HIV Infection and the Epidemiology of Invasive Pneumococcal Disease (IPD) in South African Adults and Older Children Prior to the Introduction of a Pneumococcal Conjugate Vaccine (PCV)

Introduction Streptococcus pneumoniae is the commonest cause of bacteremic pneumonia among HIV-infected persons. As more countries with high HIV prevalence are implementing infant pneumococcal conjugate vaccine (PCV) programs, we aimed to describe the baseline clinical characteristics of adult invasive pneumococcal disease (IPD) in the pre-PCV era in South Africa in order to interpret potential indirect effects following vaccine use. Methods National, active, laboratory-based surveillance for IPD was conducted in South Africa from 1 January 2003 through 31 December 2008. At 25 enhanced surveillance (ES) hospital sites, clinical data, including HIV serostatus, were collected from IPD patients ≥ 5 years of age. We compared the clinical characteristics of individuals with IPD in those HIV-infected and -uninfected using multivariable analysis. PCV was introduced into the routine South African Expanded Program on Immunization (EPI) in 2009. Results In South Africa, from 2003–2008, 17 604 cases of IPD occurred amongst persons ≥ 5 years of age, with an average incidence of 7 cases per 100 000 person-years. Against a national HIV-prevalence of 18%, 89% (4190/4734) of IPD patients from ES sites were HIV-infected. IPD incidence in HIV-infected individuals is 43 times higher than in HIV-uninfected persons (52 per 100 000 vs. 1.2 per 100 000), with a peak in the HIV-infected elderly population of 237 per 100 000 persons. Most HIV-infected individuals presented with bacteremia (74%, 3 091/4 190). HIV-uninfected individuals were older; and had more chronic conditions (excluding HIV) than HIV-infected persons (39% (210/544) vs. 19% (790/4190), p<0.001). During the pre-PCV immunization era in South Africa, 71% of serotypes amongst HIV-infected persons were covered by PCV13 vs. 73% amongst HIV-uninfected persons, p = 0.4, OR 0.9 (CI 0.7–1.1). Conclusion Seventy to eighty-five percent of adult IPD in the pre-PCV era were vaccine serotypes and 93% of cases had recognized risk factors (including HIV-infection) for pneumococcal vaccination. These data describe the epidemiology of IPD amongst HIV-infected and -uninfected adults during the pre-PCV era and provide a robust baseline to calculate the indirect effect of PCV in future studies.


Introduction
Streptococcus pneumoniae is the most common cause of bacteremia and bacterial pneumonia among HIV-infected persons, and has been shown to occur between 30 and 100 times more frequently in HIV-infected than HIV-uninfected individuals. [1][2][3][4] Recurrent disease is seen more frequently with HIV infection, with approximately 25% of patients having a recurrence within 12 months. [1,5,6] This study was conducted in South Africa, which reports a HIV prevalence of 18% amongst adults in 2008. [7] In population groups with a high HIV prevalence the epidemiological pattern of invasive pneumococcal disease (IPD) seems to change, with a peak seen in the young adult population as well as the typical but more marked peaks seen in the infant and elderly population groups. [4,6,8] Despite the higher rate of IPD in HIV-infected individuals the majority of studies have not shown higher case-fatality ratios amongst people co-infected with HIV. [1,3,6] Some differences in IPD epidemiology have been described in HIV-infected versus HIVuninfected individuals. Firstly, HIV-infected adults tend to be infected more commonly with pneumococcal serotypes that are typically seen in young children, and in relation to this, HIVinfected people tend to have a higher proportion of penicillin non-susceptible isolates. [3,[9][10][11][12] In contrast, serotype 1 disease tends to be proportionally less common amongst HIV-infected than HIV-uninfected persons. [3,4,13] Besides HIV as a stand-alone risk factor, HIV-infected individuals with IPD tend to have less underlying disease than HIV-uninfected individuals. [14] HIV Infection and Adult Pneumococcal Disease in the Prevaccine Era PLOS ONE | DOI: 10 Competing Interests: The authors have read the journal's policy and the authors of this manuscript have the following competing interests: Anne von Gottberg has received research funding from Pfizer, however this does not alter the authors' adherence to PLOS ONE policies on sharing data and materials. All other authors have no conflict of interest.
Use of the polysaccharide pneumococcal vaccine (PPV23) in HIV-infected African adults is controversial. [5] However, studies using the 7-valent pneumococcal conjugate vaccine (PCV7) are promising, showing a 74% vaccine efficacy against vaccine serotypes in HIV-infected adults with a previous episode of IPD, albeit with subsequent waning of immunity. [15] Some countries, including South Africa, have already shown an indirect effect in the adult population following infant PCV immunization. [16][17][18] In South Africa and the United States of America, this decrease was seen in both HIV-infected (-25 to -59% for vaccine serotypes attributed to childhood PCV immunization) and HIV-uninfected adults (-52 to -86% for vaccine serotypes). [18,19] Reporting baseline clinical characteristics of IPD amongst adults in a setting with high HIV prevalence is important to elicit the indirect and long-term effects, including serotype replacement, of PCV introduction on IPD in adult populations. In this analysis we aimed to describe the baseline epidemiology of IPD in the pre-PCV era amongst HIV-infected and -uninfected persons 5 years of age using data derived from national surveillance.

