Article Text
Abstract
Objectives Herpes simplex virus (HSV) infections in newborns are an uncommon but potentially devastating consequence of genital HSV infection in women. Current practice focuses on preventing perinatal transmission by women with prevalent HSV, but transmission risk is greatest when genital HSV is acquired for the first time late in pregnancy. The objective of this study was to assess the effectiveness and cost effectiveness of identifying pregnant women at risk of de novo HSV acquisition as a means of preventing vertical HSV transmission.
Methods A Bayesian decision tree model was parameterised using the best available health and economic data relating to HSV in pregnancy and was used to evaluate the cost effectiveness of screening to identify individuals susceptible to HSV infection in a hypothetical cohort of 100 000 pregnant women in their second trimester of pregnancy. Final outcomes were the projected incidence of maternal and neonatal HSV, quality-adjusted life expectancy and life-time costs associated with neonatal HSV.
Results In the absence of testing, model projected incidence of neonatal HSV was 34 cases per 100 000 births, similar to available surveillance data. Screening pregnant women and their partners was projected to decrease the incidence of HSV-1 and HSV-2 infections in women and infants and to save costs. These findings were robust under alternative assumptions and in wide-ranging sensitivity analyses.
Conclusions The use of accurate and relatively inexpensive serological tests for HSV to identify women vulnerable to incident HSV infection in pregnancy has the potential to reduce neonatal HSV incidence and reduce health-related costs.
- Cost–utility analysis
- economic analysis
- herpes simplex virus
- mathematical model
- neonates
- pregnancy
- screening
Statistics from Altmetric.com
- Cost–utility analysis
- economic analysis
- herpes simplex virus
- mathematical model
- neonates
- pregnancy
- screening
Herpes simplex virus (HSV) infections in newborns are an uncommon but potentially devastating consequence of genital HSV infection in women. Surveillance efforts in Canada and the USA have resulted in incidence estimates that range from five to over 100 cases per 100 000 live births; 15–20% of affected infants do not survive, whereas surviving infants are at high risk of developmental delay and other neurocognitive problems.1–4 Approximately 45% of neonatal infections are due to HSV type 1, with the remainder due to infection with HSV type 2.3
To date, most medical strategies aimed at the prevention of neonatal herpes virus infection have focused on preventing transmission by women with prevalent genital herpes infection, either through the use of suppressive antiviral drugs, or through Caesarian section delivery in women with genital HSV and visible lesions at the time of parturition.5 In this way, the identification of ‘infected’ individuals is consistent with the usual paradigm for perinatal identification of infectious diseases, including HIV, hepatitis B and syphilis.6 However, more recent epidemiological evidence suggests that infants at highest risk of neonatal HSV infection are actually those born to women with incident HSV acquired during the third trimester of pregnancy, without an adequate maternal immune response before birth.3 7 Infants born to women experiencing first episode HSV infection in the third trimester of pregnancy have a 33-fold increased risk of neonatal HSV infection.3
Given this improved understanding of neonatal HSV epidemiology, the identification of pregnant women who are HSV uninfected but at risk of de novo HSV acquisition may be a more effective and cost-effective means of preventing vertical HSV transmission. In particular, accurate, relatively inexpensive serological tests for HSV could be used to identify women who are vulnerable to incident infection in pregnancy. In an observational study, knowledge of partner risk status reduced infection risk by 50%.8 Identification of discordant HSV infection status in a couple, through serological testing, or serological testing that identifies vulnerability to HSV infection in a pregnant woman, could similarly reduce the risk of HSV acquisition. If the timing of such testing were such that incident third trimester HSV infection was prevented, this could be a novel, effective and cost-effective strategy for preventing neonatal HSV.
Our primary objective was to evaluate whether a strategy of screening pregnant women to identify those susceptible to HSV infection would be effective and economically attractive relative to currently accepted health interventions. We used decision analytical methods to synthesise the best available health and economic data related to HSV in pregnancy.
