HIV and Influenza Virus Infections Are
Associated With Increased Blood Pneumococcal
Load: A Prospective, Hospital-Based
Observational Study in South Africa, 2009–2011
Nicole Wolter,1,2 Cheryl Cohen,1,2 Stefano Tempia,1,8 Shabir A. Madhi,1,4 Marietjie Venter,1,5 Jocelyn Moyes,1
Sibongile Walaza,1 Babatyi Malope Kgokong,1 Michelle Groome,4 Mignon du Plessis,1,2 Marthi Pretorius,1
Halima Dawood,6 Kathleen Kahn,3,10,11 Ebrahim Variava,7 Keith P. Klugman,1,9 and Anne von Gottberg1,2
Centre for Respiratory Diseases and Meningitis, National Institute for Communicable Diseases, National Health Laboratory Service, 2Faculty of Health
Sciences and 3MRC/Wits Rural Public Health and Health Transition Research Unit (Agincourt), Faculty of Health Sciences, University of the

Downloaded from by guest on April 17, 2015
Witwatersrand, and 4Department of Science and Technology/National Research Foundation: Vaccine-Preventable Diseases, Johannesburg, 5Department
of Medical Virology, University of Pretoria, Pretoria, 6Edendale Hospital, Pietermaritzburg Metropolitan Hospitals, Pietermaritzburg, and 7Klerksdorp-
Tshepong Hospital, Klerksdorp, South Africa; 8Influenza Division, Centers for Disease Control and Prevention, and 9Hubert Department of Global Health,
Rollins School of Public Health, and Division of Infectious Diseases, School of Medicine, Emory University, Atlanta, Georgia; 10Center for Global Health
Research, Umeå University, Umeå, Sweden; and 11INDEPTH Network, Accra, Ghana

Background. Increased pneumococcal loads are associated with severe outcomes. We determined the preva-
lence of pneumococcal DNA in blood specimens from patients hospitalized with acute lower respiratory tract infec-
tion and identified factors associated with invasive pneumococcal pneumonia, bacterial loads, and death.
Methods. A total of 8523 patients were enrolled as part of prospective hospital-based surveillance. Blood was
collected for quantitative pneumococcal (lytA) detection, and nasopharyngeal specimens were collected for detec-
tion of influenza virus and other respiratory viruses by real-time polymerase chain reaction.
Results. Of 6396 cases (75%) with lytA results, 422 (7%) were positive for pneumococcal DNA. The prevalences
of human immunodeficiency virus (HIV) and influenza virus were 51% (2965/5855) and 8% (485/6358), respective-
ly. On multivariable analysis, HIV infection (adjusted odds ratio [aOR], 2.4; 95% confidence interval [CI], 1.6–3.6),
influenza virus coinfection (aOR, 1.4; 95% CI, 1.2–2.1), oxygen therapy during admission (aOR, 1.6; 95% CI, 1.1–
2.3) and in-hospital death (aOR, 2.1; 95% CI, 1.1–4.0) were significantly associated with increased pneumococcal
load. Among lytA-positive patients, after adjustment for length of hospitalization, duration of symptoms, and
oxygen therapy during admission, pneumococcal loads ≥10,000 DNA copies/mL (aOR, 3.6; 95% CI, 1.8–7.2) were
associated with increased risk of death.
Conclusions. HIV and influenza virus infections were associated with elevated pneumococcal loads, which, in
turn, were associated with increased risk of death.
Keywords. Streptococcus pneumoniae; pneumococcal pneumonia; bacterial load; HIV; influenza.

In 2004, lower respiratory tract infections were the for approximately 4.2 million deaths [1]. In South
third most common cause of death globally, accounting Africa, in 2008, influenza and pneumonia ranked the
second most common natural cause of death in all ages,
accounting for 8% of deaths [2]. Streptococcus pneumo-
Received 25 April 2013; accepted 12 July 2013; electronically published 6 niae is a common, often predominant cause of bacterial
August 2013.
Correspondence: Nicole Wolter, PhD, Centre for Respiratory Diseases and Men- pneumonia [3–5].
ingitis, National Institute for Communicable Diseases, Private Bag X4, Sandring- The etiological diagnosis of community-acquired
ham, 2131, Gauteng, South Africa (
pneumonia is difficult to elucidate because of inade-
The Journal of Infectious Diseases 2014;209:56–65
© The Author 2013. Published by Oxford University Press on behalf of the Infectious quate diagnostic tests. As a result, the incidence of
Diseases Society of America. All rights reserved. For Permissions, please e-mail: pneumococcal pneumonia is underestimated [6]. Real-
DOI: 10.1093/infdis/jit427
time polymerase chain reaction (PCR) offers a fast and

56 • JID 2014:209 (1 January) • Wolter et al

We used quantitative real-time PCR to determine the ryngeal and nasopharyngeal swabs (Dryswab. reverse primer. using the million people [24]. CA). 20] and are considered a risk factor for pneumococcal pneu- monia [21. and results were retrieved from the patient’s hos- sociated with increased rates of invasive pneumococcal disease pital records. by guest on April 17. and 2. 22]. influenza virus circulation occurs mainly Roche MagNA Pure LC 2. Swabs collected from the same patient were placed in the same virus transport medium. South Africa) upon collection. A case in individuals aged ≥5 years included any hospi- tially useful diagnostic tool [7–11]. [28]. pneumoniae in the blood specimen. 2009 and continued through April 2011. 16]. Johan- cal loads. A case in children aged 2 days to <3 months included from a known quantity (confirmed spectrophotometrically) of any hospitalized patient with physician-diagnosed sepsis or S. Corsham. placed in virus transport medium (Highveld Biological. Samples with a cycle zation’s (WHO’s) acute lower respiratory tract infection case threshold (Ct) value of >40 were recorded as negative. approximately III for bacteria (Roche. A case in children aged 3 lytA-negative cases. Blood cultures were performed at the discretion of mortality [17. resulting in severe disease and increased years of age. pneumoniae ATCC49619 and was used to calculate pneumo- physician-diagnosed acute lower respiratory tract infection. and 40 cycles of 95°C Patients were enrolled if they met the World Health Organi. pneumoniae In South Africa. using a single-target (lytA) quantitative real-time PCR assay scribe the etiology of and risk factors for acute lower respiratory adapted from Carvalho et al. A stan- definition and had