The Three Approaches in STD Diagnosis and Treatment
Diagnosis of STDs is still problematic, partially because of minimal symptoms and asymptomatic cases (up to 80% of gonococcal and chlamydial infections in women may be asymptomatic). 14 There are different approaches to STD diagnosis and treatment, with different therapeutic and cost implications. The most accurate way to diagnose STDs is through laboratory diagnostic techniques, and the gold standard methods refer to those with the highest sensitivity and specificity. For example, new tests such as ligase chain reaction (LCR) and polymerase chain reaction (PCR) are able to detect Neisseria gonorrhoeae and Chlamydia trachomatis in urine or cervical specimens with 97% to 100% sensitivity and equally high specificities, 15 but they tend to be very expensive.
While not all laboratory tests are prohibitively expensive, the logistics required to analyze specimens often restrict access. Trained staff members must collect specimens, conduct the tests, and read the results. The samples and supplies require specialized equipment and storage. One of the most critical drawbacks is the need to send the specimens to a central reference laboratory for analysis, thus forcing patients to return for the results. Many patients do not return and are lost to follow-up.
Because of the inherent difficulties in providing laboratory diagnosis, the World Health Organization (WHO), in collaboration with field researchers, developed a less technical and less costly approach for STD diagnosis and treatment. Known as “syndromic management,” it was based on the theory that a specified set of signs and symptoms constitute a syndrome, and treatment is accordingly provided for the most likely organisms responsible for the syndrome. 16 Syndromic management reduces the costs and delays associated with laboratory tests and provides immediate treatment, thus preventing further transmission and complications that may arise due to loss to follow-up.
The WHO and others created four main algorithms for use in the diagnosis of urethral discharge, vaginal discharge, lower abdominal pain, and genital ulcer disease. While syndromic management is less expensive and logistically less complicated than gold standard diagnosis and treatment, it is not the most accurate test, especially for vaginal discharge syndromes. 17,18 In numerous studies, sensitivities for various vaginal discharge algorithms range from 23% to 96%, and specificities range from 34% to 84%. 19 In general, the higher the sensitivity, the lower the specificity, and the highest sensitivities and specificities are detected among high-risk populations.
In low-risk populations, such as antenatal clinic attendees, sensitivity of vaginal discharge algorithms can go as low as 12%. The low validity and reliability of the vaginal discharge algorithm are especially problematic, because this is usually the most common “syndrome” encountered. In one study, when the relative distribution of the four main STD syndromes was standardized, it was calculated that 40% of all episodes would be vaginal discharge syndromes. 20 In another study of STD syndromes among 2570 females in rural South Africa, 49% had discharge-only syndromes. 21
A third method for treatment considered in light of the difficulties of detecting all positive STD cases (both symptomatic and asymptomatic) is mass treatment. There are two main strategies for mass treatment: community-wide mass treatment and targeted mass treatment, which is sometimes referred to as presumptive treatment. One of the major objections to the strategy of mass treatment is related to overtreating individuals who are not infected and thus possibly contributing to drug-resistant strains of microorganisms. Resistant strains have been reported frequently against some of the cheaper drugs in the treatment of STDs. 22 The prevalence of STDs in the general population and in high-risk groups and the acceptability of mass treatment have been important concerns with regard to its implementation.
This study uses a decision tree analysis model to compare the cost-effectiveness of three protocols for the diagnosis and treatment of vaginal discharge and cervical infection due to C trachomatis and/or N gonorrhoeae. First we constructed decision tree models for each of the protocols. Then, we fed the probability for each decision into the analysis by doing an extensive literature search. For each of the protocols, univariate and multivariate analyses were run to calculate some output measures of cost-effectiveness and program impact.
We were also particularly interested in the use of azithromycin or doxycycline for the treatment of chlamydia in each of the protocols. Therefore, our analyses and results involve a comparison of the tree protocols with two different medication schemes. In the following section, first we describe how the decision trees were constructed. Then, model population and assumptions, cost data, and data analysis are elaborated in specific subsections.
The decision tree analysis models in Figures 1 to 3 show how the decision model was set up. For example, in the gold standard protocol, we started with a population of 1 million women in the community (Fig. 1). On the basis of the prevalence of C trachomatis infection (CTI) and N gonorrhoeae infection (NGI), a certain percentage of women will have the infection(s). The next step is the decision about the probability of having symptoms. Because the most common symptoms of NGI and/or CTI (i.e., vaginal discharge and/or dysuria) are not specific to these infections, women in both groups (those who have the infection and those who don’t) may have these symptoms, albeit with different probabilities.
Then, the next decision is about the probability that women will seek care. Here, again, there are differences between women who have symptoms and women who don’t in terms of the probability of seeking care. According to our model, we assume that a woman may seek care for her other health concerns and complaints. If this same woman has NGI/CTI (but not necessarily the symptoms), there is a certain probability that her infection will be recognized during the course of her seeking care for something else. The most realistic example is a woman with NGI/CTI (but without symptoms) who goes to a family planning clinic to get a contraceptive method and whose infection is recognized during the course of the services she receives at the clinic (e.g., risk assessment, history-taking, or physical examination).
It is also plausible that a woman without NGI or CTI, but with other infections or no infection at all, may have either vaginal discharge or dysuria and seek care for these symptoms. Our tree takes into consideration all of these scenarios. Once the probabilities about care-seeking behavior are set for those women with or without the infection and with or without the symptoms, the next decision point is about the capability of the diagnostic approach used in correctly diagnosing or ruling out the infection. This is where the different levels of sensitivity and specificity with the three approaches come into play.
Then, the type of medication used for treatment (doxycycline or azithromycin for chlamydia and ciprofloxacin for gonorrhea in our case) and the therapeutic cure rate are the next decision points. The issues of patient compliance and treatment efficacy are reflected in these subbranches. With additional consideration for the spontaneous cure rate, the last points on the tree yield those women who are either cured (with treatment or spontaneously) or not cured.
It is also possible to continue the tree with further decisions, such as the probability of having complications and the type of the complications. However, for the sake of simplicity and manageability, we preferred to finish the model at the cure/no cure level. It is important to clarify here that time horizon is limited in our models and that we did not include sequelae of NGI/CTI or the possibility of reinfection.
The decision trees of syndromic management (Fig. 2) and mass treatment (Fig. 3) are constructed in a very similar way, with the necessary adjustments. The decision tree models in Figures 1–3 also show the probabilities used in the analyses.
