Blandford, John M. PhD*; Gift, Thomas L. PhD*; Vasaikar, Sandeep PhD†; Mwesigwa-Kayongo, Dan MD†; Dlali, Pumla MS†; Bronzan, Rachel N. MD*
CONGENITAL TRANSMISSION OF SYPHILIS INFECTION remains a significant source of adverse pregnancy outcomes in high-prevalence, underresourced settings where optimal antenatal screening and treatment have not been implemented.1–5 Syphilis prevalence in South Africa has been shown to be high, including among pregnant women.6–8 In South Africa’s rural Eastern Cape, the limited availability of services providing screening and treatment for maternal syphilis infections, the reliance of existing programs on off-site evaluation of specimens, and the lack of reliable transport result in many antenatal clinic (ANC) patients not receiving appropriate treatment to avert congenital transmission of syphilis infections.9 Early detection and prompt treatment of infections in pregnant women may substantially reduce adverse pregnancy outcomes, avert costs of subsequent treatment of live-born infants infected with syphilis, and benefit the health of mothers.10–13
A recent study in rural South Africa has demonstrated that programs offering on-site rapid syphilis testing and treatment in ANCs may effectively lead to the detection and treatment of the majority of maternal syphilis infections, averting a significant portion of adverse pregnancy outcomes (see companion paper).14 Fully informed health policy decisions must weigh relative effectiveness of alternative testing strategies against the cost of each approach.
We constructed a decision-analytic model in DATA Pro (TreeAge Software, Williamstown, MA) to estimate the costs and effectiveness of 3 approaches to screen and treat ANC patients for syphilis infection: (1) off-site evaluation of collected specimens using quantitative rapid plasma reagin (RPR) and confirmatory testing of all RPR-positive specimens using T. pallidum hemagglutination assay (TPHA), with patients returning for results and treatment (off-site RPR/TPHA); (2) on-site evaluation of specimens using a qualitative RPR, with same-day treatment of those testing positive (on-site RPR); and, (3) on-site evaluation of specimens using Determine immunochromatographic strip (ICS) assay (Abbott Laboratories, Chicago, IL), with same-day treatment of those testing positive (on-site ICS). The first approach, off-site RPR/TPHA, is used by the ANCs in South Africa’s Eastern Cape that currently offer syphilis testing. Because rural clinics are typically distant from district laboratories, and because transportation of specimens and test results can be unreliable, many ANCs in rural areas of the province offer no antenatal syphilis testing. Therefore, as a base measure for comparison, we modeled the expected disease outcomes and associated treatment costs in the absence of an antenatal syphilis-screening program.
We evaluated the cost-effectiveness of the 3 program alternatives from the perspective of the Eastern Cape Provincial Department of Health, the health authority responsible for implementation and funding of ANC services. The time frame of the analysis was the period of gestation, and our analytic horizon considered pregnancy outcomes through birth. The principal outcomes considered were averted adverse congenital syphilis events, specified as spontaneous abortion, perinatal death, and infants born infected with syphilis. Variables included in the cost-effectiveness model were based on primary data collected in the study and on published estimates (Table 1). Study-specific data comprised prevalence of syphilis infection among ANC attendees; cost data, including supply, equipment and labor costs; test performance characteristics, based on implementation during the study period; and treatment rates. Expected health outcomes were modeled from published estimates of mother-to-child transmission of T. pallidum, congenital syphilis outcomes, and efficacy of benzathine penicillin for antenatal treatment of syphilis infections.
The study period was from December 3, 2001, to March 22, 2002. Syphilis prevalence data among Eastern Cape ANC attendees were established by collection of serum specimens from participants and testing of specimens at the reference laboratory. Staging of disease was based on results of quantitative RPR and confirmatory TPHA conducted by the reference laboratory and was characterized as high-titer active, low-titer active, past or previously treated disease or biologic false positive.14
The likelihood of transmission of syphilis infection to the fetus has been shown to be dependent on the stage of maternal infection.1,15 Probability values used in the model are summarized in Table 1. We estimated the probability of fetal transmission given high-titer active maternal infection at 0.94; for low-titer active maternal infection, transmission probability was 0.37; for past infections or biologic false positives, we assumed transmission probability to be zero.4,14,16 The largest portion of adverse outcomes consists of syphilis-associated spontaneous abortion, followed by perinatal death and infants born with active syphilis infections.4,15,16 Low-birth weight outcomes were not considered.
