Online appendix
The purpose of this appendix is to describe the mathematical model that is used to simulate STIs in South Africa, and the data that are used to set the model parameters. The model is a deterministic cohort component projection model that divides the population by sex and by 5-year age group, with demographic parameters obtained from the ASSA2003 AIDS and Demographic model of the South African population.1 The sexually active population is further divided according to marital status, propensity for concurrent partnerships, number of current partners and risk group of partner(s), with rates of transition between these sexual activity states changing over the life course. The population is projected from 1985, when HIV is assumed to be introduced into the population, and the spread of HIV is simulated based on specified probabilities of HIV transmission per act of unprotected sex. Adults who acquire HIV are assumed to progress from acute HIV infection to asymptomatic infection and then develop pre-AIDS symptoms, after which they either progress to untreated AIDS or initiate antiretroviral treatment.In the model presented here, STI transmission probabilities are assumed to be independent of HIV status, and HIV transmission probabilities are assumed to be independent of infection with other STIs. A more detailed explanation of the sexual behaviour and HIV assumptions is provided elsewhere.2
Syphilis
Individuals who acquire syphilis are assumed to develop genital ulcers (primary syphilis) after a short incubation phase, after which they progress to symptoms of secondary syphilis and ultimately latent syphilis, if treatment is not sought. Because the model is calibrated to seroprevalence data, it is necessary to make assumptions about the presence of syphilis antibodies in different stages of infection. In the incubation phase, all individuals are assumed to be seronegative, and in subsequent stages all individuals are assumed to be seropositive (the low sensitivity of the non-treponemal tests in primary syphilis is allowed for in the sensitivity assumptions). All individuals who are cured of syphilis in the secondary and latent phases – and a proportion of those cured in the primary phase – are assumed to remain seropositive for several months after recovery. Since the rate of seroreversion is more rapid in patients treated in earlier disease stages, the model has separate ‘post-treatment’ states to represent individuals who are still seropositive after treatment in early syphilis and those who are still seropositive after recovery from latent syphilis. Individuals are assumed to be immune to reinfection while they are still seropositive, based on evidence suggesting that individuals previously infected with syphilis acquire a degree of immunity.3 Individuals are also assumed to be infectious only during the primary and secondary syphilis stages.4 Treatment is sought at rate υ if the individual is symptomatic, and is effective with probability ψ; alternatively, individuals can be cured through screening at rate η. The states that are defined for the purpose of simulating the natural history of syphilis are shown in Figure 1, and the prior distributions that are chosen to represent the uncertainty around these parameters are specified in Table 1.
Figure 1: Multi-state model of the natural history of syphilis
Table 1: Syphilis parameters
Parameter / Symbol / Prior distributionType / Mean / SD / Ref.
Average time (in weeks) from
Infection to primary / / - / 4.4* / - / 4-6
Primary to secondary / / Gamma / 6.6 / 2.0 / 4
Secondary to latent / / Gamma / 15.6 / 4.0 / 4
Latent to spontaneous resolution / / Gamma / 520 / 150 / 5
Recovery in early disease to seronegative / / Gamma / 26.0 / 8.0 / 8-11
Recovery in late disease to seronegative / / Gamma / 52.0 / 16.0 / 6-8
Proportion of primary cases seronegative
immediately after successful treatment / φ / Beta / 0.40 / 0.10 / 9, 10
Transmission probability per act of sex
Male-to-female / - / Beta / 0.18 / 0.05 / 11-13
Female-to-male / - / Beta / 0.15 / 0.05 / 11-13
Proportion of cases correctly treated prior to
introduction of syndromic management / - / Beta / 0.70 / 0.10 / 14-17
* Fixed parameter, not included in Bayesian analysis.
SD = standard deviation.
Gonorrhoea
Individuals who acquire gonorrhoea are assumed to either develop symptoms or remain asymptomatic, and eventually experience spontaneous resolution of infection if treatment is not sought. As there is some evidence of strain-specific immunity following recovery from gonorrhoea,18, 19an additional state is defined to represent individuals who are temporarily immune following recovery. However, since successful early treatment of gonorrhoea does not appear to be followed by immunity,20 individuals are assumed to be immune only if they have experienced spontaneous resolution of infection. This model of natural history and immunity is illustrated in Figure 2, and the prior distributions for the associated model parameters are specified in Table 2.
