Like most studies of bacteremia in older children and adults in Africa, pneumococcus was the most common pathogen identified, accounting for 58% of isolates [1, 4–9]. Because our study was population-based with active surveillance in the community, we could adjust crude rates to better assess actual rates of pneumococcal bacteremia, which was almost double that measured in the clinic. Moreover, home-based HIV testing in the community allowed calculation of rates of pneumococcal bacteremia in HIV-infected persons, which showed that almost one in four HIV-infected adults likely suffer pneumococcal bacteremia each year. The extrapolated rate of pneumococcal bacteremia among HIV-positive persons in our population (2,399 per 100,000 person-years) rivaled the crude rates among prospective cohorts of South African miners (2,136 per 100,000 person-years) and HIV-infected commercial sex workers in Nairobi (2,380 per 100,000 person-years) and HIV-infected adults in a trial of pneumococcal polysaccharide vaccine in Uganda (1,200 per 100,000 person-years), and was higher than the crude rate observed in urban South Africa (197 per 100,000 persons) [6, 12–14]. In rural western Kenya, HIV infection seems to drive the high overall rate of pneumococcal bacteremia among adults, being 19.7 times more common among HIV-positive than HIV-negative adults; this elevated rate ratio is similar to South Africa (rate ratios of 6-55) and urban Kenya (rate ratio 17.8) [6, 7, 13–15].
The majority (79%) of pneumococcal bacteremic episodes occurred in patients with SARI with a predominance among those hospitalized. Only 5 (10%) of pneumococcal bacteremia patients did not report any respiratory symptoms. However, culturing only hospitalized patients would have missed 31% of pneumococcal bacteremia. Studies of invasive pneumococcal disease in Africa have yielded case-fatality proportions averaging 14-16%; the lower case-fatality proportion we found (7.7%) likely results from inclusion of outpatients, as well as not collecting cerebrospinal fluid on higher mortality meningitis cases [6, 7, 9, 12, 14, 24]. The serotype distribution in western Kenyan is similar to that found in other adult populations in Africa [14, 24–31]. Serotype 1 is uniformly the most common serotype found among African adults, ranging from 18-36% . Serotype 1 predominated in large outbreaks of pneumococcal disease in the pre-antibiotic era and has more recently caused outbreaks of meningitis in West Africa [32, 33]. We did not find evidence for serotype 1 outbreaks in our population as no temporal or spatial clustering of cases occurred. The increased prevalence of serotype 1 in HIV-negative compared with HIV-positive individuals, and the converse for the so-called pediatric serotypes in PCV7, has been shown before [14, 15, 26, 31]. It has been postulated that HIV-infected adults, like young children, have reduced immunity to pneumococcal colonization, leading to increases in disease due to pediatric serotypes that commonly colonize the nasopharynx . Indeed, nasopharyngeal colonization among HIV-infected adults has been shown to be elevated, with rates above 30% in Kenya and Zambia [34, 35]. The high proportion of pneumococci resistant to cotrimoxazole has been shown before, and likely reflects widespread use of cotrimoxazole in the community and use of sulfadoxine-pyrimethamine for malaria treatment in Kenya . Beta-lactam resistance remains low in Kenya and likely reflects the fact that apart from amoxicillin, few beta-lactam antibiotics are used outside of the inpatient setting.
In Kenya, several large-scale interventions that are being implemented could lead to a decrease in adult pneumococcal disease. First, until recently most Kenyans did not know their HIV status. The 2007 Kenya AIDS Indicator Survey showed two-thirds of Kenyans did not know their HIV status and 65% of HIV-infected persons who needed HAART were not receiving it . However, HIV testing is now becoming more widespread in Kenya and HAART is increasingly available for those who need it due to large multinational funding initiatives. HAART has been shown to decrease invasive pneumococcal disease, as well as bacterial pneumonia, in HIV-infected persons [7, 36]. Second, Kenya plans to introduce PCV10 (Synflorix™, GlaxoSmithKline, Rixensart, Belgium) for children in 2010 with support from the Global Alliance for Vaccines and Immunizations. In the U.S., a large indirect effect of childhood PCV7 (Prevnar™, Wyeth Pharmaceuticals, New Jersey, U.S.) vaccination was realized among adults, with a significant decrease in invasive pneumococcal disease of 32% among persons 20-39 years, the age group likely to be parents of small children and also that had the highest rate of pneumococcal bacteremia in our study . After PCV7 was introduced in the U.S., HIV-infected adults showed a 62% decrease in pneumococcal disease due to vaccine serotypes . PCV10 will include serotype 1, the most common serotype among Kenyan adults. Unlike most serotypes included in PCV7, serotype 1 is not often found in the nasopharynges of children . Because the indirect impact of conjugate vaccines work through reduction of carriage, and thus transmission to other persons, whether the same degree of herd immunity in Kenyan adults will be realized for serotype 1 is unclear.
Our study had several limitations. First, we did not have HIV status on all patients and assuming the same HIV prevalence among those not tested might have been incorrect. In particular, patients who died before the home-based HIV testing initiative began might have been more likely to be HIV-positive, which would have resulted in even higher extrapolated rates of pneumococcal bacteremia in HIV-infected persons than we calculated. Due to the small number of patients with pneumococcal bacteremia who had HIV testing in our study, the HIV-specific results should be interpreted with some caution. Moreover, our projection of the national number of pneumococcal bacteremia cases among adults assumed that no persons ≥65 years old were HIV-positive, so likely underestimated the true number of cases. Second, in making extrapolations of bacteremia rates, we made some assumptions that could have biased our adjusted rates. We assumed isolation rates would be the same among those with the same syndromes who did not get blood cultures. At Lwak, cultures might have been performed more frequently among patients more likely to have pneumococcal bacteremia (e.g. sicker patients). Also, persons in the community might have been more likely to go to Lwak than other clinics if they were sicker, knowing that admission was free at Lwak. Third, we only cultured the blood of the first two febrile adults, which possibly could have led to a systematic bias if the first two patients were more or less likely to have bacteremia than other febrile patients. Moreover, by culturing the majority of SARI patients, while only a sample of febrile patients, we likely biased our results towards finding more pneumococcus and less mycobacteria and non-typhi salmonellae, which have also been shown to be prevalent, sometimes more so than pneumococcus, among febrile, HIV-infected persons admitted to hospital [4, 9–11]. Lastly, because we did not do lumbar punctures, our analysis did not include pneumococcal meningitis, the other major manifestation of invasive pneumococcal disease.