A maternal syphilis infection was estimated to increase the likelihood of spontaneous abortion to 3–4 times greater than the background rate.16,17 A subsequent study has suggested that syphilis-associated adverse birth outcomes—in particular, for low-titer active infections—may be lower than those used in our base-case analysis.5,12 We address these findings in the sensitivity analysis of our cost-effectiveness model.
The likelihood of cure of maternal syphilis given a recommended course of therapy—for early maternal infection: single treatment of 2.4 million units benzathine penicillin administered intramuscularly; for late maternal infection: 3 treatments of benzathine penicillin administered at weekly intervals—was estimated to be 0.97 (Table 1).10 We assumed the fetus received full efficacy of therapy (0.97) following a single maternal treatment. Additional probabilities incorporated into the model included return rates to the clinic for subsequent treatments, likelihood of valid test results being received by clinics doing off-site evaluation, and treatment refusal rates.
We estimated performance of on-site RPR and on-site ICS using data gathered during the study period (Table 1). The methods used to assess test performance are detailed elsewhere.14 We assumed perfect sensitivity and specificity of off-site RPR/TPHA, because the testing protocol was identical to that employed in the study reference laboratory, the criterion standard for evaluating the test performance of on-site tests.
We assessed costs for testing and treatment supplies and equipment using prices quoted by local suppliers (Table 1). Equipment costs were straight-line amortized over the expected useful life. Labor costs were estimated for each screening strategy based on time-motion studies of clinic staff working on the project. The labor cost was calculated as the mean time required for a particular task multiplied by the median wage for the staff person(s) conducting the task.
The cost of treating live-born congenitally infected infants was estimated using the recommended South African regimen of 10 days of aqueous penicillin administered intravenously; daily hospitalization costs were based on regional averages (Table 1). We anticipated that the proportion of live-born infected infants being diagnosed and receiving recommended treatment would be equal to the proportion of neonates manifesting symptoms of congenital syphilis sufficiently severe as to require hospitalization, previously estimated at 30% of live-born infected infants.1,4,16
We did not consider medical costs for congenital syphilis beyond initial treatment of infected infants born with evident symptoms, nor did we include long-term direct medical costs of maternal syphilis infections that would progress to tertiary manifestations. Additionally, we did not consider the indirect costs of congenital syphilis, measured as productivity lost to syphilis-related disabilities or long-term sequelae of untreated infections, or as the productivity losses associated with the women’s time spent receiving screening and treatment.
Cost data were gathered in terms of South African Rand (ZAR) and converted to equivalent constant 2002 US Dollars (USD). During the time the study was underway, the mean conversion rate was 9.93 ZAR per 1 USD. We conducted one-way and multiway sensitivity analyses on all variables, using the full range of values with 95% confidence intervals where possible, and 50%–150% of baseline values where confidence intervals could not be calculated (Table 1). The study received formal approval from the institutional review boards of the CDC (USA) and the University of Witwatersrand (Republic of South Africa).
Summary measures of the cost and cost-effectiveness of the competing screening approaches to prevent congenital syphilis (CS) are presented in Table 2. With no screening program, the model predicted that 33 CS cases per 1000 pregnancies would result; this rate reflects measured prevalence rates of high-titer and low-titer active syphilis among ANC patients in the study clinics and likelihood of mother-to-child transmission based on stage of maternal infection. Off-site RPR/TPHA was the least expensive of the testing strategies—at a cost of USD2841 per 1000 ANC patients screened—and would prevent 18 of 33 CS cases expected with no screening and treatment. On-site RPR would be both less effective and more expensive than off-site RPR/TPHA, averting 15 of 33 expected CS cases at a cost of USD2950; in this context, implementation of on-site RPR would be a suboptimal use of health resources (Fig. 1). On-site ICS was the most expensive of the screening approaches and also the most effective, averting 27 of 33 expected CS cases, at a cost of USD3779 per 1000 ANC patients screened.