Figure 2: Multi-state model of the natural history of gonorrhoea
Table 2: Gonorrhoea parameters
Parameter / Symbol / Prior distributionType / Mean / SD / Ref.
Proportion of cases symptomatic
Male / φ1 / Beta / 0.90 / 0.05 / 20-25
Female / Beta / 0.40 / 0.15 / 26, 27
Average duration if untreated (weeks)
Male / * / Gamma / 15.0 / 5.0 / 27, 28
Female / Gamma / 15.0 / 5.0 / 27
Average duration of immunity (weeks) / / Gamma / 52.0 / 26.0 / -
Proportion immune after cure / φ2 / - / 0.0† / - / 20
Transmission probability per act of sex
Male-to-female / - / Beta / 0.20 / 0.05 / 12, 13, 29-32
Female-to-male / - / Beta / 0.40 / 0.10 / 12, 13, 26,
31-33
Proportion of cases correctly treated
prior tointroduction of SM / - / Beta / 0.70 / 0.10 / 14, 16, 17, 34
* Same parameter is used for symptomatic duration (1/σ1) and asymptomatic duration (1/σ2). † Fixed parameter, not included in Bayesian analysis.
SD = standard deviation, SM = syndromic management.
Chlamydial infection
The model of chlamydial infection is identical in structure to that used for gonorrhoea (see Figure 2), but parameters differ. As there is substantial evidence of partial immunity to chlamydial infection following recovery,35-39 a longer average duration of immunity is assumed. Since immunity is thought to be more significant when treatment is initiated in late disease than in early disease,40-42it is assumed that only a proportion φ2 of those who are successfully treated acquire immunity (since treated symptomatic individuals would tend to have a shorter duration of infection than individuals who experience spontaneous resolution). The prior distributions assigned to the various parameters are shown in Table 3.
Table 3: Parameters for chlamydial infection
Parameter / Symbol / Prior distributionType / Mean / SD / Ref.
Proportion of cases symptomatic
Male / φ1 / Beta / 0.30 / 0.15 / 22, 24, 27, 43
Female / Beta / 0.15 / 0.08 / 22, 27
Average duration of untreated infection
Symptomatic (weeks) / / Gamma / 16.0 / 5.0 / 27, 44
Asymptomatic (weeks) / / Gamma / 90.0 / 15.0 / 45-47
Average duration of immunity (weeks) / / Gamma / 520 / 200 / -
Proportion immune after cure / φ2 / Uniform / 0.50 / 0.29 / -*
Transmission probability per act of sex
Male-to-female / - / Beta / 0.16 / 0.10 / 15, 16, 35, 36
Female-to-male / - / Beta / 0.12 / 0.06 / 15, 16, 35
Proportion of cases correctly treated
prior tointroduction of SM / - / Beta / 0.70 / 0.10 / 14, 16, 17, 34
* Due to the lack of evidence, a vague prior (uniform on the interval [0, 1]) is used.
SD = standard deviation, SM = syndromic management.
Trichomoniasis
The model of trichomoniasis is identical in structure to that used for gonorrhoea (see Figure 2), but parameters differ. As with gonorrhoea, there is little evidence of immunity following successful treatment of the infection,48, 49 but it is possible that individuals may be temporarily immune following the spontaneous resolution of infection. Prior distributions for the various trichomoniasis parameters are shown in Table 4.
Table 4: Trichomoniasis parameters
Parameter / Symbol / Prior distributionType / Mean / SD / Ref.
Proportion of cases symptomatic
Male / φ1 / Beta / 0.40 / 0.10 / 23-25, 43, 50
Female / Beta / 0.30 / 0.10 / 51, 52
Average duration of untreated infection
Symptomatic males (weeks) / / Gamma / 2.0 / 0.7 / 53, 54
Symptomatic females (weeks) / Gamma / 20.0 / 7.0 / -
Asymptomatic males (weeks) / / Gamma / 15.0 / 5.0 / 50, 55
Asymptomatic females (weeks) / Gamma / 150 / 50.0 / 56, 57
Average duration of immunity (weeks) / / Gamma / 52.0 / 26.0 / -
Proportion immune after cure / φ2 / - / 0.0* / - / 48, 49
Transmission probability per act of sex
Male-to-female / - / Beta / 0.15 / 0.08 / 58
Female-to-male / - / Beta / 0.04 / 0.02 / 58
Proportion of cases correctly treated
prior tointroduction of SM / - / Beta / 0.40 / 0.15 / 14, 17, 34
* Fixed parameter, not included in Bayesian analysis.