Model Population and Assumptions
In our analysis, we used a model population of 1 million South African women of reproductive age (15–45 years). We used South Africa as our model country because of its high burden of STDs and the relatively greater availability of data in the literature. In cases where it was not possible to find reliable data on South Africa, we either extracted data from other sub-Saharan African countries or used estimates from other STD models. In cases of highly disconcordant values for the same indicator, we followed one or more of the following approaches to decide which values to use in our estimates: (1) preferred data from relatively larger and more recent studies, (2) estimated a median representative value for our base case scenario, and (3) counted on the general estimates used in previously published STD models, such as for the probability of having mixed NGI and CTI or of seeking treatment.
If no estimate was available, we calculated an estimate on the basis of the most relevant study. For example, in the case of estimating the proportion of women with NGI/CTI but without symptoms (vaginal discharge and/or dysuria) and seeking treatment for some other health service or complaint, we calculated our own values according to our inferences from a study about the health-seeking behavior of women living in a periurban area of Cape Town for acute and chronic diseases. 23 Once the baseline values were set, the other values in the literature were used to establish range estimates for sensitivity analysis (Table 1). 6,11,15,18,26,55–78
Costs for STD treatment were derived from a review of the literature, and the costs are estimated from the perspective of the health care system and cover only incremental programmatic costs. We use this perspective because governments and organizations often have to decide the most cost-effective program to implement, i.e., whether STD management is a cost-effective intervention in relation to the other health needs of the country. Costs presented have been converted from the original figures to 2002 U.S. dollars, with use of the U.S. Consumer Price Index when the year of the original cost was indicated.
Table 2 24–26,30,39,78–84 provides a summary of studies that provide information on cost of treating STDs, and Table 3 85–90 specifically lists the cost of drugs used in NGI/CTI treatment. Most of the cost information relates to the syndromic approach. It should be noted from the table that the cost of treating an STD case lies in a wide range and can be anywhere from $2.48 per clinic client in Zimbabwe 24 to $55.39 per visit for STD treatment through freestanding referral services in Kenya. 25 In a more recent study, cost per STD treated in clinics was estimated between $9.26 and $15.00 for low and medium income countries in Africa for the year 2005. 26
For our estimates, we did not use those costs indicated as “per patient correctly treated” or “per true case treated,” because these costs also take into account the cost of overtreatment. For our analysis, we were interested only in the cost per client or visit, regardless of whether the diagnosis (with syndromic management) was a true positive or not.
On the basis of information in Tables 2 and 3, staff time for syndromic management was estimated at $9.00 per visit. This reflects an incremental, economic cost and includes training and supervision of staff members and their clinic time used for patients. All the costs used in our analyses are annualized unit costs. For our estimates, we considered that several of the syndromic management schemes would include some form of pelvic and/or speculum examination, sometimes coupled with simple laboratory tests such as microscopic examination of wet mount smears.
We estimated staff time for the gold standard protocol at $10 per client because some of the diagnostic tests could take more time or a revisit for results. Staff time for mass treatment was set at $4.50 per patient (or person in the community). In our model, mass treatment is defined at the community level, targeted to reproductive-age women. Thus, in our model, all women of reproductive age who agree to participate will be visited at their homes and provided drugs for one course of treatment for both NGI and CTI.
Cost of Diagnostic Tests
Diagnostic tests can be extremely expensive, especially if some of the more recent technologies such as PCR or enzyme immunoassay (EIA) are employed for either urine or cervical specimens. For example, costs for PCR can be approximately $21 per test, and for EIA, approximately $12. 27 However, with the newer tests available, such as AMPLICOR PCR, the prices are going down and the cost per test can go as low as $5, making it $10 per patient to include NGI and CTI diagnosis. One drawback of this new test is relatively higher levels of false-positives and the requirement of a confirming test for gonorrhea. 28,29 However, for our purposes, gold standard is defined as any diagnostics with more than 90% sensitivity and specificity and does not necessarily mark theoretical gold standards.
Over and Piot 12,22 estimate the costs for more advanced laboratory diagnostic tests, such as culture, to be $5 to $9. The costs of the tests also heavily depend on the volume of the purchase. Based on these and other published figures and also on the assumption that large quantities will be purchased for an intervention program, the cost of diagnostic tests for a gold standard protocol (culture for gonorrhea and PCR for chlamydia or PCR for both chlamydia and gonorrhea) is estimated to be $7.50 per test and $15 per patient.
Cost of Drugs
One of the major determinants of cost in our analyses was drug prices. Table 3 provides a summary of the cost of drugs for the treatment of gonococcal and chlamydial infections. For gonococcal infections, ciprofloxacin, a first-line drug with single-dose oral administration and relatively low cost, was chosen to be used as the drug of choice and was used as such in some previous studies. 6,30 Chlamydia treatment is more problematic in terms of combining efficacy and cost. Both doxycycline and azithromycin are indicated in the first line of drugs for the treatment of chlamydial infections, 17 and theoretically they are equally effective, with reported treatment efficacy rates of 96% to 100% for doxycycline and 93% to 100% for azithromycin. 31–33 Azithromycin is not included in the essential drug list of the WHO, 34 and it can be up to 100 times more expensive than doxycycline.
