base-case is unduly pessimistic on the benefits of treatment of prevalent infections.
Secondly, the model was not stratified by gender, meaning that the same prevalence, testing coverage and test positivity was assumed for women and men according to their sexual activity class. However, it is well documented that there exist significant differences in testing coverage, prevalence and positivity between men and women both in Scotland and England[11]. Women are much more likely to have been tested for chlamydia and given that the above sensitivity analysis demonstrated a tendency towards improved cost-effectiveness, higher testing/diagnoses rate in women would result in better returns in terms of female reproductive health damage prevented, which are the main outcomes costed in their analysis (see below). We appreciate the pragmatic need to maintain a parsimonious approach. However we feel that it should be recognised that a model with gender stratification may have resulted in a different conclusion regarding cost-effectiveness of screening.
Finally, chlamydia is not only associated with the adverse sequelae of PID and TFI but also ectopic pregnancy, neonatal infections and epididymitis in men. The cost-effectiveness of screening should include these potential outcomes within the model. Welte et al[12] for example, included the following chlamydia-related sequelae (their individual contribution by proportion to the overall chlamydia sequelaemanagement cost in brackets): symptomatic PID (31%), EP (19.4%), TFI (3.8%), chronic pelvic pain (13.3%), neonatal diseases (2.8%), epididymitis (6.3%), urethritis (18.8%) and cervicitis (4.6%). In their model, the probability of EP was 8% among those with PID, and the probability of PID was 25%, of which 40% were assumed to be symptomatic. It cost ~£1,400 (GBP 2013/14 values) to managesymptomatic PID, ~£1,150 to manage TFI and £3,000 to manage EP in their model. The findings presented by Looker et al highlight the combination of uncertainty with regards to management cost, which includes the unit cost of a single disease management, and the probability of disease occurrence applied to the model. It also demonstrates the importance of other chlamydia sequelae not considered by[1], since symptomatic PID and TFI aside, the remainder 65% of the costs come from other chlamydia-related sequelae. Uncertainty in these values may be a rationale for excluding these outcomes from the model. However, their exclusion almost certainly results in underestimating the cost-effectiveness of chlamydia screening.
Looker et al’s analyses highlights some key features that influence cost-effectiveness of chlamydia screening and testing strategies, but does not provide a complete measure of cost-effectiveness of chlamydia testing as practised in Scotland.
References (1) Looker KJ, Wallace LA, Turner KM. Impact and cost-effectiveness of chlamydia testing in Scotland:a mathematical modelling study. Theor Biol Med Model 2015 Jan 15;12(1):2.
(2) Aghaizu A, Adams EJ, Turner K, et al. What is the cost of pelvic inflammatory disease and howmuch could be prevented by screening for chlamydia trachomatis? Cost analysis of the Prevention of Pelvic Infection (POPI) trial. Sex Transm Infect 2011 Jun;87(4):312-7.
(3) National Institute for Health and Clinical Excellence. Fertility: assessment and treatment forpeople with fertility problems (update). Costing report. Implementing NICE guidance. 2013.
(4) Jackson LJ, Auguste P, Low N, Roberts TE. Valuing the Health States Associated with Chlamydiatrachomatis Infections and Their Sequelae: A Systematic Review of Economic Evaluations and Primary Studies. Value Health 2014 Jan;17(1):116-30.
(5) de VR, van Bergen JE, de Jong-van den Berg LT, Postma MJ. Systematic screening for Chlamydiatrachomatis: estimating cost-effectiveness using dynamic modeling and Dutch data. Value Health 2006 Jan;9(1):1-11.
(6) National Institute for Health and Care Excellence. Guide to the methods of technology appraisal2013. 2013.
(7) Price MJ, Ades AE, De AD, et al. Risk of pelvic inflammatory disease following Chlamydiatrachomatis infection: analysis of prospective studies with a multistate model. Am J Epidemiol 2013 Aug 1;178(3):484-92.
(8) Haggerty CL, Gottlieb SL, Taylor BD, Low N, Xu F, Ness RB. Risk of sequelae after Chlamydiatrachomatis genital infection in women. J Infect Dis 2010 Jun 15;201 Suppl 2:S134-S155.
(9) Oakeshott P, Kerry S, Aghaizu A, et al. Randomised controlled trial of screening for Chlamydiatrachomatis to prevent pelvic inflammatory disease: the POPI (prevention of pelvic infection) trial. BMJ 2010;340:c1642.
(10) Herzog SA, Althaus CL, Heijne JC, et al. Timing of progression from Chlamydia trachomatisinfection to pelvic inflammatory disease: a mathematical modelling study. BMC Infect Dis 2012;12:187.