SD = standard deviation
STI treatment
It is assumed that symptomatic individuals seek treatment at a rate that depends on their age and sex. Based on evidence showing that women seek treatment less frequently than men,59-63 and youth seek treatment less frequently than older adults,60, 61 the weekly rate at which treatment is sought in adults aged 20 and older is assumed to be 0.57 in men and 0.23 in women. The rate at which teenagers with STI symptoms seek treatment is assumed to be half of that in adults, and sex workers are assumed to seek treatment at a rate of 0.90 per week, due to their more frequent experience of STI symptoms.59 Based on South African surveys of sources of STI treatment,64-67 it is further assumed that the public health sector treats 45% of male STI cases and 60% of female STI cases, and that the formal private health sector treats 40% of male STI cases and 30% of female STI cases. The remainder of STI cases are assumed to be treated by traditional healers.
As syndromic management protocols have been adopted more slowly in the private health sector than in the public health sector, the assumed proportions of health workers following syndromic management protocols in each year have been specified separately for the private and public health sectors (Table 5). The assumed proportion of private practitioners following syndromic management protocols is assumed to increase at 2% per annum after 2003, as the proportion of doctors trained in syndromic management increases,68 while the treatment practices in the public health sector are assumed to remain unchanged after 2003. The model also makes allowance for drug shortages in public STI clinics, which have gradually declined over time (Table 5).
Table 5: Assumed use of syndromic management protocols
Year / % of providers correctlyusing syndromic
management protocols / % of public
clinics with
STI drug
shortages / Reference
Private / Public
1993 / 0% / 0% / 20%
1994 / 3% / 10% / 20%
1995 / 7% / 30% / 18%
1996 / 11% / 50% / 16% / Harrison et al69*
1997 / 15% / 65% / 13% / Dartnall et al70†
1998 / 18% / 75% / 10% / Pick et al71*
1999 / 21% / 78% / 8% / Chabikuli et al34†
2000 / 23% / 80% / 6%
2001 / 25% / 80% / 5% / Schneider et al68†
2002 / 27% / 80% / 4% / Ramkissoon et al72*
2003 / 29% / 80% / 4% / Reagon et al73*
* Public health sector. † Private health sector.
The assumed effectiveness of treatment is specified for both syndromic and non-syndromic approaches to STI treatment, for each STI. As there is significant uncertainty regarding the effectiveness of STI treatment prior to the adoption of syndromic management protocols, beta priors are used to represent the uncertainty around the proportion of STI cases that are correctly treated in the absence of syndromic management (see Tables 1 to 4), based on limited data from South Africa and other African countries (summarized in Table 6). It is assumed that health workers who follow syndromic management protocols always provide correct treatment, except in the case of male trichomoniasis, which is assumed to be correctly treated in only 50% of cases (since current protocols74 recommend metronidazole for male urethral discharge only if the patient fails to respond to the initial treatment). Patients who are correctly treated are assumed to be cured in 90% of cases,6, 7, 75, 76 and treatment provided by traditional healers is assumed to be ineffective.
Table 6: Appropriateness of STI treatment in African studies, prior to the introduction of syndromic management protocols
Study / Location / Syndrome / n / % of cases treatedappropriately
NG / CT / TP / TV
Somsé et al17 / Bangui,
Central African
Republic / MUD / 64 / 69% / 19%
GUD / 28 / 82%
VD / 161 / 28% / 17% / 56%
Buvé et al16 / Mwanza,
Tanzania / GUD / 80 / 85%
MUD, VD / 235 / 16% / 21%
Chilongozi et al15 / Malawi / GUD / - / 56%
Mathews et al14 / Cape Town,
South Africa / MUDa / - / 86% / 75%
GUDa / - / 68%
VDa / - / 69% / 81% / 16%
Chabikuli et al34 / Gauteng,
South Africa / MUDb / 65 / 55% / 80% / 34%
MUDc / 65 / 80% / 66% / 32%
CT = Chlamydia trachomatis. GUD = genital ulcer disease. MUD = male urethral discharge. NG = Neisseria gonorrhoeae (gonorrhoea). TP = Treponema pallidum (syphilis). TV = Trichomonas vaginalis (trichomoniasis). VD = vaginal discharge.
a Syndrome was not recorded, but infections have been grouped according to the syndrome with which they are normally associated. b Treatment provided to cash patients. c Treatment provided to medically insured patients.