Methods
GERMS-SA conducts national, active laboratory-based surveillance across South Africa in a network consisting of approximately 130 public and private microbiology laboratories. Each laboratory is responsible for sending S. pneumoniae isolates, along with demographic details of the patient to a reference laboratory (the Centre for Respiratory Diseases and Meningitis (CRDM) at the National Institute for Communicable Diseases (NICD)) where confirmatory tests, serotyping and antimicrobial susceptibility testing are performed. Data from individuals diagnosed with IPD from 1 January 2003 through to 31 December 2008 were included in this analysis. From 2003 to 2005 our surveillance network strengthened and remained relatively stable thereafter. For this reason trends in incidence by serotype were only examined from 2005 onwards. Comprehensive audits of cases identified but not reported were conducted for all the public laboratories: unreported cases ranged from 16% to 21% over the 6 years and their demographic data were included in our database. [20] PCV7 was introduced into the South African EPI in 2009, and replaced by PCV13 in 2011. There was minimal use of either PCV7 or PPV23 during the study period.
IPD was defined as identification of S. pneumoniae from any usually sterile site (cerebrospinal fluid (CSF), blood, pleural fluid, joint fluid, ascitic fluid, vitreous fluid, etc.). Multiple isolates of S. pneumoniae identified from the same person were counted as one case provided they occurred within 21 days of the first isolate. S. pneumoniae were identified by either: culture of the organism, consistent Gram stain plus detection of S. pneumoniae antigen by latex agglutination, or detection of S. pneumoniae by direct PCR (polymerase chain reaction) of the specimen. Antimicrobial disc diffusion susceptibility testing of potentially resistant strains was confirmed by broth microdilution. Antimicrobial susceptibility was reported as susceptible or non-susceptible (intermediate and resistant isolates) based on the Clinical and Laboratory Standards Institute (CLSI) interpretive criteria. [21] Serotyping of viable organisms were determined using the Quellung reaction (Statens Serum Institut, Copenhagen, Denmark). Serotypes were grouped according to those covered by PCV13 (1, 3, 4, 5, 6A, 6B, 7F, 9V, 14, 18C, 19A, 19F and 23F) and PPV23 vaccines (1, 2, 3, 4, 5, 6B, 7F, 8, 9N, 9V, 10A, 11A, 12F, 14, 15B, 17F, 18C, 19A, 19F, 20, 22F, 23F, 33F and serotype 6A cross-protection was assumed). When calculating incidence of individual serotypes we assumed the age-specific proportion of serotypes with non-viable or missing pneumococcal isolates to be that of those with available isolates and imputed serotype data for 5 041 cases.
Enhanced surveillance (ES) was conducted at 25 sentinel-hospital sites situated in all 9 provinces of South Africa. Where available, consenting persons with IPD (or their legal guardians) were interviewed using a standardized questionnaire. Clinical details were retrieved from medical record review performed on cases of IPD at these hospitals. As clinical details were only available from patients presenting to ES sites, a comparison of cases from enhanced versus non-enhanced surveillance sites was done to show representativeness of data from the ES sites.
The following demographic and clinical details were collected from each person under surveillance: age/ date of birth, sex and specimen type (CSF, blood, other for any other specimen from a sterile site). Additional details were collected at ES sites: clinical syndrome (meningitis, bacteremic pneumonia, bacteremia without focus, other), Pitt bacteremia score for severity of illness (score 0 for mild, 1-3 for moderate, 4-12 for severe illness), [22] HIV status, in-hospital outcome and any underlying condition predisposing to IPD. Predisposing conditions were further classified into ACIP (Advisory Committee on Immunization Practices) conditions (asplenia/ sickle cell disease; chronic lung, renal, liver or cardiac condition; diabetes; solid organ transplant; immunotherapy; malignancy; primary immunodeficiency; CSF leak following a head injury; alcohol dependency; tobacco use) and other risk factors (cerebrovascular accident; burns). [23] Demographic, laboratory and clinical details for each patient were linked using Epi Info, version 6.04d, and later updated to Access.