symptom onset within 7 days from admis. markers of community-acquired pneumonia [8. [19. For this study. pneumoniae started in May osystems. and pleural ef- of pneumococcal DNA in blood has been shown to be a poten. including oropha- loads. blood culture. The 25-μL PCR reaction tract infection in South Africa.5 μL of extracted mented at 6 public referral hospitals at 4 surveillance sites in 4 DNA. England) for patients aged ≥5 years with acute lower respiratory tract infection and to identify factors or nasopharyngeal aspirates for children aged <5 years. We conducted a prospective hospital-based observational A case of invasive pneumococcal pneumonia was defined as study as part of the Severe Acute Respiratory Illness surveil. the identification of S. and is physician-diagnosed acute lower respiratory tract infection. fusion. In 2006. blood specimen contained 1× TaqMan gene expression mastermix (Applied Bi- collection for the detection of S. for 15 seconds and 60°C for 1 minute. The study was imple. coccal loads (DNA copies/mL).oxfordjournals. 2015 tality [9. 95°C for 10 minutes. The following universal cycling conditions were used: provinces of South Africa. nesburg. HIV and Influenza Virus Infections Increase Pneumococcal Load • JID 2014:209 (1 January) • 57 .sensitive alternative to the gold standard. ir. pneumonia. Mannheim. 9. 14]. 100 μL of elution buffer and stored at −20°C. Respiratory specimens. For blood-culture positive. Extracted DNA was eluted into years) was infected with HIV [26]. bronchiolitis. which may serve as useful ings of pneumonia. there are few data on the effect of HIV traacetic acid–containing vacutainer tubes within 24 hours of infection and respiratory virus coinfections on pneumococcal hospital admission. 12]. Infection fied from the patient’s Road to Health card. Foster City. which was available with influenza virus is commonly complicated by secondary for review on admission in 69% of children (1860/2691) <5 pneumococcal infection. and PCR compared favor. lance program that started in February 2009 and aims to de.0 instrument and DNA Isolation kit during winter (May to August) [25]. Pneumococcal loads in clinical specimens have Sample Collection and Processing been shown to be directly related to severity of illness and mor. However. Study Design and Setting Detection and Quantification of S. Respiratory viral infections are specifically as. 50°C for 2 minutes. Vaccination history was veri- risk factor for invasive pneumococcal disease [15. 18]. 23]. the clinician. dard curve was prepared using serially diluted DNA extracts sion [27]. In addition. 13. Whole blood samples were collected in ethylenediaminete- Downloaded from http://jid. and death. pneumococ. respective of signs and symptoms. A standardized questionnaire was used to collect demo- Human immunodeficiency virus (HIV) infection is a known graphic and clinical information. a low to middle income country of about 50 DNA was extracted from 200 μL of whole blood. Specimens were stored at 4°C and transported to the MATERIALS AND METHODS National Institute for Communicable Diseases for testing. blood specimens were reextracted and PCR months to <5 years included any hospitalized patient with repeated. The detection including bronchitis. Germany) according to the 11% of the South African population (4% of children aged <5 manufacturer’s instructions. 200 nM each of forward primer. real-time PCR has been used to difficult breathing with or without clinical or radiographic find- quantify bacterial loads in blood. and probe (5′FAM). not dependent on the viability of the organism. talized patient presenting with manifestation of acute lower ably with culture and serological tests when used to diagnose respiratory tract infection with recent (≤7 days) onset of fever pneumococcal infection in children with lower respiratory tract (>38°C) and cough or sore throat and shortness of breath or infections [10. Medical Wire prevalence of pneumococcal DNA among patients hospitalized and Equipment. were associated with invasive pneumococcal pneumonia.

153/1725 [9%]) and spring (September to No- ≥100 000 DNA copies/mL. statistical analysis was implemented using Stata. Patients’ age ranged between 2 Statistical Analysis days and 92 years (median age. Ethics Approval Determination of HIV Status The protocol was approved by the University of the Witwaters- HIV results were obtained from a combination of 2 sources: (1) rand Human Research Ethics Committee (M081042) and the patient clinical records. The coccal cases positive on blood culture only were excluded from influenza virus A–positive specimens were subtyped using the analysis because no bacterial load data were available. 52%) and patients with clinical records (1348/2674. 10 000. and 2965 (51%) load. which group (9%. Of these. and parainfluenza virus types 1–3. assessed for significance in the multivariable analysis. 165 patients. 1000 to <10 000. 1000. 2015 Demographic and Epidemiologic Characteristics of the Study the ELISA was reactive. (StataCorp. Pairwise interactions ing viruses: influenza virus types A and B. and (3) mortality among lytA-positive was obtained from 474 positive samples (98%). 100. Table 1). we compared the prevalence (and asso. human metapneumovirus. version 11 Centers for Disease Control and Prevention. investigate the effect of different pneumococcal loads on mor. respiratory remaining in the final multivariable additive models. lowed by influenza A(H1N1)pdm09. The lytA positivity and mortality models were imple. 24 years).Detection of Respiratory Viruses Covariates with a P value <. as follows: with the circulation of influenza virus (Figure 1) during winter <100. 2–5 years. A total of 6358 of to identify factors associated with (1) lytA positivity among 6396 cases (99%) had available influenza results. 