Even though the very low cost of doxycycline would make it very appealing, there is a growing literature on the cost-effectiveness of using azithromycin in treating chlamydial infections, especially when the cost of complications is taken into account. 35,36 It is conventional and conservative to assume that low compliance rates with the 7-day regimen of doxycycline will significantly reduce efficacy rates in comparison with single-dose azithromycin, even though there is one report of a high therapeutic success rate (94%) with doxycycline, even with low levels of compliance. 37 Additionally, in the mass community trial in Rakai, Uganda, azithromycin was preferred for its activity against Haemophilus ducreyi, many strains of N gonorrhoeae, and possibly incubating, early syphilis, in addition to C trachomatis. 6
In our cost-effectiveness analyses we ran two different scenarios, one using doxycycline and the other using azithromycin. We followed the assumption that the decrease in efficacy will be gradual, and the more days a patient forgets to take medication, the lower will be the efficacy. We used the best estimate for efficacy of 86% for doxycycline (with a plausible range of 80%–90%) and 96% for azithromycin (with a plausible range of 94%–98%), as in a previous cost-effectiveness analysis. 38
Relying on the figures in Table 3, we set drug costs at $0.50 for one course of doxycycline and $2.00 for one course of ciprofloxacin. For comparison, the cost of treating cervical infections according to Tanzanian recommendations (with cotrimoxazole and doxycycline) was indicated as $2.00 for drugs. 39 Azithromycin prices are more problematic to estimate because of the wide range and unclear information. For example, the cost of azithromycin for the Rakai trial was claimed to be low, and the trial was suggested as cost-effective by the study authors. 6,7 However, a critique of the Rakai trial claims that the drugs used (one course of intervention regimen, to include ciprofloxacin, azithromycin, and metronidazole) should cost around $22.50 per client, which definitely is very expensive for Africa. 9
Although azithromycin does not seem to be available for less than $10.00 for more developed countries, on the basis of information from a research team that had recently purchased and used azithromycin in South Africa, we have estimated that azithromycin costs $5.00 for one course of single-dose therapy. 40 Although other less expensive drugs are available to treat gonorrhea and chlamydia, many of the drugs recommended can be used only under specific circumstances, and strains resistant against some have been reported. 41
In order to calculate the total cost of NGI/CTI treatment for each protocol, we multiplied the total cost for diagnosis and treatment by the number of positive cases detected. For example, for mass treatment, we assumed a community-based protocol in which all individuals covered would have received treatment for both infections. Thus, the total cost of the mass treatment protocol would be the number of women covered in mass treatment (those who were reached and agreed to participate), multiplied by cost per woman ($4.50 + $2.50 = $7.00). For syndromic management, it was assumed that all individuals who come to a clinic undergo diagnosis ($9/patient) but that only those with diagnosed infection (true or false positives) receive treatment and incur an additional cost of $2.50 for drugs. Under the gold standard protocol, clinic time and laboratory costs (totaling $24) were charged for all individuals, but an additional drug cost of $2.00 for gonorrhea and/or $0.50 or $10 for chlamydia was calculated only for those who received treatment.
Analyses were conducted with use of three software programs: DATA (version 3.0; TreeAge Software, Williamstown, MA), a decision analysis software program; Microsoft Excel (version 7.0; Microsoft, Seattle, WA); and Crystal Ball 2000 (Decisioneering, Denver, CO), a stochastic modeling supplement for Excel. After the baseline values were set according to the literature, univariate sensitivity analyses were performed for the parameters used. Next, multivariate sensitivity analyses were undertaken to test the robustness of the cost-effectiveness estimates to changes in underlying assumptions. For both types of analyses, we ran 10,000 Monte Carlo simulations, using triangular distributions with upper and lower extremes.
In the following sections, only the results of multivariate analyses are presented. (The results of the univariate analyses can be found at http://big.berkeley.edu and are available upon request from the authors). Among the several outcome measures for cost-effectiveness, we primarily focus on total cost of the program and cost per NGI/CTI case cured by treatment for each protocol, by applying the two scenarios (i.e., doxycycline or azithromycin scenario). For an assessment of the impact of each protocol, percentage of positive NGI/CTI cases cured (all cured cases [including cases with treatment and spontaneous cures]/all positive cases) and the ratio of NGI/CTI cases cured with treatment (cases cured with treatment/all positive cases) are presented. A further analysis deals with explaining the factors that contribute to the variance of some output measures.
We are reporting here on 15 forecast variables in order to get a more concise picture of the comparison of the three treatment protocols (Table 4). In all three protocols and for all six possible scenarios (i.e., three protocols each with doxycycline or azithromycin), total cost of syndromic management was found to be the lowest, and azithromycin scenarios yielded higher total costs than doxycycline scenarios, results which were in line with those of the univariate analysis. However, analysis of the cost per NGI/CTI case cured with treatment yielded interesting results, which were also different from the results of the univariate analysis.
For example, in the doxycycline scenario, mass treatment became the most cost-effective, with a mean value of $60 (range: $36–$99), followed by syndromic management with $73 (range: $22–$535). In the same scenario, the gold standard yielded the highest cost per NGI/CTI case cured with treatment, at $99, with a range of $39–$337. In the azithromycin scenario, the rank order changed, and this time syndromic management appeared to be the most cost effective at $78 (range: $25–$428), followed by mass treatment at $87 (range: $53–$143). Gold standard again was the most expensive approach, with azithromycin at $92 per NGI/CTI cured with treatment (range: $39–$313).
For the gold standard protocol, even though the cost per woman in the community increased slightly with the use of azithromycin (from $3.23 to $3.36), as in the case of univariate analysis, treatment of chlamydial infections with azithromycin again appeared more cost-effective than treatment with doxycycline. Even when the azithromycin cost is estimated to be higher, at $10, the gold standard with azithromycin produces a lower cost per cure ($95) than with doxycycline ($99) (data not presented).
For the other two protocols, azithromycin proves to be less cost-effective than doxycycline, and for the mass treatment protocol, it actually causes a 45% increase in the cost per NGI/CTI cured with treatment. At the $10.00 value for azithromycin, cost per cure for syndromic management was $92, and it was $125 for mass treatment (data not presented). Figure 4 demonstrates the shift in the distribution of cost-per-cure values of mass treatment, from being the most cost-effective approach with a narrow range in the doxycycline scenario to becoming a costly intervention with a wider range.
For policy-makers, along with cost-effectiveness, another major concern in the implementation of a program would be the overall impact of the program (Table 4). In our model, there were close to 142,000 women among the 1 million who had either or both of the cervical infections. (Note that the mean estimate and value of the ranges show minor fluctuations in each protocol as a function of the multivariate analysis software). In this analysis, syndromic management seems to have the lowest impact, by curing only 40% to 41% (with doxycycline and azithromycin scenarios, respectively) of all positive NGI/CTI cases. The same forecast value was calculated as 47% to 49% for gold standard and 80% to 86% for mass treatment protocols, which clearly demonstrates that mass treatment overall would have the greatest impact in the cure of NGI/CTI.
The difference between the three protocols becomes more pronounced when the ratio of NGI/CTI cases cured with treatment to all positive cases is calculated. The difference between this estimated value (cases cured with treatment/all positives) and the previous value (all cured/all positives) is that the latter includes spontaneous cure and thus curtails the real effect of intervention (but nevertheless provides a final complete picture). In this instance, mass treatment again had the highest impact, where 71% to 80% of all cases would be cured with treatment, depending on the use of doxycycline or azithromycin (with respective ranges of 60%–80% and 71%–88%). Gold standard had a medium level of impact, curing 24% to 27% of all cases (with respective ranges of 8%–44% and 10%–49%). Syndromic management, however, performed poorly and could effectively treat only 14% to 16% of all positives (with respective ranges of 2%–35% and 2%–40% for doxycycline and azithromycin).
