Methods

Historical review. The study was approved by ethics committees of the Central African Republic Ministry of Health and the Centre Hospitalier Universitaire de Sherbrooke in Canada. Historical data on the epidemiology of trypanosomiasis in the Nola-Bilolo focus were reviewed [9–18]. Annual incidence was measured as the total number of new cases and as Indice de Contamination Nouvelle (ICN), a measure of risk (new cases in a year divided by number of individuals never diagnosed as having the disease) [19]. Using ICN measures, we developed a model to calculate the proportion of individuals in each birth cohort treated for trypanosomiasis before 1951. We summed each birth cohort's risk for trypanosomiasis during each calendar year and assumed that (1) before 1936, the first year for which ICN was available, 20% of individuals aged ⩾6 years had had trypanosomiasis; (2) risk during the first 6 years of life was one-fifth that of older children and adults; and (3) 25% of patients with trypanosomiasis died of the disease or its treatment [13–18, 20].

Field survey. In October-November 2006, participants were recruited by house visits in (1) 31 villages west of Nola up to Bigéné (8 km from the Cameroonian border) known as pays Mbimou (Mbimou land), the area with the highest incidence of trypanosomiasis in the 1936–1950 period; and (2) Nola town and 3 villages along the Kadei river, grouped in forthcoming analyses (Figure 1). Inclusion criteria were age of ⩾55 years and willingness to consent. Exclusion criteria were dementia or inability to communicate. A questionnaire gathered sociodemographic characteristics, dates of migration, scarifications, circumcision, excision, and medical history. For participants unaware of their age, an estimate was made from historical events widely known locally. Vaccination scars were recorded. Capillary blood was deposited on filter papers and stored at 4°C for 1–2 months before testing. Questionnaires and samples were identified only by a number.

Laboratory assays. Samples were screened for HCV antibodies using Detect-HCV, version 3 (Adaltis). Reactive samples were further tested with the Ortho HCV 3.0 enzyme-linked immunosorbent assay (ELISA; Ortho) and Monolisa Anti-HCV Plus Version 2 (Bio-Rad): dually nonreactive samples were considered HCV negative, samples reactive with all 3 ELISAs were considered HCV positive, whereas samples with discordant results were tested by INNO-LIA HCV Score (Innogenetics), whose result was definitive unless only indeterminate bands were present, in which case polymerase chain reaction (PCR) was performed. HCV genotyping relied on amplification of a 380—base pair fragment of the NS5B gene [21].

Samples were screened for HTLV-1 antibodies using DETECT- HTLV (Adaltis). Reactive samples were further tested with Ortho HTLV-I/HTLV-II Ab-Capture ELISA (Ortho) and INNO-LIA HTLV I/II Score (Innogenetics). INNO-LIA reactive (HTLV-1 and HTLV untypable) samples were considered HTLV-1 positive; those showing only bands against HTLV-2 were considered HTLV-1 negative. Samples reactive by Ortho and nonreactive by INNO-LIA, as well as those indeterminate by INNO-LIA, were tested by PCR targeting the Tax gene [22]; PCR-positive samples were considered positive and the others negative. HIV serologic testing used Detect-HIV (Adaltis), followed by Genie II HIV-1/HIV-2 (Bio-Rad) and INNO-LIA HIV I/II Score (Innogenetics). Two participants had insufficient samples and only HCV could be tested.

Sample size estimate and data analysis. We estimated that the prevalence of HCV or HTLV-1 infection would be 6% in unexposed individuals and 12% among those treated for trypanosomiasis and that the latter would include one-third of participants. Aiming for 80% power, we needed to recruit 840 individuals. Data were analyzed with Stata statistical software, version 10.0 (Stata). Proportions were compared with the χ² or Fisher exact test. Adjusted odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated by logistic regression. Models were built sequentially, starting with the variable most strongly associated with the outcome. Models were compared using likelihood ratio tests, keeping variables significant at the P⩽.05 level.