Methods
Model overview
We constructed a Bayesian decision tree model (using Treeage Pro Suite, Williamstown, Massachusetts, USA) that incorporated the probability of baseline HSV infection status in women and their partners, the likelihood of HSV transmission to a pregnant woman during pregnancy, the risk of vertical HSV transmission at parturition and future costs and consequences of neonatal HSV infections (see figure 1 for simplified model schematic). Bayesian models allow probabilities to change in light of new evidence, in a causal and evidential manner, to define important clinical outcomes and costs. The model was used to evaluate the cost effectiveness of plausible testing strategies in a target population of hypothetical pregnant women presenting for prenatal care in their second trimester of pregnancy (24–28 weeks).
Strategies
We compared two strategies: no intervention and serological screening of pregnant women and their partners to identify serodiscordant couples. After testing, women could be correctly or incorrectly classified as infected or uninfected; incorrect classification on either measure would result in the inappropriate use of resources. In all strategies, prevalent HSV infection in pregnancy was managed according to current guidelines from the American College of Obstetricians and Gynecologists, with Caesarean section recommended for women with visible genital lesions.9 The incidence of HSV and transmission to the infant, with associated costs and consequences, were measured as final outcomes.
Identification of serodiscordant couples was assumed to reduce the risk of HSV transmission.8 We assumed that HSV-1 infection did not protect against infection with HSV-2, and vice versa, and that co-infection with both strains did not modify a woman's risk of transmission of either virus to the infant.10 Infected men were assumed to be equally infectious whether or not they were infected with one or both strains. We assumed that each couple was long term, monogamous and heterosexual.
Probabilities, outcomes and costs
Determination of model parameters, including data on baseline prevalence and the proportion of couples with discordant infection status, have been described in detail elsewhere.11 Baseline probabilities were derived from published literature and expert opinion (table 1).12 13 Natural history data for HSV transmission rates from mother to infant, by mode of delivery (Caesarean section vs vaginal) and serological status, were derived from a Washington State cohort study by Brown et al.3
Derivation of costs associated with screening, delivery and the occurrence of neonatal herpes virus infections has been described previously.11 All costs were converted to 2009 US dollars using the medical care component of the consumer price index.14 Quality-adjusted life expectancy for infants with neonatal HSV was calculated using estimates of the probability of disease severity (normal/mild, moderate, severe and dead),21 with utility weights estimated using the health utilities index 2.22 Disease outcomes were rated along the dimensions of sensation, mobility, cognition and self-care, with moderate diminution in each of these domains in individuals with a moderately severe outcome and severe diminution in each domain in severely affected individuals. Base case utility weights were 1, 0.84, 0.41 and 0 for the normal/mild, moderate, severe and dead outcomes, respectively, and these weights were used to transform survival on a life-year scale into survival in quality-adjusted life-years (QALY). A base case discount rate of 3% was applied to future costs and outcomes associated with neonatal HSV infection.14
Simulations
Base case cost-effectiveness analysis was performed through deterministic simulations using average or most likely parameter values as appropriate. To evaluate the frequency of HSV transmission, and the proportion of transmissions attributable to HSV-1 and HSV-2, we performed second-order Monte Carlo simulations. These simulations use a random number generator to create unique patients, and move them through a series of chance events over time. The term ‘second-order’ refers to the fact that parameter values were sampled from distributions (β or triangular) for probabilities (table 1). Results are presented based on 1000 simulations of cohorts of 100 000 women for each strategy.
Sensitivity analyses
We conducted sensitivity analyses to test model assumptions, variance in costs and the generalisability of given parameters to the population. Plausible ranges were derived from available CI, the highest and lowest values found in the published literature, or plausible ranges and are included in table 1. In addition to one-way or multi-way sensitivity analyses, in which individual parameters, or groups of parameters, when appropriate, were varied, we conducted probabilistic sensitivity analysis, with all parameters sampled from distributions simultaneously.
Although there is no evidence in the literature that knowledge of a pregnant woman's risk status alone (ie, without knowledge of the partner's infection status) is sufficient for reducing the risk of HSV acquisition, we considered a strategy of serological screening of pregnant women only in sensitivity analyses. In particular, we determined the necessary risk reduction, compared with that observed when screening couples, required to make screening pregnant women only a feasible alternative strategy.