One step further in this exercise is to find out the percentage of women who have the symptoms and seek treatment yet are not cured because of flaws in the diagnostic method or treatment protocol. Women in this situation can be termed as missed opportunities. Mass treatment, again, is most dependable in these circumstances, and only 4% (range: 2%–6%) of those women who are assumed to participate in the intervention (took their medication) would be left uncured with the azithromycin scenario. When doxycycline is used for mass treatment, missed opportunities rise to 15% (range: 10%–20%). With the gold standard protocol, the figures for missed opportunities are 19% (range: 12%–27%) for doxycycline and 9% (range: 4%–15%) for azithromycin. Most dramatically, for syndromic management almost half of all infected women (51% with doxycycline and 45% with azithromycin) who have symptoms and who are assumed to seek care will not be cured.
Among the whole population of 1 million women, of which 11.7% were assumed to be positive for NGI and/or CTI with the gold standard protocol, 6% will still remain positive after the implementation because they did not seek treatment and they did not have spontaneous cure. With the syndromic management, 2% to 3% of the population would be treated by the implementation of the protocol, but another 5% would be overtreated. Mass treatment, however, would effectively treat 9% to 10% of the 11.7% who are infected, but at the same time 75% of the population would be overtreated.
Table 5 is a summary table of findings. For the sake of simplicity, it compares the six scenarios and shows the rank order of some selected outcome measures, based on mean values resulting from the multivariate analyses.
For each of the 15 forecast variables in Table 4, the ranges of the mean estimate are very wide. In further analyses, we wanted to understand which factors more significantly account for the variance observed. For both the gold standard and syndromic management protocols, the total cost of the program was most sensitive to the percentage of women seeking STD treatment and the prevalence of non-STD vaginal discharge, whereas the total cost of mass treatment was almost exclusively determined by coverage rates. For the cost per NGI/CTI cured by treatment, prevalence of women with vaginal discharge for reasons other than NGI/CTI was the most important for the gold standard protocol, accounting for 60% and 61% of the variance (for doxycycline and azithromycin, respectively), and was a significant factor in syndromic management (30% and 31% for doxycycline and azithromycin, respectively). For syndromic management and mass treatment, the factors that accounted for the highest variance were sensitivity (52%–46%) and prevalence of CTI (82%–84%), respectively.
For both gold standard and syndromic management protocols, the variance in cost of overtreatment was explained mainly by the specificity of the protocols, prevalence of women with vaginal discharge for reasons other than NGI/CTI, and percentage of women seeking NGI/CTI treatment. Prevalence of CTI and coverage were important determinants of the cost of overtreatment in the mass treatment protocol. Health-seeking behavior of women (seeking care for NGI/CTI and seeking other health services), together with the percentage of women who were symptomatic, explained 98% and 99% of all the variance observed in the percentage of women cured by treatment with the gold standard protocol.
For syndromic management, health-seeking behavior and sensitivity explained 95% and 96% of the variance, while for mass treatment the variance observed was explained by the coverage (71% and 95%) and treatment efficacy (29% and 5%) with doxycycline and azithromycin, respectively. In all of the analyses of the factors that contribute to variance, the parameters were more sensitive to the prevalence of CTI than NGI, which is due to the higher prevalence of CTI.
Our results have several important implications. The inference for policy-makers is that some priorities must be set in choosing one program over the other, on the basis of the specific circumstances of each field situation. It should be noted that this study involved a static analysis, with a case considered cured if it were cured on the initial visit. Cost issues related to partner treatment, reinfection, and complications are not addressed in this study. Obviously, these perspectives, together with their long-term effects, could bring into attention other results and implications. However, the models we have used are simple enough to replicate, but they still cover the major factors of cost and impact of STD services. Therefore, they can be carried over easily into other settings if the appropriate adjustments are made.
According to the Results, Which Protocol is the Best in Controlling NGI/CTI?
It should be clear from the presentation of the results that no single protocol carries with it all the desired conditions of a cost-effective program. The best program for a given locale may vary, but it is possible to lay out some characteristics of each protocol, as evident from the results of the multivariate analyses.
The gold standard protocol with azithromycin is never an initial choice and could be considered only as an alternative. Especially when compared with the syndromic management protocol with azithromycin, the gold standard protocol would cause a $14 increase in the mean cost per NGI/CTI cured but would provide the safety of a much narrower range for the highest cost possible. Additionally, the gold standard protocol would cure 27% of all positive cases, as compared with the 16% cured with syndromic management. However, we have used incremental costs in our analyses, and the clinic envisioned for gold standard service provision is assumed to have the required storage and transport facilities for samples, in addition to trained staff.
Furthermore, we adopted a health service perspective and thus excluded costs incurred by individuals and households. To what extent a specific clinic will fulfill these assumptions will be a significant factor in the decision to choose the gold standard protocol. In the case of sub-Saharan Africa, in an urban hospital where microbiology facilities and trained staff would be more readily available and there is a high load of patients, the gold standard protocol does not seem out of scope. However, in a rural and distant village, the gold standard would not seem realistic. If the gold standard protocol is being considered, our results imply that use of single-dose azithromycin for the treatment of chlamydia infections proves to be more cost-effective in terms of cost per NGI/CTI cured.
Azithromycin has an additional advantage of being effective against H ducreyi and some strains of N gonorrhoeae, even with a 1000-mg dose, and can be used against N gonorrhoeae at a 2000-mg dose. Azithromycin can also be considered as the drug of choice if multiple infections are targeted. The cost-effectiveness of azithromycin is also expected to increase with the inclusion of the cost of treating complications (e.g., pelvic inflammatory disease) in the cost-analysis model. 36,38,42
Syndromic management, to the contrary, has the advantage of overall low programmatic costs and relatively easier implementation. Actually, to date, most of the STD control programs in sub-Saharan Africa and in other resource-poor settings have used some type of syndromic management (e.g., Table 2). Our results suggest that in terms of impact, syndromic management functions quite poorly in comparison with the other protocols. If the sensitivity is low, there may be drastic increases in the cost per NGI/CTI cured because a significant proportion of the budget will be spent for treating false-positives, while a significant proportion of real positives will remain uncured.
In terms of cost per NGI/CTI cured by treatment, in our analyses the wide variation in costs with the syndromic management protocol, of which approximately half is explained by the sensitivity of the method, should warn decision-makers to be careful about the validity and reliability of syndromic management in their unique circumstances. Several studies have revealed that the sensitivity (and specificity) of syndromic management can vary widely, depending on several factors such as the prevalence of infection, the additional scoring or risk-assessment schemes used along with syndromic management algorithms, and the performance of the providers in recognizing the symptoms. 19
Sensitivity and specificity of syndromic management can go so low in some settings that the protocol performs no better than a random guess. Therefore, a thorough analysis of how the algorithms function in specific well-defined circumstances is essential in making the final decision about whether to use syndromic management.