Women with asymptomatic syphilis are assumed to be treated through antenatal screening programmes, at a rate equal to their age-specific fertility rate, multiplied by the proportion of pregnant women who are screened (75%) and the proportion of those testing positive who receive treatment (80%).77-82
Sensitivity and specificity assumptions
Table 7 specifies the assumed sensitivity and specificity of the diagnostic techniques most commonly used for each STI, together with the assumed standard deviation of the sensitivity and specificity parameters when the tests are applied in different conditions by different investigators. In most cases the average sensitivity and specificity are calculated as the unweighted average of the estimates obtained from various studies, and the standard deviation is calculated from the sample variance of the same studies. In cases in which there are fewer than three estimates from which to calculate a sample variance, the standard deviation has been subjectively chosen to be similar to that of the same diagnostic used to detect a different STI, or to that of a related diagnostic used to detect the same STI. Studies have generally not been included if the ‘gold standard’ used to calculate the sensitivity and specificity parameters was judged to be insufficiently sensitive. A number of the assumptions are based on published reviews of sensitivity and specificity estimates.83-86
Table 7: Assumed sensitivity and specificity of different diagnostics
STI / Diagnostic / Sex / Sensitivity / Specificity / RefMean / SD / Mean / SD
Syphilis / Non-treponemal
+ treponemal
assays / M, F / 0.956 / 0.02 / 1 / 0 / 87-90
Non-treponemal
assay / M, F / 0.956 / 0.02 / 0.98 / 0.02 / 9, 10, 90-92
Gonorrhoea / Culture / F / 0.742 / 0.193 / 0.998 / 0.005 / 84
LCR on urine / M / 0.921 / 0.029 / 1 / 0 / 22, 93
F / 0.838 / 0.191 / 1 / 0 / 84
PCR on urine / M / 0.904 / 0.029 / 0.997 / 0.007 / 83
F / 0.556 / 0.202 / 0.987 / 0.018 / 83
PCR on swabs / F / 0.942 / 0.046 / 0.992 / 0.012 / 83
Chlamydial
infection / Direct immuno-
fluorescence / F / 0.763 / 0.02 / 0.988 / 0.008 / 94-96
LCR on urine / M / 0.875 / 0.121 / 1 / 0 / 84
F / 0.866 / 0.121 / 1 / 0 / 84
PCR on urine / F / 0.833 / 0.139 / 0.995 / 0.007 / 83
PCR on swabs / F / 0.855 / 0.116 / 0.996 / 0.005 / 83
Trichomoniasis / Wet mount / F / 0.59 / 0.123 / 0.991 / 0.021 / 85
Culture / F / 0.689 / 0.131 / 1 / 0 / 97-100
PCR on swabs / F / 0.95 / 0.05 / 0.98 / 0.024 / 86
PCR on urine / M / 0.95 / 0.05 / 0.98 / 0.024 / -
LCR = ligase chain reaction. PCR = polymerase chain reaction. SD = standard deviation.
STI prevalence data
The following tables summarize the STI prevalence data used in the Bayesian analysis. South African STI prevalence data were identified through a Medline search, through a hand search of the Southern African Journal of Epidemiology and Infection, through searches of abstracts of relevant conferences and through an earlier review of STI prevalence data in South Africa.101 This search was conducted in 2004 and subsequently published,102 and the database of STI sentinel surveillance studies has been augmented as new studies were identified. In cases in which the year of the survey was not stated, the year of the survey was imputed to be three years prior to the date of publication if published in a peer-reviewed journal, or one year prior to the date of publication if presented at a conference (based on the median publication lags for those studies in which the year of the survey was stated).
Table 8: Syphilis prevalence estimates
Study / Year / Sample / Location / n / Prev. / DiagnosticO’Farrell et al 103 / 1986-7 / ANC / Empangeni / 193 / 11.9% / Non-trep. + trep.
Donders et al 104 / 1988 / ANC / Pretoria / 256 / 9.0% / Non-trep. + trep.
Dietrich et al105 / - / ANC / Durban / 170 / 7.6% / Non-trep. + trep.
Coetzee 106 / 1990-2 / ANC / Cape Town / 1973 / 5.2% / Non-trep. + trep.
Opai-Tetteh et al 107 / - / ANC / Durban / 200 / 11.0% / Non-trep. + trep.
Bam et al79 / 1990 / ANC / Bloemfontein / 971 / 15.7% / Non-trep. + trep.