154 (33%) were influenza A(H3N2). fol- lowest bacterial load category (<100 DNA copies/mL). and 100 000 ported having received antibiotics in the 24 hours before hospi- DNA copies/mL) and included in the model as a predictor to tal admission. (75%). TX). and stat- tion real-time PCR as previously described [29] for the follow. rhinovirus. we enrolled 8523 cases 5636. (34%) were 2009 pandemic influenza A(H1N1) (influenza A mented using stepwise backward selection logistic regression. The outcome variable However. HIV prevalence was 13% (243/ outcomes or characteristics of the primary end points of the 1943). and ( June to August. compared with the log odds of being at or below the spec. adenovirus. Testing included HIV enzyme-linked immunosorbent assay (ELISA) for patients aged RESULTS ≥18 months and PCR testing for children aged <18 months if Downloaded from http://jid. Cases of invasive pneumococcal assumes that the ordinal outcome variable represents categories pneumonia were detected throughout the year (Figure 1). tality. an anonymized linked dried blood spot tested at the mittee (BF157/08). pneumoniae results were obtained from 6396 patients revealed no difference (P = . istical significance was assessed at P < .201). For a large number of patients for whom both results were available. [H1N1]pdm09). 10 000 to <100 000. and ≥45 years. 54% (95/175). Univariate comparisons were performed using logistic (646/1125) among those aged <2 years. 4 binary variables were generated for (33%) were influenza B. compared with summer (December al-odds model (and associated odds ratio [OR]) measure the to February. we implemented 3 multivariable models 19–44 years. a 58 • JID 2014:209 (1 January) • Wolter et al . 86% (1884/2181).org/ by guest on April 17. The coefficients of the proportion. 50%) and S. were assessed by the inclusion of product terms for all variables enterovirus. duction of influenza A(H1N1)pdm09 in the country. and death. where available. regression. NICD. 6–18 years. In the same year. and (2) for consenting University of KwaZulu Natal Biomedical Research Ethics Com- patients. (2) pneumococcal load among were positive for influenza virus. following the intro- level. Of these. 100 to <1000. An influenza virus subtype lytA-positive patients. the NICD result was used.05. respectively. and 485 (8%) cases with available lytA results. South ified level. In addition. In 2009. The overall pneumococcal detection rate was 7% (422/6396). of an underlying continuous variable. 52/1351 [4%]) and autumn (March to May. we used a proportional-odds (ordinal) model. A total of 5855 of To identify factors associated with lytA positivity. bacterial 6396 patients (92%) had a known HIV status. The first wave was dominated by influenza A(H3N2). higher numbers of cases were detected concomitantly ( pneumococal load) was categorized into 5 levels. In addition.2 in the univariate analysis were Respiratory samples were tested by multiplex reverse-transcrip. Africa experienced 2 distinct waves of influenza virus circula- ciated OR) of covariates in each bacterial load level to the tion.oxfordjournals. 72/ effect of a predictor on the log odds of being above a specified 1588 [5%]. For the analysis of factors associated with pneumococcal The highest detection rate was observed in the 19–44-year age load. 23% (97/431). and 57% analysis. Five percent of patients (329/6348) re- the pneumococcal load (cutoffs. 145/1732 [8%]). and 2753 (43%) were children ≤5 years (Table 1). Pneumo- syncytial virus (RSV). A comparison of HIV prevalence Population and Factors Associated With lytA Positivity among patients with anonymized linked testing results (2926/ From May 2009 through April 2011. 203/2280. Table 1 and Figure 1). and 155 For the mortality analysis. we included potential determinants for and were HIV positive (Table 1). 2969 (46%) were males. The methods from the WHO Collaborating Center for Influenza. College Station. vember.

1) . 5 1 1 ≥5 2774 (44) 241 (57). 5 0.1–5.001 Hospitalization duration.402 . Yes 1519 (24) 81 (19)...022 Oxygen receipt n = 6380 n = 421 No 4391 (69) 224 (53).9 (1.. Spring 1732 (27) 145 (34).. ≥45 1169 (18) 91 (22).8 (.8–13. 8 2.. Age.4–.9–2. Summer 1351 (21) 52 (12).4–2. 1 1 1 Yes 6119 (96) 418 (99).8 (1. Edendale 442 (7) 25 (6)..9 (.6–... 6–18 194 (3) 4 (1).2) <.1–2.0 (1. 2 0. Downloaded from http://jid.5 (1.Table 1. 6 0...002 6.6) <.. 2010 ( Jan–Dec) 3143 (49) 216 (51)..5 (.358 .001 .001 .7 (.0 (1.5–2.6 (2..7–2. HIV infection n = 5855 n = 392 No 2890 (49) 111 (28). (% of Total).. (%) No. d n = 6396 n = 422 0–2 1909 (30) 65 (15).001 . 6 1.6 (1. 4 0.... 4 1 . 6 1.oxfordjournals..3 (1.5 (1...001 2.3) . 5 1 . South Africa. 9 2.166 . 8 2.8 (1.001 Antibiotic use (during n = 6392 n = 422 admission) No 273 (4) 4 (1).8) .9–3.4 (.5 (1.043 . 7 1 .0 (1.2–1.9) .035 . 2–5 499 (8) 31 (7).5–2..001 . 7 1 .5–24...5) <. 4 0.8 (1..6) .8) ..7) <. Male 2969 (46) 175 (42)...5) .177 .3 (1..9 (1. 7 1 . 19–44 2280 (36) 203 (48).1–2.011 Symptom by guest on April 17. 9 2. Winter 1725 (27) 153 (36).3) .1–2. d n = 6375 n = 421 0–4 3601 (56) 180 (43). Yes 784 (12) 33 (8)..8–2. 7 4. Seasond n = 6396 n = 422 Autumn 1588 (25) 72 (17).1–1. Influenza virus infection n = 6358 n = 420 No 5873 (92) 374 (89).2) <. 7 1. Rhinovirus infection n = 6343 n = 420 No 4824 (76) 339 (81).5–2. Sex n = 6369 n = 422 Female 3427 (54) 247 (58)..6 (.4) <.1 (1.9) ..3–2..3) ..2) . 2015 Hospital n = 6396 n = 422 Agincourt 1036 (16) 42 (10).. Chris Hani Baragwanath 4681 (73) 325 (77).6 (.001 .. lytA-Positive Cases.001 HIV and Influenza Virus Infections Increase Pneumococcal Load • JID 2014:209 (1 January) • 59 .. 4 1 1 Yes 2965 (51) 281 (72).5–2..1) <. Univariate and Multivariable Analysis of Factors Associated With Pneumococcal Infection (lytA Positive) Among Patients Hos- pitalized With Acute Lower Respiratory Tract Infection.6) <.8 (. 7 0.2) . 6 1 1 Yes 485 (8) 46 (11). % lytA Positive OR (95% CI) P (95% CI) P Year n = 6396 n = 422 2009 (May–Dec) 2003 (31) 150 (35).4) .010 RSV infection n = 6343 n = 420 No 5559 (88) 387 (92).9) .001 1. 13 3...3) <.9) <.021 .7–1. Adjusted ORb Variable No..6) <.9 (1. 10 2.008 1. y n = 6396 n = 422 <2 2254 (35) 93 (22).6–1..1) . 9 1.001 ..004 .001 1.5–2.7–. 8 1. 4 1 1 3–7 4487 (70) 357 (85).001 . 2009–2011 Univariate Analysisa Multivariable Analysis Enrolled Cases.1–3.4–2.. 9 2.0 (1. 2011 ( Jan–Apr)c 1250 (20) 56 (13). 10 1. 4 1 .001 1.7) <.. 7 1 . 5 1 1 Yes 1989 (31) 197 (47).4–.4 (2. 4 0. Klerksdorp 237 (4) 30 (7).