Results
Model calibration and projected incidence of HSV infection
In the absence of testing, approximately 1.3% of women were projected to acquire HSV infection after the first trimester of pregnancy (figure 2), which is comparable to reported rates of 2%.7 The model projected mean incidence of neonatal HSV of 34 cases per 100 000 births was similar to available surveillance data.3 Among neonatal cases, the projected proportion due to HSV-1 was comparable (49%) to that reported previously (45%).3 Testing decreased the incidence of HSV-1 and HSV-2 infections in the mother and infant.
Cost and cost effectiveness
The clinical outcomes and costs of the two strategies in a hypothetical cohort of 100 000 pregnant women are presented in table 2. In the base case, testing couples dominated the no-testing strategy. A strategy is dominated when it costs more, but provides less health benefit than a competing strategy; a dominated strategy would not be chosen regardless of available health resources.23
Sensitivity analyses
The model was sensitive to assumptions about the reduction in risk associated with the identification of serodiscordance or seronegativity (figure 3A). Our analysis showed that screening pregnant women only was more cost effective than screening women and their partners, assuming similar levels of compliance in the absence of knowledge of partner HSV status. This observation is attributable to test characteristics (test sensitivity and specificity), with an increased likelihood of misclassification of at-risk women when also testing their partners, resulting in failure to apply the intervention appropriately. When the relative risk of HSV acquisition following identification of serodiscordance or seronegativity was greater than 0.95 or 0.98, respectively, testing was no longer cost effective, assuming a societal ‘willingness to pay’ of US$50 000 per QALY.
In threshold analysis, testing couples was more cost effective than screening pregnant women alone when testing in the pregnant woman alone inflated the relative risk of HSV acquisition by 9% or more (ie, if the RR for couple testing was 0.5, the RR associated with testing the pregnant woman alone was 0.55 or more). When we performed sensitivity analyses on the degree to which reduced compliance, resulting from lack of knowledge about the partner's HSV status, diminished the attractiveness of screening pregnant women alone, we found that the effectiveness of the intervention itself determined the compliance threshold at which testing of the pregnant individual would be preferred to testing the couple (figure 3B).
For all remaining sensitivity analyses testing couples was compared with no testing, again assuming a willingness to pay of US$50 000 per QALY. In univariate and multivariate sensitivity analyses evaluating the attractiveness of testing over a wide range of test costs and characteristics, costs of neonatal infection, and quality-adjusted life expectancy associated with neonatal HSV, the model was robust for all parameters over the ranges examined.
Baseline estimates of the proportion of couples in serodiscordant partnerships, the risk of HSV acquisition by vulnerable women and the probability of non-seroconversion after HSV infection were not adjusted for the timing of testing in the second trimester. Varying these estimates to exclude HSV acquisition events occurring in the first trimester did not alter the dominance of the testing strategy. In addition, no effect was seen on acceptability over a wide range of partnership and transmission probabilities.
Given the robustness of model outputs to a range of parameter estimates, we performed a probabilistic sensitivity analysis in which we assessed the effect of varying all model parameters simultaneously, with values drawn from distributions with the characteristics outlined in table 1 (figure 4). The testing of pregnant women and their partners was cost effective in over 99% of simulations, assuming a willingness to pay of US$50 000 per QALY, with approximately 90% of simulations projecting testing to be cost saving. We also generated worst and best case estimates of cost effectiveness by using either their minimum or maximum plausible parameter values. In the worst case scenario (in which test sensitivity was low, RR of HSV acquisition was 0.8 following identification of serodiscordance, Caesarean section did not reduce the risk of HSV transmission, and the lifetime costs and consequences of neonatal HSV were low), the incremental cost-effectiveness ratio for testing compared with no testing was US$130 130 per QALY. In the best case scenario (perfect test sensitivity, highly effective intervention, and high lifetime costs and consequences of neonatal HSV infection), testing couples was a highly cost-saving intervention, resulting in increased QALY at lower costs than the alternative of not testing.