Hoosen et al108 / - / ANC / Durban / 32 / 12.0% / Non-trep. + trep.
Qolohle et al109 / 1993 / ANC / Durban / 363 / 9.4% / Non-trep. + trep.
Govender et al110 / 1994 / ANC / Durban / 168 / 12.0% / Non-trep. + trep.
Kharsany et al111 / 1994 / ANC / Durban / 52 / 26.9% / Non-trep.
Sturm et al112 / 1995 / ANC / Hlabisa / 327 / 12.0% / Non-trep. + trep.
Sturm et al113 / 1996 / ANC / Hlabisa / 327 / 8.4% / Non-trep. + trep.
Mashiane et al114 / 1997 / ANC / Pretoria / 3000 / 12.4% / Non-trep. + trep.
Dawadi et al78 / 1997-8 / ANC / Hewu / 271 / 8.5% / Non-trep.
Myer et al82 / 1998-
2000 / ANC / Hlabisa / 7391 / 7.5% / Non-trep.
Sturm et al112 / 1999 / ANC / Hlabisa / 245 / 6.0% / Non-trep. + trep.
2002 / ANC / Hlabisa / 449 / 2.0% / Non-trep. + trep.
Ramjee et al 115 / 1996-
2000 / CSW / KZN / 395 / 31.4% / Non-trep. + trep.
Steen et al116 / 1996-7 / CSW / Virginia / 407 / 33.8% / Non-trep. + trep.
Dunkle et al117 / 1996-7 / CSW / Johannesburg / 295 / 25.6% / Non-trep. + trep.
Williams et al59 / 1998 / CSW / Khutsong / 121 / 23.3% / Non-trep. + trep.
Williams et al118 / 2000 / CSW / Khutsong / 93 / 34.4% / Non-trep. + trep.
Ndhlovu et al67 / 2001 / CSW / Khutsong / 101 / 21.0% / Non-trep. + trep.
Hoosen et al119 / 1986 / FPC / Durban / 50 / 8.0% / Non-trep. + trep.
Schneider et al120 / 1994 / FPC / Bushbuck-
ridge / 249 / 5.0% / Non-trep. + trep.
Wilkinson et al121 / - / FPC / Hlabisa / 189 / 8.0% / Non-trep. + trep.
Kharsany et al111 / 1994 / FPC / Durban / 55 / 21.8% / Non-trep.
Hoosen et al122 / - / FPC / Durban / 40 / 8.0% / Non-trep. + trep.
Fehler et al123 / - / FPC / Johannesburg / 210 / 8.6% / Non-trep. + trep.
Frohlich et al 124 / 2002 / FPC / Vulindlela / 221 / 2.2% / Non-trep. + trep.
Cronje et al125 / - / HH, F 20-49 / Urban FS / 403 / 15.5% / Non-trep. + trep.
HH, F 20-49 / Rural FS / 465 / 12.3% / Non-trep. + trep.
Colvin et al126 / 1995 / HH, F 15-49 / Hlabisa / 142 / 8.5% / Non-trep. + trep.
Williams et al59 / 1998 / HH, F 15-59 / Khutsong / 712 / 9.7% / Non-trep. + trep.
Auvert et al127 / 1999 / HH, F 15-24 / Khutsong / 622 / 4.5% / Non-trep. + trep.
Colvin et al126 / 1995 / HH, M 15-49 / Hlabisa / 86 / 9.3% / Non-trep. + trep.
Williams et al59 / 1998 / HH, M 15-59 / Khutsong / 475 / 6.1% / Non-trep. + trep.
Auvert et al127 / 1999 / HH, M 15-24 / Khutsong / 560 / 1.8% / Non-trep. + trep.
Williams et al118 / 2000 / HH, M 15-19 / Khutsong / 606 / 8.1% / Non-trep. + trep.
Ndhlovu et al67 / 2001 / HH, M 15-59 / Khutsong / 532 / 5.0% / Non-trep. + trep.
Auvert et al128 / 2002 / HH, M 15-49 / Orange Farm / 438 / 3.2% / Non-trep. + trep.
ANC = antenatal clinic attenders. CSW = commercial sex workers. F = females. FPC = family planning clinic attenders. FS = Free State. HH = households. KZN = KwaZulu-Natal. M = males. Non-trep. = non-treponemal assay. Prev. = prevalence. Trep. = treponemal assay.