Results

Characteristics according to ethnic groups. Table 1 compares characteristics of Mbimous to all other ethnic groups. Of 154 Mbimous living in Nola for whom we had detailed data on migration, only 54 (35%) were born in Nola; others had migrated from Mbimou land later in life, and their exposure to trypanosomiasis was similar to those who remained in their villages. Mbimous were older and more likely to be female, reflecting immigration into Nola of younger males from elsewhere. Mbimous were more likely to be widowed (data not shown); to have been treated for trypanosomiasis, schistosomiasis, or yaws; and to have received pentamidine chemoprophylaxis. No interethnic difference was seen for parenteral treatment of tuberculosis (10 participants), of leprosy (5 participants), or for transfusions (32 participants). Mbimou men were less likely to report past sexually transmitted infections. Almost all men were circumcised. Only 2 Mbimou women (1%) and 14 (8%) of other ethnicities had been excised.

Risk factors for HCV infection. The prevalence of HCV antibodies was 10.5% (95 of 905 participants). Sixty-five participants were infected with genotype 4, 7 with genotype 1, and 6 with genotype 2, whereas in 17 the virus could not be amplified. Genotype 4 so predominated that it would have been impossible to identify risk factors for other genotypes, and our main analyses were restricted to correlates of genotype 4 infection, excluding those infected with genotypes 1 and 2 or of unknown genotype. The only parenteral exposure associated with HCV genotype 4 was treatment of trypanosomiasis before 1951 (Table 2). Genotype 4 infection was not associated with treatments against yaws, schistosomiasis, tuberculosis, or leprosy or with sex, ethnicity, marital or educational status, or presence of scarifications (data not shown). None of the sexspecific factors (for men, circumcision, having paid for sex, past genital ulcer, and past urethral discharge; for women, excision, having “sold” sex, past genital ulcer, and past vaginal discharge) was associated with HCV genotype 4 (data not shown). No risk factor became significant after adjustment for treatment of trypanosomiasis before 1951.

Risk factors for HTLV-1 infection. Of 903 participants, 67 (7.4%) were considered HTLV-1 infected; 9 patients with only bands against HTLV-2 were considered HTLV-1 negative. Table 2 displays risk factors for HTLV-1 infection, more common among participants who had received ⩾2 pentamidine injections for trypanosomiasis chemoprophylaxis. HTLV-1 infection tended to be more common among those treated for trypanosomiasis before 1951 and transfusion recipients. HTLV-1 infection was not associated with past treatment against yaws, schistosomiasis, tuberculosis, or leprosy or with sex, ethnicity, marital status or educational status, presence of scarifications, or any of the sex-specific factors (data not shown).

In multivariate analysis, HTLV-1 infection was independently associated with having received ⩾2 injections of prophylactic pentamidine (adjusted OR, 2.03; 95% CI, 1.01–4.06) and with having received a transfusion (adjusted OR, 2.82; 95% CI, 1.04–7.67). None of 8 women married to a HTLV-1—infected man was herself HTLV-1 infected.

In a secondary analysis that excluded the 31 participants found on Innolia to be HTLV positive but nontypable, having received ⩾2 pentamidine injections for prophylaxis of trypanosomiasis was even more strongly associated with HTLV-1 infection (adjusted OR, 3.92; 95% CI, 1.64–9.38), but transfusion no longer was.

HIV infection. Only 12 participants (1.3%) were HIV infected (the oldest was aged 72 years), whereas 7 (0.8%) remained indeterminate. Given this low prevalence, no risk factor could be elicited.

Predicted and observed prevalence of past trypanosomiasis. Detailed data being available for Mbimou land, historical analyses focused on Mbimous. High incidence of trypanosomiasis persisted until the late 1940s (Figure 2) [9—18]. Risk (ICN) was a staggering 45% in 1936, decreased to 7% in 1942, increased to 18%–20% in 1943–1944, and decreased again, reaching <1% in 1949. The epidemic was eventually controlled through authoritarian measures to implement case-finding surveys every other month and half-yearly intramuscular prophylactic pentamidine to the whole population [9, 10]. Demographic impacts were noteworthy: the population decreased from 8679 inhabitants in 1934 to 3389 in 1951 because of death from trypanosomiasis or its treatment, low fertility, and emigration of individuals scared by the epidemic [9–11,23].