60 • JID 2014:209 (1 January) • Wolter et al . RSV. 2015 d Southern hemisphere seasons were classified as follows: autumn (March through May). and therefore the winter season of 2011 was excluded.3–11. *Influenza subtype accounted for at least 50% of all influenza virus subtypes detected during the period. adenovirus infection.023 Abbreviations: CI. HIV observed concomitantly with the circulation of the 2 influenza (OR.. Only covariates significant at the multivariable analysis are reported.5.2) . % lytA Positive OR (95% CI) P (95% CI) P ICU admission n = 6381 n = 421 No 6362 (99) 417 (99). kwashiorkor/marasmus. and antibiotic use ≤24 hours before admission. winter ( June through August).1–2.3) and influenza virus A subtypes (Figure 1). (% of Total). odds ratio.001 1. Weekly detection rate of Streptococcus pneumoniae (lytA-positive) and influenza virus among patients hospitalized with acute lower respiratory tract infection.8 (1.6. c Only cases enrolled from January through April 2011 were included in the study..5) . 7 1 .Table 1 continued. influenza virus subtypes. chronic renal failure.. 7-valent pneumococcal conjugate vaccine receipt. (%) No. among other factors (Table 1). respiratory syncytial virus. 2. underlying illness (including asthma. human immunodeficiency virus. 95% CI. 1.oxfordjournals. cirrhosis/liver failure. 6 1 1 Died 343 (5) 43 (10) 13 2. chronic lung diseases. lytA-Positive Cases. spinal cord injury. human metapneumovirus infection. intensive care unit. seizure disorder.. OR. 95% confidence interval [CI]. diabetes.1–3.5–3. Outcome n = 6374 n = 421 Survived 6031 (95) 378 (90). 1. Univariate Analysisa Multivariable Analysis Enrolled Cases. nephrotic syndrome.1–2. splenectomy.018 . and summer (December through February). a Only covariates statistically significant in the univariate analysis are reported. parainfluenza virus (1–3) infection. Downloaded from http://jid. 2. The following covariates were additionally assessed for significance in the univariate analysis: alcohol use. Yes 19 (1) 4 (1). heart failure. burns.0) <. 2009–2011. Adjusted ORb Variable No. virus (OR.1 (1. confidence by guest on April 17. biphasic detection of invasive pneumococcal pneumonia was In the univariate analysis. coronary heart disease.1) infections were significantly Figure 1. or cancer). emphysema. smoking. valvular heart disease. ICU. enterovirus infection. 21 3. HIV. b Calculated from observations with available data for all covariates significant at the multivariable analysis (n = 5798). South Africa. spring (September through November). immunosuppressive therapy.5 (1.

4 (. 1.2) 88 (70) 2.0) 40 (61) 2. 2. 95% CI. the lytA-positive individuals were HIV infection and pneumo- able analysis (Table 2).1 log DNA to lytA-negative patients (300/5953 [5%]. virus infection was 6% (2/32) among patients with a pneumo- . load was 82 732 (n = 14).3 (. Downloaded from http://jid.1–5.4 (1.2–3.0. . b Calculated from observations with available data for all covariates statistically significant in the multivariable analysis (n = 390). to have longer (≥5 days) tion of influenza virus subtypes with pneumococcal load. antibiotic use during admission. c The following covariates were assessed for significance in the multivariable analysis: age.5–24. 1. 2–5 years. 95% CI.6–29. 3. On univari- copies/mL among individuals aged <2 years.9) 56 (45) 1. 1. 1.1).032 Abbreviations: CI. A(H1N1) were at greater risk of death (aOR. Table 2.3 log DNA copies/mL. 95% CI.1–68.2–10.6–2.1–4.0) .6) <. RSV and rhinovirus 96% (25/26) among patients with a pneumococcal load of were significantly less associated with lytA positivity in the uni.002 infection Oxygen receipt 12 (37) 62 (44) 1. 95% CI. respectively (P = . variate analysis (Table 1). and ≥45 years.4 (1.1–2. sex. 1.1 (1.2–2.4.3) 8 (31) 6.1–2.oxfordjournals. compared with the unvaccinated group (34/793). as well significantly associated with elevated pneumococcal load. 3. the prevalence of influenza (6/183. receipt of 1 2. and outcome. the sample size was small. duration of symptoms. oxygen therapy.1–295. individuals with higher pneumococcal [aOR]. 26) among patients with a pneumococcal load of ≥100 000 tivity.associated with lytA positivity.001 Influenza virus 2 (6) 9 (6) 1. The median pneumococcal higher among lytA-positive patients (43/421.127). 10%). 95% CI.6 (1.3.c Variable (%) (%) (95% CI) (%) (95% CI) (%) (95% CI) (%) (95% CI) (95% CI) P HIV infection 16 (50) 93 (66) 1.1 (1. human immunodeficiency virus.1–2. a The reference group was subjects with <102 DNA copies/mL.1–6.8) or 2 (11/229.9–4. season of the year. smoking. to receive anti.9–19. 95% CI.1) 7 (6) 0. 95% CI.6.2–2. respectively.001). compared with 31% (8/ (PCV7) was not significantly associated with reduced lytA posi. The mean pneumococcal biotic treatment during admission (aOR. HIV and influenza virus infections were coccal loads of ≥10 000 or ≥100 000 DNA copies/mL.5–2. with a tients (6374/6396) with lytA results. No.1) and for patients infected with influenza A(H3N2). odds ratio.7. South Africa.4) and influenza virus (aOR. 6.9. patients with invasive pneumococcal A subanalysis restricted to cases positive for both pneumo- pneumonia were more likely to present late (>2 days from sym.6 (1. and 64 487 (n = 15) and to require oxygen therapy (aOR. 1.9 (.6–3. coccus and influenza virus revealed no difference in the associa- ptom onset. antibiotic use ≤24 hours before admission. admission to the intensive care unit.6 (1. 2015 Factors Associated With Pneumococcal Load Among lytA. ORa No.3–1.2) 11 (9) 1. HIV and Influenza Virus Infections Increase Pneumococcal Load • JID 2014:209 (1 January) • 61 . ORa No. 95% loads were more likely to require oxygen therapy during hospi- CI. 34 906 (n = 17). compared loads by age group were 2. Among eligible children.1–9.6–3.7) 2. In addition. length of hospitalization. 3. confidence interval.8 (. 