Our scenario here for syndromic management is a “high-quality comprehensive syndromic management made available wherever patients seek care for an STD,” as indicated by Wilkinson. 43 However, a well-documented problem with syndromic management that we did not take into account in our analyses is the imperfect performance of the service providers. According to one study, only 9% of patients presenting with syndromic symptoms were adequately treated at South African clinics. 44 Another study in South Africa reported that only about 40% of the simulated patients received recommended drugs at clinics using syndromic management. 45 Even in cases where there was an intervention for strengthening syndromic management, the proportion of specially trained health professionals who provided correct treatment was 88%. 46
In the Masaka trial in Uganda, where practitioners were again trained to use syndromic management, the use of correct drugs for abnormal vaginal discharge syndrome was mentioned in 89% of the cases. 47 In Cote d’Ivoire, however, while the therapeutic algorithms were followed scrupulously by the 10 clinicians in 76% of cases, adherence was essentially connected to the genital ulcer syndrome. 30 It is interesting to note that the clinicians in this study abided by the protocol in 98% of cases of urinary discharge and in 100% of the 26 cases of genital ulcer. However, the algorithms for vaginal discharge with or without speculum examination were poorly observed (53% of cases with speculum examination and 44% of cases without speculum examination). 30 These studies indicate that correct diagnosis and provision of correct medication seem to be problematic with syndromic management, especially with the vaginal discharge syndrome.
For our analysis, we assumed for all treatment protocols that all diagnosed cases would be given correct treatment, since we lacked data about provision of correct medication for treatment protocols. There seems to be a general understanding that in circumstances where gold standard diagnosis and treatment are provided and supplies are adequate, the problem of wrong or insufficient medication (reasons mostly attributable to poor performance of health service providers) would be negligible. Incorrect medication will not be a major problem in mass treatment, as everyone gets the exact same treatment.
If we had included provider performance in analyses, the cost-effectiveness of syndromic management would have declined. However, syndromic management performs better with higher sensitivity levels for syndromes other than vaginal discharge 19 (e.g., urethral discharge syndromes in men, lower abdominal pain syndrome in women, and genital ulcer syndromes), and different syndromes could yield different results.
Mass treatment in our analyses appeared to be the most cost-effective protocol, with the highest impact in the doxycycline scenario. Despite the recent unexpected results from the Rakai trial, mass treatment, especially when it is targeted to a specific high-risk, high-prevalence population, still seems to be a good alternative. Using doxycycline or azithromycin in individuals who are infected is not expected to be associated with any serious health risks to them. However, development of resistant strains, reinfection from untreated men, and the ethics of mass treatment are some of the major concerns of using mass treatment. 48
Even if single-dose ciprofloxacin is unlikely to promote resistance in target organisms such as N gonorrhoeae, there is a possibility of promoting resistance in nontarget organisms such as Escherichia coli and Salmonella and Shigella species. 49 This potential adverse effect is true for syndromic management protocol also because of substantial treatment of false-positive cases, although it is seldom reported as an undesirable consequence of syndromic management programs. Available data on the effects of mass treatment or syndromic management on drug resistance are not conclusive.
One important conclusion about mass treatment is that it seems more plausible to use doxycycline in the protocol rather than azithromycin if only chlamydial infections are targeted, especially if the possible opportunity costs with azithromycin are not significant. Use of azithromycin in mass treatment more than doubles the total cost of the program and the cost per NGI/CTI case cured with treatment, but it is associated with an increase of only 6% in the ratio of all cured cases to all positives, increasing the mean percentage from 80% to 86%.
We interpret these results as a very high cost for a small benefit. It would be interesting, however, to conduct an exercise with the actual figures and costs of the Rakai mass trial, in which azithromycin was used, to see if doxycycline would really be almost as effective yet half as costly as azithromycin.
Despite the drawbacks and failures of both mass treatment and syndromic management, both protocols are still being supported for their different advantages, because of a general lack of better solutions in STD management. As it becomes more clear that solutions are complex and multifaceted, different combinations of intervention protocols are being increasingly supported.
This new approach is sometimes referred to as “hybrid” STD interventions. 50 Hybridity can involve a combination of treatment and prevention interventions, as currently being field-tested in Zimbabwe, Zambia, and South Africa, 50 or it can involve combining different treatment protocols. Periodic mass treatment, coupled with syndromic management and improved STD services, is now proposed as a good alternative. 51–53
To reduce STD transmission in high-prevalence communities in a short time, mass treatment can be an effective approach in the beginning, especially when high-risk core groups are targeted. However, mass treatment would have to be backed up with continuous access to STD treatment, 4 which then can be in the form of syndromic management. Another hybrid example from South Africa involved periodic presumptive treatment (monthly selective mass treatment for bacterial STDs with a directly observed 1-g dose of azithromycin), coupled with prevention education, as a feasible approach to providing STD services in a population of CSW. 54
Cost analyses of such hybrid interventions should be studied in more detail for better decision-making.
Some of the most important determinants of cost are those which cannot be directly controlled. For both the syndromic management and gold standard protocols, the nature of the infection and the response of the population (being aware or unaware of the infection) appear to be important cost determinants. Mass treatment is largely spared because the initiative mostly lies in the hands of the program performers rather than the population it intends to serve. In this study, the probability of having symptoms and the health care–seeking behavior of the immediate population are shown to have a great impact on the total cost and effectiveness of the programs and also on the cost of overtreatment.
A study by Wilkinson, 55 which showed that only about 3% of women in South Africa with STD symptoms seek treatment, further emphasizes the challenges posed for program planners by health care–seeking behavior.
One important decision that requires further discussion is related to our use of the presented figures on probability of seeking treatment. In one small study by Colvin 56 in South Africa, he found that of 26 individuals with STD symptoms in the past 43 months, 15.4% sought no treatment, 11.5% had traditional treatment, and 76.9% used medical services. However, another study demonstrated that of a population of rural South African women with vaginal discharge–causing STDs, 48% were asymptomatic, 50% were symptomatic but sought no treatment, 1.7% were symptomatic and would seek care, and 0.3% were seeking care on that very day. 55
In this case the difference between two values was too large to reconcile. Since the variable is one of the key points in any STD modeling, we preferred to use the estimates of the Piot–Fransen model for STDs. 11,49 According to this model, of women who have STDs and who are also symptomatic, 70% would seek treatment. It can be argued that these figures may change in the face of CTI and/or NGI because, for example, those with genital ulcers may seek treatment more than those with vaginal discharge alone. However, because of a lack of precise data on treatment-seeking behavior for NGI and/or CTI, we preferred to use 70% as the baseline estimate for the percentage seeking treatment (among NGI- and/or CTI-positive women who are symptomatic) and to use 77% as the maximum range for South Africa.