(11) Sonnenberg P, Clifton S, Beddows S, et al. Prevalence, risk factors, and uptake ofinterventions for sexually transmitted infections in Britain: findings from the National Surveys of Sexual Attitudes and Lifestyles (Natsal). Lancet 2013 Nov 30;382(9907):1795-806.
(12) Welte R, Kretzschmar M, Leidl R, Van den Hoek A, Jager J, Postma M. Cost effectiveness of screening programmes for Chlamydia trachomatis. Sexually Transmitted Diseases 2000;27(9):518.
Competing interests
All authors are employed by Public Health England. Kevin Dunbar is the Director of the National Chlamdydia Screening Programme (NCSP) in England. John Saunders is the Clinical Champion of the NCSP.
Theoretical Biology and Medical Modelling
Mathematical modelling of cost-effectiveness of chlamydia testing highlights complexities and uncertainties that should not be overlooked. Sarah C Woodhall, Koh Jun Ong, John Saunders, Kevin Dunbar
23 February 2015
We welcome the recent article by Looker et al. exploring the impact and cost-effectiveness of chlamydiascreening in Scotland given current state of knowledge about the input parameters to cost-effectiveness models[1]. Their work clearly shows some of the inadequacies in current knowledge that need to be addressed before models of cost-effectiveness of chlamydia screening can be useful tools for policy decision making.
Looker et al. highlight some of the uncertainties which hinder estimating cost-effectiveness. We agree with the authors that improved parameter estimates, especially with regard to management costs and loss of quality-adjusted life years (QALY) associated with chlamydia-related outcomes, would greatly strengthen future models. While management costs for an episode of PID in England have recently been estimated from the POPI trial[2], management costs for other chlamydia-related sequelae are still uncertain, as these costs are dependent on the care pathway provided for the sequelae. With TFI in particular, since in-vitro fertilisation (IVF) is among the major treatment options, the total number ofIVFs per case of TFI may have increased, since the more recent NICE recommendation advocated for 3 cycles of IVF in women under the age of 40 who failed to conceive after 2 years[3]. Cost estimates are also further complicated by the fact that milder tubal blockage may be managed by tubal surgery, and the proportion of women with TFI being treated with this may vary depending on local practice and policy.
In addition, estimates of QALY loss associated with the sequelae of chlamydia vary greatly in the literature, especially for sequelae with longer duration of disutility, such as TFI. For example, currently three utility estimates are available for TFI, measured using health utilities index (HUI) or time-trade-off methods, giving the values of 0.76 to 0.84[4]. However, individual models then apply different duration of disutility, which can range from 1 year[1] to rest of lifetime until the age of 50[5]. Hence, we have two layers of uncertainty with regards to QALY estimates here, the first being that the standard recommended generic utility elicitation tool is the EQ5D[6] and we currently lack estimates elicited using this tool. The seconduncertainty is the duration of disutility experienced by patients. Public Health England plan to conduct data collection to improve these estimates – as a prerequisite for more accurate cost-effectiveness analyses to guide health policies.
The authors rightly point out that their estimate is based on the model assumptions. The influence of several of these assumptions is worth addressing to explore the potential influence on the cost-effectiveness to inform public health policy.
Firstly, the base case of the model assumes that PID and TFI are only prevented through interrupting transmission of chlamydia and not through treating someone who is diagnosed with chlamydia. In sensitivity analysis, the authors demonstrate a tendency towards improved cost-effectiveness estimates iftreatment is assumed to reduce risk of complications. While the natural history of infection with chlamydia is not precisely understood[7;8], there is evidence to suggest that some risk reduction would be expected from treatment of infections detected through screening. The largest prospective randomisedcontrolled trial demonstrated that treating prevalent infections reduced the risk of developing PID at 12 months (relative risk 0.17, 0.03-1.01)[9] and existing models are consistent with the theory that damage occurs throughout the duration of infection[7;10]. Therefore the base-case is unduly pessimistic on the benefits of treatment of prevalent infections.
Secondly, the model was not stratified by gender, meaning that the same prevalence, testing coverage and test positivity was assumed for women and men according to their sexual activity class. However, it is well documented that there exist significant differences in testing coverage, prevalence and positivity between men and women both in Scotland and England[11]. Women are much more likely to have been tested for chlamydia and given that the above sensitivity analysis demonstrated a tendency towards improved cost-effectiveness, higher testing/diagnoses rate in women would result in better returns in terms of female reproductive health damage prevented, which are the main outcomes costed in their analysis (see below). We appreciate the pragmatic need to maintain a parsimonious approach. However we feel that it should be recognised that a model with gender stratification may have resulted in a different conclusion regarding cost-effectiveness of screening.