Incremental cost-effectiveness ratios—the net additional cost per additional unit of health benefit gained (measured as USD per averted CS case)—provide decision-makers with a tool to assess the additional resources required to attain a desired public health outcome. Adopting off-site RPR/TPHA would, compared to no program, avert 18 of the 33 expected CS cases per 1000 pregnancies at a cost of USD82 per case. Averting all but 6 of the CS cases per 1000 pregnancies could be realized through adoption of the on-site ICS strategy; averting 9 additional cases (compared to off-site RPR/TPHA) with an incremental cost-effectiveness ratio of USD104 per case.
One-way and multiway sensitivity analyses demonstrated the results to be stable over a wide range of probability and cost estimates. In one-way analyses, the modeled effectiveness of ANC screening approaches was most affected by test sensitivity of on-site ICS in detecting high- and low-titer maternal infections. Even at lower extremes of test sensitivity values for on-site ICS, however, the number of averted congenital syphilis cases exceeds that of other screening approaches. In multiway sensitivity analyses, only extreme assumptions would drive effectiveness of on-site ICS below that of other screening approaches. Sensitivity of on-site ICS would need to be less than 65% for detection of both high-titer and low-titer maternal infections, and the sensitivity of on-site RPR would need to exceed 87% in detecting high-titer infections for on-site ICS to be supplanted as the most effective of the screening approaches. Other variables shown to influence screening effectiveness—probability of mother-to-child transmission, probability of accepting presumptive treatment, syphilis prevalence rates, penicillin treatment efficacy—did not significantly alter the relative effectiveness of the 3 strategies.
The factors most influential in determining the relative cost-effectiveness of the screening strategies were prevalence of maternal syphilis and the relative distribution of active and past disease. While these variables did not impact the effectiveness of the screening approaches in detecting maternal infections, they did influence cost and, hence, cost-effectiveness measures. As estimates of maternal prevalence increased, on-site ICS became more cost-effective. For example, at the upper boundary of the confidence interval for maternal prevalence, 7.8%, the incremental cost-effectiveness of on-site ICS decreased from USD104 to USD85. At a prevalence rate of 10%, the incremental cost-effectiveness would decrease further to USD57. On-site ICS was sensitive to the prevalence of past disease, implying that relative cost-effectiveness would increase as rates of past disease increased, driven by the costs of unnecessary treatment following misdiagnosis. At the upper-bound prevalence estimate for past disease, 8.1%, the incremental cost of on-site ICS increased to USD122 per additional case averted.
We also evaluated the effect of assuming the rate of congenital transmission for low-titer infections to be zero, reflecting the findings of a recent study.5,12 With the assumption of zero probability of congenital transmission of low-titer infections, on-site ICS would remain the most effective of the screening approaches, averting 14 of the 16 adverse outcomes estimated to occur in the absence of a screening and treatment program. Off-site RPR/TPHA would avert 8 of the expected adverse outcomes and on-site RPR would avert 9. On-site RPR would also be somewhat less expensive than off-site RPR/TPHA, implying that off-site RPR/TPHA would be dominated by the on-site screening strategies. Compared to no program, the cost per case averted for on-site RPR would be $213. The incremental cost-effectiveness ratio of on-site ICS, compared to on-site RPR would be $218, approximately double that estimated in the base model. Incorporating a further assumption of decreased likelihood of adverse outcomes resulting from high-titer active infections had no significant impact on the relative effectiveness of the screening strategies but would increase the incremental cost-effectiveness ratios of the strategies.
Our study found that screening ANC patients using on-site ICS resulted in the largest proportion of infected patients receiving the recommended course of treatment. On-site ICS also averted the greatest portion of CS cases, preventing 82% of the CS cases that would be expected with no screening program. The next most effective screening strategy, off-site RPR/TPHA, averted 55% of expected CS cases. Screening with on-site RPR proved the least effective of the 3 approaches. The greater effectiveness of on-site ICS, however, entailed higher program costs than the other screening approaches evaluated. The incremental cost effectiveness of off-site RPR/TPHA of USD82 per CS case averted was somewhat lower than the USD104 for each of the additional CS cases averted using on-site ICS. While the additional cost is not trivial, especially in resource-poor settings, the advantages in terms of lower perinatal death rates, averting of CS-related lifetime disability, and prevention of late-stage maternal disease are substantial.