Clearly, using different figures instead or using the very low figures from South Africa would have changed the results dramatically.
Our results also imply that, if the same exercise is done with a hypothetical male population, costs would be different because NGI/CTI is more symptomatic and men with STDs have different health care–seeking behaviors. 57 Health education or any other type of program designed to increase awareness of and demand for services will increase the cost. However, the further a program achieves its goal to decrease STD prevalence, our model suggests that the total program cost will decrease (with syndromic management and gold standard protocols), but the cost for treating each infected individual will be higher.
Because prevalence is an important source of variation in the results, at different stages of the epidemic, different interventions will prove more effective and cost-effective. This issue was raised previously, in relation not to cost but to the impact of STD prevention programs in controlling the HIV/AIDS epidemic. 11 The interactions between cost, impact, and prevalence need to be investigated further.
NGI, CTI, and HIV/AIDS
Prevention and early treatment of curable STDs such as N gonorrhoeae and C trachomatis infections will result in considerable health gain. Piot and Rowley 12 calculated that for each NGI case prevented, 1.0 DHLY (discounted healthy life years) would be saved in women living in an urban area in Africa, and it would be 1.3 DHLY for each CTI case prevented. However, STD treatment has always been an integral part of HIV/AIDS prevention, especially in sub-Saharan Africa, where STD prevalence is high.
When the role of nonulcerative STDs in facilitating HIV transmission was reviewed, adjusted risk ratios for HIV seroconversion generally ranged from 3.2 to 5.1 for N gonorrhoeae and 2.7 to 3.6 for C trachomatis infection. 4 In another recent modeling exercise it was reported that of the estimated 40,000 to 80,000 new HIV infections that occur each year, 5052 or more could result from facilitating effects of syphilis, chlamydia, gonorrhea, and genital herpes on HIV transmission. 58 Because our primary objective was to compare cost-effectiveness of the different protocols in STD treatment, in our cost analysis we did not incorporate the potential benefits of preventing HIV/AIDS infections through NGI/CTI treatment.
Obviously, a further step in this analysis could include an exercise about how the cost and effectiveness of STD treatment impact HIV/AIDS prevention and treatment.
In conclusion, we suggest careful evaluation of each situation on the basis of the important cost-effectiveness and impact factors analyzed in this article. The models used are sufficiently broad to carry over into other contexts, aiding governments and other decision-makers in their choice of future STD control and prevention services.
1. Dallabetta G, Laga M, Lamptey P, eds. Control of Sexually Transmitted Diseases: A Handbook for the Design and Management of Programs. Arlington: AIDSCAP/Family Health International, 1996.
2. Gerbase A, Rowley J, Heymann D, Berkley S, Piot P. Global prevalence and incidence estimates of selected curable STDs. Sex Transm Infect 1998; 74 (Suppl 1): S12–S16.
3. Lamptey P. Sexually transmitted infections in sub-Saharan Africa: HIV brings much needed attention to a hidden problem. AIDS Infothek (Switzerland) 2000.
4. Fleming D, Wasserheit J. From epidemiological synergy to public health policy and practice: the contribution of other sexually transmitted diseases to sexual transmission of HIV infection. Sex Transm Infect 1999; 75: 3–17.
5. Grosskurth H, Mosha F, Todd J, et al. Impact of improved treatment of sexually transmitted diseases on HIV infection in rural Tanzania: randomised controlled trial. Lancet 1995; 346: 530–536.
6. Wawer M, Sewankambo N, Serwadda D, et al. Control of sexually transmitted diseases for AIDS prevention in Uganda: a randomised community trial. Rakai Project Study Group. Lancet 1999; 353: 525–535.
7. Wawer M, Sewankambo N, Serwadda D, Gray R. Authors’ reply. Lancet 1999; 353: 1523–1524.
8. Nicoll A, Johnson A, Adler M, Laga M. Preventing HIV-1: lessons from Mwanza and Rakai. Lancet 1999; 353: 1522.
9. Matthys F, Boelaert M. Letter to the editor. Lancet 1999; 353: 1523.
10. Hitchcock P, Fransen L. Preventing HIV infection: lessons from Mwanza and Rakai. Lancet 1999; 353: 513–515.
11. Hudson C. Community-based trials of sexually transmitted disease treatment: repercussions for epidemiology and HIV prevention. Bull World Health Organ 2001; 79: 48–58.
12. Piot P, Rowley J. Economic impact of reproductive tract infections and resources for their control. In: Germain A, Holmes K, Piot P, Wasserheit J, eds. Reproductive Tract Infections: Global Impact and Priorities for Women’s Reproductive Health. New York: Plenum Press, 1992: 227–248.
13. Watts C, Kumaranayake L. Thinking big: scaling-up HIV-1 interventions in sub-Saharan Africa. Lancet 1999; 354: 1492.
14. World Health Organization. An Overview of Selected Curable Sexually Transmitted Diseases. Geneva: World Health Organization, 1997.
15. Quinn T. Recent advances in diagnosis of sexually transmitted diseases. Sex Transm Dis 1994; 21 (Suppl 2): S19–S27.
16. World Health Organization. Management of Sexually Transmitted Diseases. Geneva: UNAIDS, World Health Organization, 1997.
17. World Health Organization. Guidelines for the Management of Sexually Transmitted Diseases. Geneva: World Health Organization, 2001(WHO/RHR/01.10).
18. Brabin L. Clinical management and prevention of sexually transmitted diseases: a review focusing on women. Acta Trop 2000; 75: 53–70.
19. Pettifor A, Walsh J, Wilkins V, Raghunathan P. How effective is syndromic management of STDs?. A review of current studies Sex Transm Dis 2000; 27: 371–385.
20. Van der Veen F, Fransen L. Drugs for STD management in developing countries: choice, procurement, cost, and financing. Sex Transm Infect 1998; 74 (supplement 1): S166–S174.