Finally, chlamydia is not only associated with the adverse sequelae of PID and TFI but also ectopic pregnancy, neonatal infections and epididymitis in men. The cost-effectiveness of screening should include these potential outcomes within the model. Welte et al[12] for example, included the following chlamydia-related sequelae (their individual contribution by proportion to the overall chlamydia sequelaemanagement cost in brackets): symptomatic PID (31%), EP (19.4%), TFI (3.8%), chronic pelvic pain (13.3%), neonatal diseases (2.8%), epididymitis (6.3%), urethritis (18.8%) and cervicitis (4.6%). In their model, the probability of EP was 8% among those with PID, and the probability of PID was 25%, of which 40% were assumed to be symptomatic. It cost ~£1,400 (GBP 2013/14 values) to managesymptomatic PID, ~£1,150 to manage TFI and £3,000 to manage EP in their model. The findings presented by Looker et al highlight the combination of uncertainty with regards to management cost, which includes the unit cost of a single disease management, and the probability of disease occurrence applied to the model. It also demonstrates the importance of other chlamydia sequelae not considered by[1], since symptomatic PID and TFI aside, the remainder 65% of the costs come from other chlamydia-related sequelae. Uncertainty in these values may be a rationale for excluding these outcomes from the model. However, their exclusion almost certainly results in underestimating the cost-effectiveness of chlamydia screening.
Looker et al’s analyses highlights some key features that influence cost-effectiveness of chlamydia screening and testing strategies, but does not provide a complete measure of cost-effectiveness of chlamydia testing as practised in Scotland.
References
(1) Looker KJ, Wallace LA, Turner KM. Impact and cost-effectiveness of chlamydia testing in Scotland:a mathematical modelling study. Theor Biol Med Model 2015 Jan 15;12(1):2.
(2) Aghaizu A, Adams EJ, Turner K, et al. What is the cost of pelvic inflammatory disease and howmuch could be prevented by screening for chlamydia trachomatis? Cost analysis of the Prevention of Pelvic Infection (POPI) trial. Sex Transm Infect 2011 Jun;87(4):312-7.
(3) National Institute for Health and Clinical Excellence. Fertility: assessment and treatment forpeople with fertility problems (update). Costing report. Implementing NICE guidance. 2013.
(4) Jackson LJ, Auguste P, Low N, Roberts TE. Valuing the Health States Associated with Chlamydiatrachomatis Infections and Their Sequelae: A Systematic Review of Economic Evaluations and Primary Studies. Value Health 2014 Jan;17(1):116-30.
(5) de VR, van Bergen JE, de Jong-van den Berg LT, Postma MJ. Systematic screening for Chlamydiatrachomatis: estimating cost-effectiveness using dynamic modeling and Dutch data. Value Health 2006 Jan;9(1):1-11.
(6) National Institute for Health and Care Excellence. Guide to the methods of technology appraisal2013. 2013.
(7) Price MJ, Ades AE, De AD, et al. Risk of pelvic inflammatory disease following Chlamydiatrachomatis infection: analysis of prospective studies with a multistate model. Am J Epidemiol 2013 Aug 1;178(3):484-92.
(8) Haggerty CL, Gottlieb SL, Taylor BD, Low N, Xu F, Ness RB. Risk of sequelae after Chlamydiatrachomatis genital infection in women. J Infect Dis 2010 Jun 15;201 Suppl 2:S134-S155.
(9) Oakeshott P, Kerry S, Aghaizu A, et al. Randomised controlled trial of screening for Chlamydiatrachomatis to prevent pelvic inflammatory disease: the POPI (prevention of pelvic infection) trial. BMJ 2010;340:c1642.
(10) Herzog SA, Althaus CL, Heijne JC, et al. Timing of progression from Chlamydia trachomatisinfection to pelvic inflammatory disease: a mathematical modelling study. BMC Infect Dis 2012;12:187.
(11) Sonnenberg P, Clifton S, Beddows S, et al. Prevalence, risk factors, and uptake ofinterventions for sexually transmitted infections in Britain: findings from the National Surveys of Sexual Attitudes and Lifestyles (Natsal). Lancet 2013 Nov 30;382(9907):1795-806.
(12) Welte R, Kretzschmar M, Leidl R, Van den Hoek A, Jager J, Postma M. Cost effectiveness of screening programmes for Chlamydia trachomatis. Sexually Transmitted Diseases 2000;27(9):518.
Competing interests
All authors are employed by Public Health England. Kevin Dunbar is the Director of the National Chlamdydia Screening Programme (NCSP) in England. John Saunders is the Clinical Champion of the NCSP.