Incidence of IPD was calculated using South African population denominators from Statistics-South Africa. [24] HIV-specific incidence rates for South Africa were calculated by assuming that the proportion of patients who tested positive for HIV at sentinel sites was the same in those who were not tested. HIV population denominators for 2008 were obtained from the Thembisa model 1.7 version 3. [25] Statistical analysis was done using Stata version 11 (StataCorp Inc., College Station, Texas, USA) and p-values of 0.05 were considered significant throughout. Trends in incidence rates by age-category and serotype were calculated using Poisson regression. Univariate analysis of characteristics associated with IPD among HIV-infected and HIV-uninfected persons was performed using Fisher's exact test or the Mantel-Haenszel χ 2 -test for categorical variables. Multivariable logistic regression models were evaluated comparing IPD cases at enhanced and nonenhanced surveillance sites, and comparing IPD cases amongst those HIV-infected and-uninfected. We started with all variables that were significant at p-value less than 0.05 on univariate analysis, and dropped non-significant factors with stepwise backward selection. All two-way interactions were evaluated.
Ethical clearance and permission to conduct laboratory-based and enhanced surveillance in South Africa for this study was obtained from the Health Research Ethics Committee (Human), University of Witwatersrand (Clearance number M02-10-42); the University of Stellenbosch Health Research Ethics Committee (Reference number N04/01/0021), the National Institute for Communicable Diseases Research Committee (Clearance number M060449); and the South African Department of Health (Reference H2/12/8). The following consent procedures were approved by the above mentioned ethics committees: Written consent was obtained from all patients (or parents/legal guardians of minors) at ES sites, prior to being interviewed, for participation in the ES program. Patients from ES sites who had already been discharged from hospital were interviewed telephonically and gave verbal consent, documented by surveillance officers on the written consent forms. Parents/ legal guardians of minors gave written consent (or verbal if the child had already been discharged from hospital) on behalf of their children <18 years of age. Patients from ES sites who were unable to provide written or verbal consent (due to severity of illness, death or lost to follow up) were still included in the study and their clinical data were collected only through medical record review. Participants not consenting to answer the enhanced surveillance questionnaire were still included in the laboratory-based surveillance program. All patient identifiers were removed prior to data analysis.

Comparison between HIV-infected vs. HIV-uninfected individuals at ES sites
Eighty-nine percent (4 190/4 734) of people with IPD from ESS with HIV data available were HIV-infected and 37% (1 568/4 190) of these HIV-infected patients were newly diagnosed during this episode of IPD. On multivariable analysis, age between 20 and 49 years, female sex, bacteremic pneumonia and a cotrimoxazole non-susceptible isolate were significantly more common among HIVinfected than HIV-uninfected IPD patients. HIV-uninfected patients had better survival to discharge, and more underlying conditions (apart from HIV infection) predisposing them to IPD than HIV-infected patients. There was no difference between HIV-infected and-uninfected adults regarding proportion of vaccine-preventable serotypes causing IPD (71% serotypes amongst HIV-infected persons covered by PCV13 vs. 73% amongst HIV-uninfected persons, p = 0.4, OR 0.9 (CI 0.7-1.1) and 84% serotypes amongst HIV-infected persons covered by PPV23 vs. 86% amongst HIV-uninfected persons, p = 0.3, OR 0.9 (CI 0.6-1.1)). (Table 3)

Discussion
In South Africa during the pre-vaccine era, 70% of IPD in adults and older children was caused by serotypes in PCV13 and 85% by PPV23 serotypes. IPD incidence in HIV-infected individuals was 43 times higher than in HIV-uninfected persons (52 per 100 000 vs.1 per 100 000), with HIV-infected adults 65 years at highest risk (IPD incidence of 237 per 100 000, 95% CI 171-321). IPD and HIV-coinfection was associated with more bacteremia (with or without pneumonia) and higher in-hospital case fatality than HIV-uninfected cases.
HIV-infection is a major risk factor for IPD, with 90% of IPD amongst South African older children and adults occurring in the 18% of South Africans who are HIV infected. [7,24] By 2008, HAART coverage amongst HIV-infected South African adults eligible for treatment was 40% [26], this has increased to 79% in recent years. [27] Yet in South Africa, one study showed only a 7% decrease in adult IPD attributable to the increased HAART coverage during this same period. [18] Although HAART reduces the risk of IPD in HIV-infected persons, there   seems to be an incomplete restoration of the immune system, necessitating additional measures for preventing IPD in this high risk group. [28][29][30][31][32][33][34][35][36] Sex, age and underlying medical conditions are established risk factors for IPD and its more severe outcomes. [37][38][39] Globally IPD occurs more frequently in males than females however we found that in persons with IPD HIV-infection was associated with female sex. In South Africa, HIV prevalence in females is higher (11.9% in females versus 9.2% in males for 2008) which may in part account for the disparity. [7,40] From age 65 years, IPD incidence in both HIV-infected and uninfected persons increases exponentially, however in this age category the incidence amongst HIV-infected persons is 58 times higher than in HIV-uninfected individuals (237 vs. 4 cases per 100 000 population respectively). With the rollout of ARVs, this high-risk category of elderly HIV-infected persons is likely to grow. Similarly, there is a peak of disease in HIV-infected teenagers, which is in stark contrast to their HIV-uninfected peers (81 vs. 0.6 cases per 100 000 population respectively, relative risk of 127). Many of these HIV-infected teens are potentially long term survivors of vertical HIV-transmission, with potential added risk factor of chronic lung disease, whilst a smaller proportion may be recently infected. [41] The HIV-infected teens and elderly are at an even higher risk with the potential development of chronic diseases which further predispose to IPD.