21. Wilkinson D, Connolly A, Harrison A, Lurie M, Abdool Karim S. Sexually transmitted disease syndromes in rural South Africa: results from health facility surveillance. Sex Transm Dis 1998; 25: 20–23.
22. Over M, Piot P. HIV infection and sexually transmitted diseases. In: Jamison D, ed. Disease Control Priorities in Developing Countries. New York: Oxford University Press, 1993.
23. Hoffman M, Pick W, Cooper D, Myers J. Women’s health status and use of health services in a rapidly growing peri-urban area of South Africa. Soc Sci Med 1997; 45: 149–157.
24. Maggwa N, Askew I, Simba S, et al. Demand for and Cost-Effectiveness of Integrating RTI/HIV Services with Clinic-Based Family Planning Services in Zimbabwe. Zimbabwe: The Population Council, 1999.
25. Soderlund N, Lavis J, Broomberg J, Mills A. The costs of HIV prevention strategies in developing countries. Bull World Health Organ 1993; 71: 595–604.
26. Schwartlander B, Stover J, Walker N, et al. Resource needs for HIV/AIDS. Science 2001; 292: 2434–2436.
27. Paavonen J, Puolakkainen M, Paukku M, Sintonene H. Cost-benefit analysis of first-void urine Chlamydia trachomatis screening program. Obstet Gynecol 1998; 92: 292–297.
28. Mukenge-Tshibaka L, Alary M, Bernier F, et al. Diagnostic performance of the Roche AMPLICOR PCR in detecting Neisseria gonorrhea in genitourinary specimens from female sex workers in Cotonou, Benin. J Clin Microbiol 2000; 38: 4076–4079.
29. Farrell D. Evaluation of AMPLICOR Neisseria gonorrhoeae PCR using cppB nested PCR and 16S rRNA PCR. J Clin Microbiol 1999; 37: 386–390.
30. La Ruche G, Lorougnon F, Digbeu N. Therapeutic algorithms for the management of sexually transmitted diseases at the peripheral level in Cote d’Ivoire: assessment of efficacy and cost. Bull World Health Organ 1995; 1995: 305–313.
31. Ossewaarde J, Plantema F, Rieffe M, Nawrocki R, De Vries A, Van Loon A. Efficacy of single-dose azithromycin versus doxycycline in the treatment of cervical infections caused by Chlamydia trachomatis. Eur J Clin Microbiol Infect Dis 1992; 11: 693–697.
32. Martin D, Mroczkowski T, Dalu Z, et al. A controlled trial of single dose of azithromycin for the treatment of chlamydia urethritis and cervicitis: the Azithromycin for Chlamydial Infections Study Group. N Engl J Med 1992; 327: 921–925.
34. World Health Organization. Essential drugs: WHO model list [revised December 1999]. WHO Drug Information 1999; 13: 249–262.
35. Haddix A, Hillis S, Kassler W. The cost effectiveness of azithromycin for Chlamydia trachomatis infections in women. Sex Transm Dis 1995; 22: 274–280.
36. Marra F, Marra C, Patrick D. Cost effectiveness analysis of azithromycin and doxycycline for Chlamydia trachomatis infection in women: a Canadian perspective. Canadian J Infect Dis 1997; 8: 202–208.
37. Bachmann L, Stephens J, Richey C, Hook E. Measured versus self-reported compliance with doxycycline therapy for chlamydia-associated syndromes: high therapeutic success rates despite poor compliance. Sex Transm Dis 1999; 26: 279–280.
38. Magid D, Douglas J, Schwartz S. Doxycycline compared with azithromycin for treating women with genital Chlamydia trachomatis infections: an incremental cost-effectiveness analysis. Ann Intern Med 1996; 124: 389–399.
39. Mayaud P, Uledi E, Cornelissen J, et al. Risk scores to detect cervical infections in urban antenatal clinic attenders in Mwanza, Tanzania. Sex Transm Infect 1998; 74 (Suppl 1): S139–S146.
42. Vrcic H, Zovko V, Culig J. Economic burden of STD in Croatia: incremental cost effectiveness analysis of azithromycin vs. doxycycline for treating Chlamydia trachomatis genital infections in women. In: Proceedings of ICMAS, Seville, Spain, 2000.
43. Wilkinson D. Syndromic management of sexually transmitted diseases in developing countries: what role in the control of the STD and HIV epidemics? Genitourin Med 1997; 73: 427–428.
44. Connolly C, Wilkinson D, Harrison A. Inadequate treatment for sexually transmitted diseases in the South African private health sector. Int J STD AIDS 1999; 10: 324–327.
45. Harrison A, Wilkinson D, Lurie M, Connoly A, Abdool Karim S. Improving quality of sexually transmitted disease case management in rural South Africa. AIDS 1998; 12: 2329–2335.
47. Walker D, Muyinda H, Foster S, Kengeya-Kayondo J, Whitworth J. The quality of care by private practitioners for sexually transmitted diseases in Uganda. Health Policy and Planning 2001; 16: 35–40.
48. Hudeson C, Smith A. Ethics of mass STD treatment. Lancet 2000; 356: 1115.
49. Hayes R, Wawer M, Gray R, et al. Randomised trials of STD treatment for HIV prevention: report of an international workshop. Genitourin Med 1997; 73: 432–443.
51. Steen R, Dallabetta G. The use of epidemiologic mass treatment and syndrome management for sexually transmitted disease control. Sex Transm Dis 1999; 26 (Suppl 4): S12–S20.
52. Van Dam C, Becker K, Ndowa F, Islam M. Syndromic approach to STD case management: where do we go from here? Sex Transm Infect 1998; 74 (Suppl 1): S175–S178.
53. Dallabetta G, Feinberg M. Efforts to control sexually transmitted diseases as a means to limit HIV transmission: pros and cons. Curr Infect Dis Rep 2001; 3: 162–168.
54. Steen R, Vuylsteke B, DeCoito T, et al. Evidence of declining STD prevalence in a South African mining community following a core-group intervention. Sex Transm Dis 2000; 27: 9–11.
55. Wilkinson D, Abdool Karim S, Harrison A, et al. Unrecognized sexually transmitted infections in rural South African women: a hidden epidemic. Bull World Health Organ 1999; 77: 22–28.
56. Colvin M, Abdool Karim S, Connolly C, Hoosen A, Ntuli N. HIV infection and asymptomatic sexually transmitted infections in a rural South African community. Int J STD AIDS 1998; 9: 548–550.
57. Fonck K, Mwai C, Rakwar J, Kiuri P, Ndinya-Achola J, Temmerman M. Healthcare-seeking behavior and sexual behavior of patients with sexually transmitted diseases in Nairobi, Kenya. Sex Transm Dis 2001; 28: 367–371.