95% CI.5–2.1–1.5–2. ORa No. underlying illness. DNA copies/mL (aOR [ proportional-odds model] 1.3.6. The case-fatality ratio was median of 3.4. lytA-positive patients patients with a pneumococcal load of <100 DNA copies/mL to who had a load of ≥10 000 DNA copies/mL (aOR.0 (2.4–2.8. 95% CI. 1.5–7. P < . HIV. OR. 1. ≥100 000 DNA copies/mL (aOR [ proportional-odds model]. 2009–2011 Proportional-Odds lytA-Positive Cases.3 ( by guest on April 17. 1. infection with respiratory viruses other than influenza virus. OR. and influenza B.9) 1.6–3. The as different markers of disease severity (Table 3).8. 1. HIV (adjusted OR 1.2) doses of the 7-valent pneumococcal conjugate vaccine coccal load of <100 DNA copies/mL. aOR.6 (.018 Death 3 (9) 11 (8) 0. prevalence of HIV infection increased from 50% (16/32) among On multivariable analysis (Table 3). 1.0 (.2).3) 57 (86) 6.3 (. Factors Associated With Mortality Among lytA-Positive Positive Cases Patients Among the 422 lytA-positive cases.6).0.9) 2.4 (. OR. On multivari. 19–44 years. 1.2) infections remained significantly associated with talization and were at greater risk of death. lytA positivity.4 (1.3–14.4.1–2.2 to 6. 6–18 ate analysis. 3. 0.6) 8 (31) 4.8) 1. 1. 1.2–18. Conversely. ORa Adjusted ORb.9).1) .5. Similarly.6). 95% CI.3) . by Pneumococcal Load in DNA copies/mL Model <102 102 to <103 103 to <104 104 to <105 ≥105 (n=32) (n=141) (n=125) (n=66) (n=26) No. In addition. On multivariable analysis (Table 1). hospital. hospitalization time (aOR.3) 17 (65) 3.4) 11 (17) 1. influenza virus infection.1. however.1. Multivariable Analysis of Factors Associated With Pneumococcal Load Among Patients With Pneumococcal Infection (lytA Positive).1) 11 (17) 3. alcohol consumption.7).9 (.0 (. 95% CI.1 log DNA copies/mL. the overall pneumococcal The outcome of hospitalization was available for 99% of pa- loads ranged from 1. HIV infection. and 3. pdm09. factors associated with increased mortality among years.6) 25 (96) 25.

. immunosuppressive therapy.9) . human metapneumovirus.6) <.5 (.1 (. a Only covariates statistically significant in the univariate analysis are reported.0) .0) were at increased risk of death. pneumococcal load cutoff of 102 DNA copies/mL.001 3. Symptom duration.8–7. d 0–4 26/180 (14) 1 1 ≥5 17/240 (7) 0.001 4.1–7. Pneumococcal load.3 (. seizure disorder. 2015 0–4 26/331 (8) 1 1 5–7 17/90 (19) 2. Of the 23 culture. patients who died were more likely to require In this study.Table 3..2) <. risk of death. cirrhosis/liver failure.7.oxfordjournals. heart failure. Yes 2/4 (50) 9. nephrotic syndrome. enterovirus.. indicating a fulminant progression of illness. OR.2–.. intensive care unit. ICU. coronary heart disease.. fection and influenza virus coinfection were independently as- sociated with invasive pneumococcal pneumonia and elevated Association Between lytA and Blood Culture Positivity blood pneumococcal loads. odds ratio.1–10. and antibiotic use during admission.016 0.001 105 cutoff <105 33/392 (8) 1 . South Africa.6–7.8–7.7 (2... 4. 13. From this same set of cases.2–.7 (2. 95% CI. chronic lung diseases.4 (1. underlying illness (including asthma.4) <. HIV in- 95% CI. spinal cord injury. 2009–2011 Univariate Analysisa Multivariable Analysis Variable Case-Fatality Ratio (%) OR (95% CI) P Adjusted ORb (95% CI) P Influenza virus infection No 34/373 (9) 1 .4) mens of patients hospitalized with acute lower respiratory tract and had a shorter (<5 days) length of hospitalization (OR. The following covariates were additionally assessed for significance in the univariate analysis: sex. Only covariates statistically significant at the multivariable analysis are reported. 2.4. .7 (1.029 . ≥105 10/29 (34) 5. 2% (23/1120) were recorded as culture posi. infection as a marker for pneumococcal pneumonia.3) .. The association between high blood bacterial load and severe positive cases. Univariate and Multivariable Analysis of Factors Associated With Death Among Patients With Pneumococcal Infection (lytA-Positive). smoking. confidence interval. age group.7) .1 (2. HIV infection No 5/111 (4) 1 . 30].2) and presented to hospital late (≥5 days from onset of DISCUSSION symptoms. human immunodeficiency virus.3–10. diabetes. In this these 9 patients died. One of disease.001 . rhinovirus.8 ( by guest on April 17. is well described [9..7) . alcohol use. or cancer). Abbreviations: CI. valvular heart disease..001 ICU admission No 41/416 (10) 1 .8–4. HIV infection was found to be associated with an 62 • JID 2014:209 (1 January) • Wolter et al .2–. d Downloaded from http://jid.. adenovirus.8) <. 9 (39%) were negative by real-time PCR. kwashiorkor/marasmus. including mortality.031 . 1.1–9. chronic renal failure.3) . In addition.3–66.. Yes 9/46 (20) 2.4–5. 1. study.6–7. aOR. splenectomy. emphysema..0 (2. b Calculated from observations with available data for all covariates statistically significant at the multivariable analysis (n = 419). 18% of cases (1120/6396) had blood culture results of 10 000 DNA copies/mL were associated with an increased available. 0.6 (1. In addition.9) . Yes 35/280 (12) 2.. we detected pneumococcal DNA in blood speci- oxygen therapy during admission (aOR. pneumococcal load cutoff of 103 DNA copies/mL.003 3.3) <.055 . 10% (116/ vasive pneumococcal pneumonia than blood culture alone. parainfluenza virus [1–3]). DNA copies/mL 104 cutoff <104 21/322 (7) 1 1 ≥104 22/99 (22) 4. 3. 95% CI.1–10. Of these. burns.7).. antibiotic use ≤24 hours before admission..001 Hospitalization duration.3.002 Oxygen receipt No 9/223 (4) 1 1 Yes 34/197 (17) 5. HIV.5–13. bacterial loads in excess In total. Real-time PCR detected 4-fold more cases of in- tive for pneumococcus. 1120) were PCR positive for pneumococcus.1 (1. infection with respiratory viruses other than influenza virus (respiratory syncytial virus.