The on-site RPR screening approach proved to be least effective of the 3 approaches because of relatively low sensitivity in detecting both high- and low-titer maternal syphilis infections. The low field sensitivity of the test reflects the greater difficulty in accurately interpreting test results, a limitation exacerbated by frequent staff turnover in the study clinics. Test performance was not a factor in the standard-practice clinics, where off-site RPR/TPHA was used, because sensitivity and specificity were assumed to be equal to those of the study reference laboratory. The factor limiting the effectiveness of off-site RPR/TPHA was the loss to follow-up of ANC patients who did not return to the clinics for test results and treatment.
A recent study of adverse birth outcomes associated with maternal syphilis infection in Tanzania suggested that no adverse outcomes were associated with low-titer active maternal infections.5,12 Because this was a retrospective study of women presenting for delivery, spontaneous abortion was not one of the considered adverse outcomes of infection; as a result, data from the Tanzania study may not be fully applicable to the adverse outcomes modeled in the current study. Nonetheless, we considered the impact of these findings on cost-effectiveness measures. In the end, consideration of these findings from Tanzania would influence incremental cost-effectiveness ratios but does not impact the dominant effectiveness of on-site ICS relative to other screening strategies.
The primary disadvantage of on-site ICS stemmed from the inability of the treponemal test to distinguish active from past infections. The lower specificity of on-site ICS testing approach in identifying active infections, which would result in some overtreatment of tested ANC patients, should be weighed against the high sensitivity of the test in detecting both high- and low-titer infections and the opportunity to treat infected patients while they are still in the clinic.
The cost-effectiveness of on-site ICS will depend on the particular epidemiological context. As rates of high- and low-titer syphilis infections increase, the on-site ICS becomes increasingly cost-effective relative to other approaches. The cost-effectiveness of on-site ICS will also depend on the distribution of active and past syphilis infections. As rates of past infections in a population increase, the proportion of women who are unnecessarily treated will increase also. In contrast, the off-site RPR/TPHA approach and on-site RPR are largely insensitive to differences in rates of past infection. The impact of syphilis rates, and the distribution of disease between active and past infections, will influence the relative magnitude of differences in cost-effectiveness measures. Because the epidemiology of maternal syphilis infections may vary by population and time-period, the impact of these variables should be considered in assessing the relative advantages and cost-effectiveness of these interventions.
Because we assumed off-site RPR/TPHA to perform with perfect sensitivity and specificity, the cost-effectiveness study was designed such that any bias in cost-effectiveness measures would favor the off-site RPR/TPHA strategy. Should the off-site laboratories prove less accurate, then estimates of averted and treated disease would have been overestimated and measures of cost per case averted or cured would have been underestimated. If this were to be the case, the relative performance of both on-site ICS and on-site RPR would improve.
Another source of potential bias in estimates of performance of the standard-practice clinic strategy stems from the difficulty in locating standard-practice clinics comparable to the study clinics and representative of ANC clinics in the rural Eastern Cape.14 Because antenatal syphilis testing in the province was rare before the study period, clinics where antenatal screening was already in place were disproportionately located in urban or periurban areas. As a result, our study may have underestimated transportation difficulties that would adversely impact the effectiveness of this screening approach, including off-site evaluation of specimens, timely return of test results, and presentation of ANC patients at clinics to receive results and treatment. A more representative group of clinics doing off-site evaluation might have lower rates of successful treatment and less favorable cost-effectiveness ratios than were predicted in the model. If so, the relative estimated performance of on-site ICS and on-site RPR would improve.
Because we adopted the perspective of the Provincial Ministry of Health for the analysis, we did not consider patient costs. Because patient time and travel costs were excluded under this perspective, our study underestimated the value of being able to treat patients at their initial screening visit and therefore the relative cost of the off-site RPR/TPHA. No user fees associated with ANC syphilis screening in the Eastern Cape; in a context where user fees were assessed, we would expect the uptake of the more expensive on-site ICS to be negatively affected if user fees reflected relative test costs. We also underestimated the value of averting CS cases, because lost productivity costs of parents that would be incurred in caring for CS-affected infants were not included. Considering these costs would improve the relative cost-effectiveness of on-site ICS testing compared to off-site RPR/TPHA.