The highest burden of IPD in South African persons five years and older is in the young to middle-aged adult population. The majority (88%) of disease occurred amongst HIV-infected persons; however, 43% of HIV-uninfected persons had other underlying conditions predisposing them to IPD. In our study, as seen in other studies, vaccine serotypes occur at high rates in older children and adults, with 70% of disease caused by PCV13 serotypes and 85% by PPV23 serotypes. [3,6,42,43] This was similar for both HIV-infected and uninfected individuals, despite the individual serotype diversity seen within each group. Although South Africa follows recommendations to vaccinate high risk individuals with PPV23 (including HIV-infected individuals), vaccine uptake is extremely low and no patient in our study had received PPV23. [44,45] In South Africa, we recently demonstrated an indirect effect on all serotype IPD incidence, following PCV7 vaccination of infants, of 34% in the 25-44 year age category. [18] However, this indirect effect was less evident amongst those in their early teens and in the elderly (>65 years) (6% reduction in 10-14 years and 14% reduction in 65+ age group)-two groups with high incidences of IPD amongst those HIV-infected. Since 2011, PCV13 has been used in the EPI, and it is expected that the indirect effects of PCV13 on additional serotypes will further decrease the incidence of IPD in older children and adults. However, in HIV-infected populations additional preventative interventions could be considered such as direct vaccination of selected groups with pneumococcal vaccine, increased uptake of the seasonal influenza vaccine and increased use of ART. [46][47][48][49] Previous studies in South Africa reported that HIV-coinfected persons were more likely to be infected with penicillin non-susceptible pneumococci, however when including cotrimoxazole in our multivariate model this finding fell away. Cotrimoxazole non-susceptibility remained a significant finding in HIV-coinfected patients on multivariable analysis; this may be due to the high use of cotrimoxazole prophylaxis in preventing opportunistic infections in HIV-infected patients in South Africa. [50] One important limitation of surveillance studies is that IPD incidence may be underestimated. [51] Our case definition relies on the specimen taking practices of clinicians and the health seeking behavior of the patients, which have both been found to vary greatly across our study population. [52,53] In many of the rural areas reporting of blood culture results are often delayed, therefore if patients do seek medical care, clinicians often treat those with a suspected bacteremia empirically without laboratory investigations to confirm their diagnosis. These cases would not have been reflected in this study and are therefore excluded from our incidence calculations. A study during the same time period confined to a single site in South Africa with a relatively well-defined population and adequate blood culturing practices, reported IPD incidence in adults of 58 per 100 000 population. [28] Therefore the true incidence of IPD in South African adults and older children likely lies above the 7 per 100 000 reflected here. An assumption was made for 29% (5 041/17 604) of IPD isolates that distribution of serotypes by age-category and province was similar amongst isolates that were available and those that were missing when calculating incidence by serotype. In addition, these missing isolates could influence the distribution of antimicrobial susceptibility data. HIV testing was not done routinely for surveillance purposes and patient HIV serostatus was only available at ES sites-and only for 68% of the patients with medical records available. Only a few parameters were available for comparing cases from ES versus non-enhanced sites, and although the main differences were found in specimen type and cotrimoxazole non-susceptibility there may be more unmeasured differences which we were unable to account for which could have influenced our model comparing HIV-infected to HIV-uninfected patients with IPD. ASSA2003 denominators for HIV-infected individuals tended to underestimate the number of HIV-infected individuals in the teenage and elderly populations, therefore we used a modified version of the ASSA2003 model, Thembisa 1.7v3, which accounts for the survival benefits of the rollout of HAART especially amongst the teenagers and elderly. [25,54] This gave more plausible results for IPD incidence amongst the teenager and elderly HIV-infected population but may still be an underestimate of true IPD incidence.
In summary, 70-85% of adult IPD in the pre-PCV era were vaccine serotypes and 93% of cases had recognized risk factors (including HIV-infection) for pneumococcal vaccination. These data describe the epidemiology of IPD amongst HIV-infected and -uninfected adults during the pre-PCV era and provide a robust baseline to calculate the indirect effect of PCV in future studies.