58. Chesson H, Pinkerton S. Sexually transmitted diseases and the increased risk for HIV transmission: implications for cost-effectiveness analyses of sexually transmitted disease prevention interventions. J Acquir Immune Defic Syndr 2000; 24: 48–56.
59. Williams B, Gouws E. The epidemiology of human immunodeficiency virus in South Africa. The Royal Society of Philosophy. Trans R Soc Trop Med Hyg 2001; 356: 1077–1086.
60. Dietrich M, Hoosen A, Moodley J, Moodley S. Urogenital tract infections in pregnancy at King Edward VIII Hospital, Durban, South Africa. Genitourin Med 1992; 68: 39–41.
61. Wilkinson D, Ndovela N, Harrisson A, Lurie M, Connolly C, Sturm A. Family planning services in developing countries: an opportunity to treat asymptomatic and unrecognised genital tract infections? Genitourin Med 1997; 73: 558–560.
62. Schneider H, Coetzee D, Fehler H, et al. Screening for sexually transmitted diseases in rural South African women. Sex Transm Infect 1998; 74 (Suppl 1): S147–S152.
63. Pham-Kanter G, Steinberg M, Ballard R. Sexually transmitted diseases in South Africa. Genitourin Med 1996; 72: 160–171.
64. Diallo M, Ghys P, Vuylsteke B, et al. Evaluation of simple diagnostic algorithms for Neisseria gonorrhoeae and Chlamydia trachomatis cervical infections in female sex workers in Abijan, Cote d’Ivoire. Sex Transm Infect 1998; 74 (Suppl 1): S106–S111.
65. Laga M, Nzanze H, Brunham R, et al. Epidemiology of ophthalmia neonatorum in Kenya Lancet 1986;November 15: 1145–1148.
66. Vuylsteke B, Laga M, Alary M, et al. Clinical algorithms for the screening of women for gonococcal and chlamydial infection: evaluation of pregnant women and prostitutes in Zaire. Clin Infect Dis 1993; 17: 82–88.
67. Daly C, Franco L, Chilongozi D, Dallabetta G. A cost comparison of approaches to sexually transmitted disease treatment in Malawi. Health Pol Plan 1998; 13: 87–93.
68. Gray R, Wawer M, Sewankambo N, et al. Relative risks and population attributable fraction of incident HIV associated with symptoms of sexually transmitted diseases in Rakai District, Uganda. AIDS 1999; 13: 2113–2123.
69. Paxton L, Sewankambo N, Gray R, et al. Asymptomatic non-ulcerative genital tract infections in a rural Ugandan population. Sex Transm Infect 1998; 74: 421–425.
70. Paxton L, Kiwanuka N, Nalugoda F, Gray R, Wawer M, Group RPS. Community based study of treatment seeking among subjects with symptoms of sexually transmitted disease in rural Uganda. BMJ 1998; 317: 1630–1631.
71. UNAIDS, World Health Organization. Consultation on STD interventions for preventing HIV: what is the evidence? Geneva: Joint United Nations Programme on HIV/AIDS and World Health Organization, 2001(report no. UNAIDS/01.81E - WHO/HSI/2000.02).
72. Korenromp E, Van Vliet C, Grosskurth H, et al. Model-based evaluation of single-round mass treatment of sexually transmitted diseases for HIV control in a rural African population. AIDS 2000; 14: 573–593.
73. Grosskurth H, Mwijarubi E, Todd J, et al. Operational performance of an STD control programme in Mwanza Region, Tanzania. Sex Transm Infect 2000; 76: 426–436.
74. Hanson S, Sunkutu RM, Kamanga J, Höjer B, Sandström E. STD care in Zambia: an evaluation of the guidelines for case management through a syndromic approach. Intl J STD AIDS 1996; 7: 324–332.
75. Mehta S, Bishai D, Howell M, Rothman R, Quinn T, Zenilman J. Cost-effectiveness of five strategies for gonorrhea and chlamydia control among female and male emergency department patients. Sex Transm Dis 2002; 29: 83–91.
76. Parks K, Dixon P, Richey C, Hook E. Spontaneous clearance of Chlamydia trachomatis infection in untreated patients. Sex Transm Dis 1997; 24: 229–235.
77. Grosskurth H, Gray R, Hayes R, Mabey D, Wawer M. Control of sexually transmitted diseases for HIV-1 prevention: understanding the implications of the Mwanza and Rakai trials. Lancet 2000; 355: 1981–1987.
78. Mayaud P, ka-Gina G, Cornelissen J, et al. Validation of a WHO algorithm with risk assessment for the clinical management of vaginal discharge in Mwanza, Tanzania. Sex Transm Infect 1998; 74 (Suppl 1): S77–S84.
79. Marseille E, Kahn J, Billinghurst K, Saba J. Cost-effectiveness of the female condom in preventing HIV and STDs in commercial sex workers in rural South Africa. Soc Sci Med 2001; 52: 131–148.
80. Harrison A, Abdool Karim S, Floyd K, et al. Syndrome packets and health worker training improve sexually transmitted disease case management in rural South Africa: randomized controlled trial. AIDS 2000; 14: 2769–2779.
81. Mayaud P, Ka-Gina G, Grosskurth H. Effectiveness, impact and cost of syndromic management of sexually transmitted diseases in Tanzania. Int J STD AIDS 1998; 9 (Suppl 1): 11–4.
82. Gilson L, Mkanje R, Grosskurth H, et al. Cost-effectiveness of improved treatment services for sexually transmitted diseases in preventing HIV-1 infection in Mwanza Region, Tanzania. Lancet 1997; 350: 1805–1809.
83. Bulut A. Cost benefit analysis of sexually transmitted diseases. FEMS Immunol Med Microbiol 1999; 24: 461–467.
84. Kumaranayake L, Watts C. Economic costs of HIV/AIDS prevention activities in sub-Saharan Africa. AIDS 2000; 14 (suppl 3): S239–S252.
85. Hillis S. Doxycycline and azithromycin for prevention of chlamydial persistence or recurrence one month after treatment in women: a use-effectiveness study in public health settings. Sex Transm Dis 1998; 25: 5–11.
88. Parker K, Koumans E, Hawkins R, et al. Providing low-cost sexually transmitted disease services in two semi-urban health centers in Central African Republic (CAR): characteristics of patients and patterns of health seeking behavior. Sex Transm Dis 1999; 26: 508–516.