To avert adverse congenital syphilis events and to effectively treat syphilis infections of patients seeking antenatal care in high-prevalence settings, implementation of on-site screening with ICS may be a highly effective approach. The next most effective of the approaches evaluated relied on evaluation of specimens at an off-site laboratory using RPR/TPHA. On-site RPR was the least effective of the 3 approaches considered and was dominated on cost-effectiveness measures. While the incremental cost-effectiveness of off-site RPR/TPHA was marginally less than that of on-site ICS, both screening strategies may be considered cost-effective, depending on the available resources of the relevant funding authority and public priorities for syphilis control in mothers and infants.
1. Radolf JD, Sanchez PJ, Schulz KF, et al. Congenital syphilis. In: Holmes KK, Sparling PF, Mardh P-A, et al., eds. Sexually Transmitted Diseases. New York: McGraw-Hill; 1999:1165–1190.
2. Hira SK, Bhat GJ, Chikamata DM, et al. Syphilis intervention in pregnancy: Zambian demonstration project. Genitourin Med 1990; 66:159–164.
3. McDermott J, Steketee R, Larsen S, et al. Syphilis-associated perinatal and infant mortality in rural Malawi. Bull World Health Organ 1993; 71:773–780.
4. Schulz KF, Schulte JM, Berman SM. Maternal health and child survival: Opportunities to protect both women and children from the adverse consequences of reproductive tract infections. In: Germain A, ed. Reproductive Tract Infections: Global Impact and Priorities for Women’s Reproductive Health. New York: Plenum; 1992:145–182.
5. Watson-Jones D, Changalucha J, Gumodoka B, et al. Syphilis in pregnancy in Tanzania. I. Impact of maternal syphilis on outcome of pregnancy. J Infect Dis 2002; 186:940–947.
6. Pham-Kanter GB, Steinberg MH, Ballard RC. Sexually transmitted diseases in South Africa. Genitourin Med 1996; 72:160–171.
7. Sturm AW, Wilkinson D, Ndovela N, et al. Pregnant women as a reservoir of undetected sexually transmitted diseases in rural South Africa: Implications for disease control. Am J Public Health 1998; 88:1243–1245.
8. Wilkinson D, Abdool Karim SS, Harrison A, et al. Unrecognized sexually transmitted infections in rural South African women: A hidden epidemic. Bull World Health Organ 1999; 77:22–28.
9. Karuhije P. A pilot study on the feasibility, cost and cost-effectiveness of transporting specimens from clinics to the nearest laboratory. Northern Transkei, Eastern Cape Province, Republic of South Africa: National Health Laboratory Service; 2003.
10. Alexander JM, Sheffield JS, Sanchez PJ, et al. Efficacy of treatment for syphilis in pregnancy. Obstet Gynecol 1999; 93:5–8.
11. Sheffield JS, Sanchez PJ, Morris G, et al. Congenital syphilis after maternal treatment for syphilis during pregnancy. Am J Obstet Gynecol 2002; 186:569–573.
12. Watson-Jones D, Gumodoka B, Weiss H, et al. Syphilis in pregnancy in Tanzania. II. The effectiveness of antenatal syphilis screening and single-dose benzathine penicillin treatment for the prevention of adverse pregnancy outcomes. J Infect Dis 2002; 186:948–957.
13. Wilkinson D, Sach M. Improved treatment of syphilis among pregnant women through on-site testing: An intervention study in rural South Africa. Trans R Soc Trop Med Hyg 1998; 92:348.
14. Bronzan R, Mwesigwa-Kayongo D, Narkunas D, et al. Onsite rapid antenatal syphilis screening with an immunochromotagraphic strip improves case detection and treatment in rural South African clinics, in press.
15. Ingraham NR. The value of penicillin alone in the prevention and treatment of congenital syphilis. Acta Demato-Venereologica 1951; 31:60–88.
16. Schulz KF, Cates W Jr, O’Mara PR. Pregnancy loss, infant death, and suffering: Legacy of syphilis and gonorrhoea in Africa. Genitourin Med 1987; 63:320–325.
17. Temmerman M, Lopita MI, Sanghvi HC, et al. The role of maternal syphilis, gonorrhoea and HIV-1 infections in spontaneous abortion. Int J STD AIDS 1992; 3:418–422.