Discussion
Using a model parameterised with the best available data, we have demonstrated that a screening strategy identifying women at risk of incident HSV during pregnancy is projected to result in reductions in neonatal HSV and net savings in health costs. Our findings are robust under alternative assumptions and in the face of wide-ranging sensitivity analyses. We have also shown that, assuming a similar reduction in HSV acquisition risk to that observed in studies of serodiscordant couples, screening of pregnant women alone tends to be more cost effective than screening couples. Screening couples would be favoured only when intervention compliance is substantively enhanced when partner status is known, and the effectiveness of couple screening is actually projected to be less effective than screening women alone, due to ‘false reassurance’ resulting from falsely negative testing in partners of uninfected women. Importantly, the projected gains seen in our model are achieved in the absence of specific interventions targeted to pregnant women with chronic HSV infection, who have historically been the target of strategies aimed at preventing perinatal HSV transmission.
Our analysis differs from previous evaluations of the effectiveness of serological testing for the prevention of neonatal HSV,24–26 in that we consider both HSV-1 and HSV-2 transmission and incorporate more recent data on the likelihood of discordant partnerships and the probability of HSV transmission based on partner serostatus. Given the increasing frequency of HSV-1 genital infections,27 28 the inclusion of HSV-1 in models of neonatal HSV transmission is critical. The approach to the prevention of infection proposed here is also, to our knowledge, unique: in contrast to screening strategies related to the prevention of perinatal or congenital transmission of HIV, syphilis, toxoplasmosis and hepatitis B, which seek to identify women with otherwise unrecognised prevalent infection to prevent transmission, our focus here is on preventing incident perinatal infection by targeting uninfected women. The attractiveness of such a strategy relates to the markedly elevated risk of mother-to-infant HSV transmission when virus is present in the birth canal, in the absence of a fully developed immune response in the mother.7
As with any mathematical model, ours includes simplifying assumptions and incorporates parameter values that are subject to uncertainty. Nonetheless, the projection that the identification of pregnant women at risk of HSV acquisition would decrease the risk of neonatal HSV in a highly cost-saving manner remains robust in the face of wide-ranging sensitivity analyses. For instance, although we did not distinguish between genital and orolabial infections, which may account for up to 50% of HSV-1 infections,28 but are not associated with neonatal HSV, or account for the potential impact of horizontally acquired neonatal HSV, their impact on model projections can be inferred from the sensitivity analysis: in the face of reduced HSV-1 transmission rates to seronegative women and reduced intervention effectiveness following the identification of serodiscordant couples, screening remained highly cost effective. We did not evaluate the effect of other potential interventions, such as the use of suppressive antiviral therapy in infected partners, which might be expected to reduce the risk of neonatal HSV infection further.4 However, evidence of the effectiveness of partner antiviral therapy as a strategy for reducing neonatal HSV is presently lacking.29 Our analysis focused on the costs associated with neonatal HSV, and we did not evaluate the costs associated with HSV management in infected women.
In summary, we have shown that the use of accurate and relatively inexpensive serological tests for HSV could be used to identify women vulnerable to incident HSV infection in pregnancy. While clinical trials of such a strategy would be an ideal means of affirming or refuting our model projections, the rarity of neonatal HSV would make such trials a challenging undertaking. Models such as that presented here provide an efficient means for synthesising data that can guide policy while trials are pending, and can be a useful tool for the design of such trials. In some countries, such as Canada, there are existing sentinel surveillance systems for neonatal herpes,30 and such systems may be a useful interim means of identifying the impact of screening pregnant women to identify HSV seronegativity.
Key messages
A Bayesian decision tree model was parameterised using the best available health and economic data relating to HSV during pregnancy.
Identification of women in serodiscordant partnerships in their second trimester was projected to reduce neonatal HSV incidence in a cost saving manner.
Screening of pregnant women alone to identify seronegativity tended to be more cost effective than screening to identify serodiscordant couples.
Accurate, relatively inexpensive serological tests may be used effectively to identify women vulnerable to incident HSV infection in pregnancy and reduce neonatal HSV incidence.
References
Footnotes
Funding This work was supported in part by funds from the Public Health Agency of Canada. The views expressed in this manuscript are those of the authors and do not necessarily represent the views of the Public Health Agency of Canada.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.