Table 9: Gonorrhoea prevalence estimates
Study / Year / Sample / Location / n / Prev. / DiagnosticO’Farrell et al103 / 1987 / ANC / Empangeni / 193 / 5.7% / Culture
Donders et al104 / 1988 / ANC / Pretoria / 256 / 3.9% / Culture
Dietrich et al105 / - / ANC / Durban / 170 / 4.1% / Culture
Hoosen et al108 / - / ANC / Durban / 32 / 6.0% / Culture
Govender et al110 / 1994-5 / ANC / Durban / 168 / 3.0% / Culture
Kharsany et al111 / 1994 / ANC / Durban / 52 / 5.8% / Culture
Sturm et al113 / 1996 / ANC / Hlabisa / 327 / 7.8% / Culture
Sturm et al112 / 1999 / ANC / Hlabisa / 245 / 7.0% / PCR on swabs
Sturm et al100 / - / ANC / Hlabisa / 185 / 7.6% / -*
Sturm et al112 / 2002 / ANC / Hlabisa / 449 / 4.0% / PCR on swabs
Ramjee et al115 / 1996-
2000 / CSW / KZN / 387 / 10.3% / Culture
Steen et al116 / 1996-7 / CSW / Virginia / 407 / 17.3% / LCR on urine
Dunkle et al117 / 1996-7 / CSW / Johannesburg / 295 / 23.3% / LCR on urine
Williams et al59 / 1998 / CSW / Khutsong / 121 / 15.7% / LCR on urine
Williams et al118 / 2000 / CSW / Khutsong / 93 / 16.1% / LCR on urine
Ndhlovu et al67 / 2001 / CSW / Khutsong / 101 / 10.0% / LCR on urine
Hoosen et al119 / 1986-7 / FPC / Durban / 50 / 10.0% / Culture
Schneider et al120 / 1994 / FPC / Bushbuck-
ridge / 249 / 3.0% / LCR on urine
Wilkinson et al121 / - / FPC / Hlabisa / 189 / 4.0% / Culture
Kharsany et al111 / 1994 / FPC / Durban / 55 / 5.5% / Culture
Hoosen et al122 / - / FPC / Durban / 40 / 5.0% / Culture
Fehler et al123 / - / FPC / Johannesburg / 210 / 8.6% / LCR on urine
Kleinschmidt et al129 / 1999-
2001 / FPC / Orange Farm / 538 / 3.9% / LCR on urine
Frohlich et al124 / 2002 / FPC / Vulindlela / 221 / 2.2% / PCR on swabs
Colvin et al126 / 1995 / HH, F 15-49 / Hlabisa / 137 / 5.8% / LCR on urine
Williams et al59 / 1998 / HH, F 15-59 / Khutsong / 712 / 6.9% / LCR on urine
Auvert et al127 / 1999 / HH, F 15-24 / Khutsong / 622 / 10.9% / LCR on urine
Williams et al118 / 2000 / HH, F 15-49 / Khutsong / 893 / 8.6% / LCR on urine
Ndhlovu et al67 / 2001 / HH, F 15-59 / Khutsong / 878 / 11.0% / LCR on urine
Pettifor et al130 / 2002-3 / HH, F 15-19 / Peri-urban
townships / 2624 / 3.5% / PCR on urine
HH, F 20-24 / 2002 / 3.5% / PCR on urine
Hurkchand et al131 / 2002 / HH, F 20-49 / Mbalenhle / 399 / 4.7% / PCR on urine
Colvin et al126 / 1995 / HH, M 15-49 / Hlabisa / 85 / 2.4% / LCR on urine
Williams et al59 / 1998 / HH, M 15-59 / Khutsong / 475 / 3.4% / LCR on urine
Auvert et al127 / 1999 / HH, M 15-24 / Khutsong / 560 / 2.9% / LCR on urine
Williams et al118 / 2000 / HH, M 15-49 / Khutsong / 606 / 3.3% / LCR on urine
Ndhlovu et al67 / 2001 / HH, M 15-59 / Khutsong / 532 / 4.0% / LCR on urine
Pettifor et al130 / 2002-3 / HH, M 15-19 / Peri-urban
townships / 2389 / 1.1% / PCR on urine
HH, M 20-24 / 1455 / 3.2% / PCR on urine
Hurkchand et al131 / 2002 / HH, M 20-49 / Mbalenhle / 291 / 3.9% / PCR on urine
* Diagnosed by culture and a series of genetic tests, which in combination would have had very high sensitivity and specificity.