Using ICN measures of risk and previously mentioned assumptions, we calculated the predicted cumulative prevalence of past trypanosomiasis for each birth cohort to be compared to its observed prevalence among Mbimous surveyed in 2006, 56 years after the disease was brought under control. Table 3 stratifies participants according to their age in 2006. Among those aged ⩾65 years, in each stratum the proportion who reported a diagnosis of trypanosomiasis before 1951 was much lower than predicted. This was true both for those living in Mbimou land and in Nola (data not shown). Overall, for individuals ⩾65 years, we predicted that 59.3% would report a diagnosis of trypanosomiasis before 1951; only 11.1% did so. The accuracy of our model could be verified against the 1948 census; the model predicted that 37.0% of Mbimous then had a history of trypanosomiasis, whereas according to the contemporary medical report 39.1% (1690 of 4324) reported such a history [18].

Discussion

Our study attempted to indirectly delineate whether parenteral transmission played a role in the emergence of HIV-1 in Equatorial Africa several decades ago. As proxies for parenteral transmission of HIV-1, we used 2 other blood-borne viruses, HCV and HTLV-1, both compatible with a prolonged survival and more prevalent in Equatorial Africa than elsewhere, for reasons hitherto unclear. The efficacy of parenteral transmission of HCV is only twice that of HIV-1; that of HTLV-1 is unknown [24]. We selected as the study site the only area where extremely high incidence of trypanosomiasis persisted well into the 1940s, so that some patients could still be alive.

When Europeans colonized the heartland of Equatorial Africa, population displacements provoked epidemics of trypanosomiasis. In 1917–1919, of 89,743 individuals screened in Oubangui-Chari (now Central African Republic), 5347 were diagnosed as having trypanosomiasis and treated (mostly with subcutaneous drugs) using only 6 syringes [20]. In an area east of Yaoundé (Cameroon), of 194,889 inhabitants, 42% had trypanosomiasis diagnosed and were treated in 1920–1928 with subcutaneous and/or intravenous injections [19]. French colonial physicians belonged to the military: decisions about therapeutic regimens were transmitted down the chain of command. From 1928 onward, trypanosomiasis patients in Afrique Équatoriale Française were to receive 12 weekly injections of subcutaneous or intravenous orsanine or, if cerebrospinal fluid was altered, 12 weekly injections of intravenous tryparsamide, to be repeated annually for 3 years [25]. In the Nola district, treatment was centralized in the Bilolo hypnoserie, where 600 patients could be treated at any one time [9, 18]. This focus was known for a high frequency of resistance to arsenicals. Until melarsoprol was introduced (1953), many patients received combination therapies [9, 10, 18]. Transmission of blood-borne viruses might have been compounded by the use, for relapsing cases, of hétérohémothérapie, the intramuscular administration of whole blood (10 mL) from convalescent patients; in 1945, 1097 such injections were administered [13–17].

Pentamidine was used as chemoprophylaxis of sleeping sickness from 1946 until the late 1950s. Hoping to eradicate the disease, intramuscular injections were given to all inhabitants of endemic foci (except pregnant women and infants), every 6 months, for several years. Approximately 3.7 million injections were given in the Belgian Congo, 2.8 million in Afrique Équatoriale Française and 835,000 in Cameroon [8]. In Mbimou land, pentamidinization started in 1947 and continued until 1953: some received up to a dozen injections [9, 10, 12]. It was abandoned when countries became independent and resources were allocated to more pressing needs. The injections were exquisitely painful, and outbreaks of gas gangrene, including one near Nola, made the practice unpopular. Pentamidine was available in bottles of 25 g, enough for 100 injections. A contemporary description gives a flavor of the potential for transmission of blood-borne pathogens: “principles of mass production and time and motion study should be invoked to ensure the maximum speed and efficiency in getting through, say, 250 injections in a morning. The man actually giving the injection should merely have to turn half around in order to hand over his used syringe and take a freshly charged one. As he turns back again, a freshly iodined buttock, and the appropriate dose, should present themselves before him” [26, p 723–724].