PCV is advocated for rapidly identified and treated. the detection of lytA in the blood of cases with a duration of symptoms >7 days in this study may may reflect transient bacteremias. In the multivariable model. compared with infection This study has a number of limitations. the assay is not with influenza virus or pneumococcus alone [36]. In addition. [10]. in which real-time PCR was 4-fold tality. findings were not available. more sensitive than blood cultures when used to diagnose ed pneumococcal load. but results were similar to marker for invasive pneumococcal pneumonia-associated mor- findings by Resti et al. Limited studies targeting the therefore have resulted in an underestimation of the number of ply gene have shown that pneumococcus can be detected in the secondary pneumococcal infections. the determination of pneumococcal blood culture alone. the number of patients who were ed with an increased risk of invasive pneumococcal pneumonia eligible to receive PCV7 and were positive for pneumococcus and an elevated pneumococcal load. 43]. because creased pneumococcal load has been observed in squirrel monkeys it is dependent on pneumococcus crossing into the blood- coinfected with influenza virus [37]. First. 11]. results could not be stratified by serotype. pneumococcal pneumonia was estimated to resent a minimum estimate. the true part. when coinfected with both pathogens. specimens [7. Third. we had limited data account for 8% of clinical pneumonia cases and 36% of chest on CD4+ T-cell counts and HIV treatment. tion rate differed significantly among the sites on univariate fluenza virus and pneumococcus. but this was the blood of healthy children colonized with pneumococcus not observed in our study. in this study influenza virus infection was associat. In addition. serotyping data. In this study. This is consistent with previous findings pneumococcal community-acquired pneumonia. which may be of particular use in devel- PCVs have been shown to be safe and effective in reducing oping countries. Our study design was not appropriate invasive pneumococcal pneumonia and associated mortality in for the assessment of PCV7 protection because lytA-positive this group. the ability to quantify pneumo- pneumococcal disease in HIV-infected children [31–33] and coccus enables individuals at risk of severe outcome to be adults [34]. 39]. real-time time PCR on blood specimens as a marker of pneumococcal PCR detected 5 times as many cases than were detected by pneumonia. surveillance programs such most likely because of the strong age association of these respira. likely to detect all cases of pneumococcal pneumonia. [33]. loads at the time of diagnosis may have a role as a prognostic formed only on a subset of cases. by guest on April 17. Studies have reported a blood of healthy individuals [40. On [42. accounting for HIV Downloaded from http://jid. A number of mechanisms were too few for meaningful analysis. did not identify pneumococci in and other respiratory viruses. Real-time [13] in which HIV-infected children with pneumonia had PCR enables diagnosis and serotyping from culture-negative higher pneumococcal loads than HIV-uninfected children. 2015 and increased inflammatory responses [35]. The exclusion underestimated. Based on vaccine. the protective effect of RSV and rhinovirus such as oxygen saturation on admission and chest radiograph that was observed in the univariate analysis was not significant. this study highlights the usefulness of real- burden of pneumococcal pneumonia is thought to be. 41]. in resulted in false-negative lytA results. Experimentally in. number of reasons. In addition. veillance staff. and age. As a result. HIV and influenza virus infection were associated with HIV and Influenza Virus Infections Increase Pneumococcal Load • JID 2014:209 (1 January) • 63 . studies targeting similar synergistic interaction in children between pneumococcus the lytA gene. In addition. spective sites. including increased bacterial analysis. explain the lethality of the dual infection [18]. our estimates of mortality likely rep- probe studies.oxfordjournals. In addition. virus-induced conditions that promote bacterial growth. This may be due to a tory viruses. in comparison with blood culture. 10. there was no significant difference observed between fected mice demonstrated a significant reduction in survival rates the sites. In addition. the effect radiograph–positive pneumonia cases in children aged <5 years of these covariates could not be evaluated. as in our study. Therefore. as ours tend to underestimate mortality. such as RSV [19. vaccination PCV7 vaccination was not associated with a reduced risk of of HIV-infected individuals of all ages may reduce the risk of pneumococcal disease. Although these estimates are still lower than what the true In conclusion. In addition. The increased risk of invasive stream. However. Based on the results of this study. detailed information on clinical characteristics multivariable analysis. which included all age groups. small volumes of blood were tested in the pneumococcal pneumonia and elevated pneumococcal load in PCR assay. Second. Recent studies disease prevalence and the magnitude of the identified risk of the 1918 influenza epidemic have indicated that the majority factors for invasive pneumococcal pneumonia may have been of deaths occurred 7–14 days after infection [38]. including the fact that severely ill patients The blood pneumococcal detection rate observed in patients may be less likely to consent to inclusion in the surveillance or aged ≤18 years during the early introduction of PCV7 (5%) that a proportion of patients may die before hospital admission was lower than the 10% described in a similar study performed or shortly after admission but before being approached by sur- in Italian children ≤16 years of age [10]. blood cultures were per. therefore.increased risk of invasive pneumococcal pneumonia and elevat. The pneumococcal detec- have been proposed for the synergistic interaction between in. owing to the lack of Similarly. Currently in South Africa. use in children. most likely because of differences in the HIV preva- colonization as a result of epithelial cell damage by influenza lence and age distribution of the patients presenting to the re- virus. which may have influenza virus–infected patients observed in this study may.