HTLV-1 seems to have been transmitted during prophylactic injections of pentamidine. HTLV-1 subtype B predominates in Equatorial Africa, sharing >98% homology with the P. troglodytes simian T cell lymphotropic virus—1 [27]. HTLV-1 infects CD4 cells and is an interesting proxy for HIV-1. Equatorial Africa has high HTLV-1 prevalences, with a north-south gradient: 0%–2% in Chad, northern Central African Republic, and Cameroon to 5%–10% in forested areas of Gabon, Equatorial Guinea, and southern Cameroon [28]. Age gradients suggest a cohort effect [28, 29]. Apart from breast-feeding, modes of transmission remain unclear. Among Kinshasa prostitutes, HTLV-1 infection was not more prevalent than in other women [30]. Patients with sickle cell anemia and leprosy have a high prevalence, suggesting iatrogenic transmission. In the Mbandaka leprosarium (Figure 1), 37% of patients were HTLV-1 infected in 1969; throughout 22 years, the mortality ratio as sociated with HTLV-1 was only 1.4 [31]. Consequently, we could link HTLV-1 infections documented in 2006 to at least 1 exposure that occurred ⩾53 years earlier.

After Egypt, Equatorial Africa has the highest HCV prevalence in the world [32]. Our data suggest that HCV genotype 4 may have been transmitted during trypanosomiasis treatment in the 1930s and 1940s. Molecular clock analyses, based on our samples, revealed exponential spread from 1935 through 1965 [21]. Occasions for HCV transmission may have varied among areas, depending on the local epidemiology of tropical diseases. In Cameroon, where the HCV prevalence among elderly individuals is ⩾50%, intravenous antimalarial agents have been incriminated [33].

The main limitation is that our findings (HCV transmission during treatment of trypanosomiasis and HTLV-1 transmission during its chemoprophylaxis) are based on small numbers of participants. This was expected given the extremely long interval (>50 years) between exposures and measurement of outcomes. During this half-century, many residents of Mbimou land or Nola died for all kinds of reasons, biasing downward our measures of association. Furthermore, if within this cohort HIV-1 was iatrogenically transmitted, these individuals would have long disappeared. Thus, we measured only a fraction of the transmission of blood-borne pathogens that may have occurred during the treatment and chemoprophylaxis of trypanosomiasis. For some exposures, recall biases are likely given this long interval. However, the distribution of diseases outlined in Table 1 is congruent with contemporary medical reports [13–17].

Among Mbimous, there was a substantial difference between predicted and observed prevalence of past trypanosomiasis. Trypanosomiasis was an important event in one's life, even as a child: it seems hardly possible that a recall bias could have been of such a magnitude. Measures of risk could have been overestimated if the denominator was underestimated, but medical censuses were considered more reliable than the administrative ones. Differential migration of cases outside the study area seems unlikely. This finding may thus reflect excess mortality among individuals treated for trypanosomiasis in the 1930s and 1940s, far beyond what could be attributed to the disease or its treatment. Our model assumed that 25% of patients died of trypanosomiasis or its treatment; this was certainly not an underestimate [18]. Hepatitis B would be an unlikely culprit: this virus is universal in Central African Republic and most infections resolve. The excess mortality associated with HCV is modest in nonaddicts, and HTLV-1–associated mortality is marginal [31, 34]. By elimination, it is plausible that SIV_cpz was introduced into these cohorts and amplified exponentially by parenteral transmission among patients, causing their premature mortality a decade or so later.

Serial passage of SIV_cpz between trypanosomiasis patients or pentamidine recipients could have generated a pool of infected individuals large enough for the virus to disseminate sexually along trading routes, the tributaries of the Congo river, reaching Kinshasa/Brazzaville not later than 1959–1960 [1]. This may have occurred in other trypanosomiasis foci in southern Cameroon or Afrique Équatoriale Française, where, even if less remarkable than in Nola, incidence was high enough to prompt pentamidinization [12]. Other therapeutic regimens may have facilitated dissemination of the virus in specific locations: intravenous antimalarials, arsenicals against syphilis or yaws, streptomycin against tuberculosis, and parenteral treatments of leprosy [8].

In conclusion, our data suggest that iatrogenic transmission of HCV genotype 4 and HTLV-1 occurred during mass interventions for the control of sleeping sickness in the first half of the 20th century. These interventions, and perhaps others for the control of tropical diseases, might have jump-started the HIV-1 epidemic in regions of Equatorial Africa endemic for SIV_cpz,, from a handful to a few hundred infected individuals, a threshold beyond which sexual transmission could prosper, facilitated by urbanization and ensuing changes in behavior. Unfortunately, the mortality associated with HIV-1 infection precludes epidemiologic testing of this hypothesis.