2009–2010. Severity of meningococcal disease ing by real-time polymerase chain reaction using blood samples. as well as Akhona Tshangela and Phumza Tikilili from the Centre pathogens: implications for pandemic preparedness. 5. et al. moniae lower respiratory infection in hospitalized children by culture. Campos JM. Wkly Epidemiol Rec 2007. Peters RP. 2010. et al. Ma Fat D. Alpert G. Grabowska K. Simonsen L. Jordan R. Molecular detection methods veillance standards for influenza. disease. Janoff EN. Pediat- Disease Control and Prevention (cooperative agreement 5U51IP000155). et al. Use of a rapid test of pneu. N. Lancet Infect Dis 2008. 2011. rics 2008. Khoosal M.mrc. A trial of a 7-valent pneu- are associated with mortality in Malawian children with invasive pneu. and urinary antigen de. 1984–2008. Feasibility of real-time polymerase chain reaction in whole blood to identify Streptococcus pneumoniae in ated with an increased risk of death. Palladino S. Carrol Madhi SA. Impact of the 2009 influenza pandemic on pneumococcal pneumonia Br Med Bull 2009. de Boer by guest on April 17. O’Brien KL. Rello J. Madhi SA.statssa. 82:93–104. and serotyping performed directly on clinical samples improve diag. Madhi SA. Clin Infect Dis 2009. 61:72–5. et al. Viboud C. 26:416– cation of Streptococcus pneumoniae as the cause of community-ac. Boerma T. Petersen K. Diagn 32. O’Brien KL. Svensson A. 198:962–70. Bradshaw D. Gordon SB. Daniel TJ. Dis 2008. J Infect Dis Clin Infect Dis 2011. Statistics South Africa. 47:3308–12. were associ. Longo D. J Infect Dis 2012. Walters MI. et al. et al. Conflicts that the editors consider relevant to the to influenza. This may be a pathway by patients with community-acquired pneumonia. Cohen C. DNA load in blood as a marker of infection in patients with communi.South Africa. Dorrington R. 368:1048–50. Severity of pneumococcal pneumonia reaction assay in patients hospitalized with severe acute respiratory associated with genomic bacterial load. Atwell-Melnick N. All authors: No reported conflicts. Salama P. Tondella ML. Novel approaches to the identifi. et al. et al. J Med Microbiol 2008. High pneumococcal DNA loads 34. J Infect Dis 2012. for detection of pneumococcal DNA. Morens DM. 2012. et al. The demographic leading killer of children. Michelow IC. 2012. 15.oxfordjournals. Severe pneumococcal pneu- Downloaded from http://jid. which. Evaluation and im- Streptococcus pneumoniae in Italian children. Seasonal invasive pneumococcal (investigator-initiated research agreement WS1167521) and the Centers for disease in children: role of preceding respiratory viral infection. impact of HIV on Streptococcus pneumoniae bacteraemia in a South African population. Diagn Microbiol Infect which the immune dysregulation caused by HIV and/or influ. Disclaimer. Streptococcus pneumoniae in children younger than 5 years: global esti- tection. Albrich WC. Mbelle N. Lancet Infect Dis for Respiratory Diseases and Meningitis. Bender J. pneumonia as a cause of death in pandemic influenza: implications for stitutions or the agencies funding the study. Accessed 12 June 2012. This work was supported by Pfizer South Africa 19. 374:893–902. Kay I. and psaA genes 57:1205–12. 64 • JID 2014:209 (1 January) • Wolter et al . Lisboa T. 205:458–65. 25. J Clin Microbiol 2007. Albrich WC. Pediatrics 1985. Wolfson LJ. Weinberger 33. WHO interim global epidemiological sur- 7. N Engl J Med 2003. Interactions between influenza and bacterial respiratory agement. 8:67–80. et al. 47(Suppl 3):S202–6. Watt JP. 206(Suppl 1): 10. 136:832–40. 206(Suppl 1):S153–8. 17. 122:229–37. Pierce 11. Klugman K. Guiver M. 19:1141–7. teremic pneumococcal pneumonia in children: diagnosis and serotyp. The evidence for using conjugate Microbiol Infect Dis 2008. McCaustland K. O’Brien KL. 2. 23. 6:303–12. 2015 monia in previously healthy children: the role of preceding influenza References infection. Routine quantitative blood 2008: Findings from death notification. The findings and conclusions in this article are those of 18. Mwalukomo T. Huebner R. Occur- All authors have submitted the ICMJE Form for Disclosure of Potential rence of invasive pneumococcal disease and number of excess cases due Conflicts of Interest. Sheng X. Ekdahl K. 76:901–4. Mid-year population estimates 2011. Clin associated with genomic bacterial load. Lujan M. Predominant role of bacterial the authors and do not necessarily represent the views of their affiliated in. Penttinen P. Lancet 2006. Wasas A.who. Pneumococcal conjugate vaccine for childhood immunization—WHO 24. 20. http://www. Mortality and causes of death in South Africa. Moriondo M. identified by a 10-plex real-time reverse-transcription polymerase chain 9. Community-acquired Financial support. 21. Crewe-Brown H. Fauci AS. Impact of Notes human immunodeficiency virus type 1 on the disease spectrum of Streptococcus pneumoniae in South African children. ://www. 2006. Azzari C. et al. resources/documents/INFSURVMANUAL. Lancet 2009. Indolfi G. Cortimiglia M. Pediatr Infect Dis J 2007. 3. et al. Johnson L. 362:812–22. Klugman KP. Respiratory Illness study for the collection of specimens and database man. 92:7–32. Accessed 3 June 2013. Kohberger R. Taubenberger JK. 349:1341–8. 22. Clin Infect Dis 2008. Huebner RE. 13. 54:601–9. Johansson EW. Mathers CD. We thank all members involved in the Severe Acute J 2000. French N. 4. Nkhoma S. The coccal pneumonia patients. 31. Moyes J. J Clin Microbiol 2009. et al.pdf. Diagnosis of Streptococcus pneu. ply. 20:58. Klugman KP. 30. Steiner C. Indicators. Mason E. 51:1042–9. Chest 2009. 12:2177–84. 8. AIDS 1998. Adrian PV. Carvalho MG. mates. Rouphael NG. Miller M. Olsen K. N Engl J Med mococcal disease. Ampofo K. Moriondo M. Bell LM. http://www. 48:587–94. illness . 29. Klugman KP. A trial of a 9-valent pneumococcal conjugate vaccine in children ase chain reaction for the detection of Streptococcus pneumoniae in with and those without HIV infection. Resti M. impact of HIV/AIDS in South Africa: National and provincial indica- 6. Kee C. in turn. tors for 2006. Pediatr Infect Dis Acknowledgments. nostic sensitivity and reveal increased incidence of invasive disease by 28. Respiratory viral coinfections ty-acquired pneumonia. pneumoniae bacteremia. 1. McAnerney JM. Brundage JF. Streptococcus pneumoniae 45:2460–6. tient influenza surveillance in South Africa. 14. Twenty-five years of outpa- mococcal colonization density to diagnose pneumococcal pneumonia. et al. 30:784–9. et al. Global and regional causes of death. J Infect Dis 2008. Bliss SJ. mococcal conjugate vaccine in HIV-infected adults. provement of real-time PCR assays targeting lytA. Hogberg L. serological testing. et al. Sellman J. A real-time polymer. et al. Guiver M. Potential conflict of interest.pdf. 34:E1–11. 27. cultures in children with Haemophilus influenzae or Streptococcus Accessed 4 May 2012. Pretorius MA. quired pneumonia. Darton T. Clin Infect Dis 2000. Jones N. Burden of disease caused by polymerase chain reaction. 62:23–5. Madhi SA. Infect Dis 2010. vaccines to protect HIV-infected children against pneumococcal 12. World Health Organization. Pneumonia: the 26. et al. S159–65. Wardlaw content of the manuscript have been disclosed. enza virus infection results in severe outcomes among pneumo. Madhi SA. BMC Infect Dis 2006. Clin Infect Dis 2002. Schuurman T. Cohen increase in pneumococcal loads. pandemic influenza preparedness. 2006. http://www. Madhi A. Statistics South Africa.statssa. Lozano J. http position paper. blood using a mouse model: a potential new “gold standard”. Naylor S. Plotkin SA. Accessed 4 May 2012. hospitalizations in the United States.

McCullers JA. 37. Rouphael N. et by guest on April 17. A role for Streptococcus pneumoniae in virus. Nat Med 2004. respiratory samples from adults with community-acquired pneumonia. 27(Suppl 3): 43. 36:669–73. Klugman KP. 186: assay for detection of Streptococcus pneumoniae in respiratory and non- 341–50. children colonized with Streptococcus pneumoniae. Madhi SA. Madhi SA. Shriker O. Pneumococcal pneumonia targeting the lytA gene. 41:63–6. Dagan R. Pneumococcal DNA is Infect Dis 1975. Evaluation of a PCR role of platelet-activating factor receptor.oxfordjournals. 2015 HIV and Influenza Virus Infections Increase Pneumococcal Load • JID 2014:209 (1 January) • 65 . Clin Microbiol Rev 2006. pneumonia due to Streptococcus pneumoniae in the squirrel monkey. Long GG. J Clin Microbiol 1998. Bangert M. Beynon KA. et al. Moriondo M. Rehg JE. Walker JS. et al. Downloaded from http://jid. Berendt RF. Steyn S. Cortimiglia M. et al. Anderson TP. 10:811–3. McCullers JA. and influenza: a deadly combination. Influenza alone and in sequence with J Clin Microbiol 2003. Vaccine 2009. Klugman KP. 132:689–93. J Med Microbiol 2011. ical relevance of detection of pneumococcal DNA in sera of children by 36. Chien YW. Streptococcus pneumoniae: characterization of a mouse model and the 41. J Infect Dis 2002. common protein real-time polymerase chain reaction assays in healthy 39. Use of 2 pneumococcal C9–14. Murdoch DR.35. 60:710–4. J 42. 70:452–4. Diagn Microbiol associated pneumonia. Insights into the interaction between influenza virus and 40. Hazan I. Azzari C. Prospective study to determine clin- pneumococcus. Lethal synergism between influenza virus and PCR. Infect Dis 2011. 19:571–82. not detectable in the blood of healthy carrier children by real-time PCR 38.