Breiman RF, Keller DW, Phelan MA, Sniadack DH, Stephens DS, Rimland D, Farley MM, Schuchat A, Reingold AL. Evaluation of Effectiveness of the 23-Valent Pneumococcal Capsular Polysaccharide Vaccine for HIV-Infected Patients. Arch Intern Med. 2000;160(17):2633–2638. doi:10.1001/archinte.160.17.2633
We conducted a retrospective case-control study to evaluate effectiveness of pneumococcal vaccine against invasive disease among adults with human immunodeficiency virus (HIV) infection in San Francisco, Calif, and Atlanta, Ga.
Case patients were 18- to 55-year-old subjects with HIV infection who were admitted to selected hospitals in Atlanta or San Francisco from February 1992 to April 1995 from whom Streptococcus pneumoniae was isolated from a normally sterile site. Controls were HIV-infected patients of similar age matched to cases by hospital of admission and CD4 lymphocyte count (<0.20, 0.20-0.499, ≥0.50 × 109/L [<200, 200-499, ≥500 cells/mm3]) or clinical stage of acquired immunodeficiency syndrome. Case and control subjects were restricted to persons known to have HIV infection before hospital admission. Analysis used matched univariate and conditional logistic regression.
One hundred seventy-six case patients and 327 controls were enrolled. By univariate analysis, persons with pneumococcal disease were more likely to be black, be current smokers, and have close contact with children. Adjusted for these factors and CD4 cell count, pneumococcal vaccine effectiveness was 49% (95% confidence interval [CI], 12%-70%). Adjusting for all variables and key interaction terms, vaccine effectiveness among whites was 76% (95% CI, 35%-91%), whereas effectiveness among blacks was 24% (95% CI, −50% to 61%). Among controls, vaccination was significantly less common among blacks (29% vs 45%; P<.005).
Pneumococcal vaccine demonstrated protection against invasive pneumococcal infections among white but not black HIV-infected adults. Failure to demonstrate effectiveness among blacks may be due to limited power because of low use of the vaccine in this population, immunization at more advanced stages of immunosuppression, or unmeasured factors. These data support current recommendations for use of pneumococcal vaccine in HIV-infected persons and highlight a clear need for strategies to improve vaccine-induced protection.
PERSONS INFECTED with the human immunodeficiency virus (HIV) are at markedly increased risk for invasive infections with Streptococcus pneumoniae. Among adults, HIV-infected persons have an incidence of invasive pneumococcal disease that is 150 to 300 times greater than that of the general population,1,2 and it is estimated that 15% to 20% of HIV-infected children will have pneumococcal bacteremia during their HIV disease.3
Protective immunity to S pneumoniae is conferred by antibody to capsular polysaccharides, which is the basis for the use of the currently licensed 23-valent pneumococcal polysaccharide vaccine. This vaccine prevents invasive pneumococcal disease caused by vaccine serotypes among immunocompetent persons for whom the vaccine is recommended4,5; its effectiveness in preventing invasive disease among immunosuppressed persons, however, is less clear.5 Nonetheless, since 1989 the Advisory Committee on Immunization Practices6 and the American College of Physicians have recommended the use of the 23-valent pneumococcal vaccine for HIV-infected persons.6,7 In addition, use of the vaccine is recommended by the Infectious Disease Society of America and the US Public Health Service Task Force on Opportunistic Infections.8
To provide information useful for making evidence-based recommendations on pneumococcal vaccine use, we conducted a case-control study of the effectiveness of the 23-valent pneumococcal polysaccharide vaccine among HIV-infected adults admitted to hospitals in Atlanta, Ga, and San Francisco, Calif; in this article, we report the results of that investigation.
This study was a matched case-control study that prospectively identified cases of invasive pneumococcal disease and retrospectively determined vaccination status. It was conducted at 4 hospitals in Atlanta and 6 hospitals in San Francisco from February 1992 to April 1995. The study was part of a larger program for surveillance of invasive pneumococcal disease among all persons 18 to 55 years old in those hospitals. The study was approved by the institutional review boards of the participating hospitals and that of the Centers for Disease Control and Prevention (CDC).
A case of invasive pneumococcal disease was defined as the isolation of S pneumoniae from a normally sterile site, including blood, cerebrospinal fluid, pleural fluid, open or transthoracic lung biopsy material, and joint aspirate material. Cases were identified from records in the microbiology laboratories at participating hospitals. A case patient was eligible for enrollment into the case-control study if he or she had invasive pneumococcal disease and was 18 to 55 years old, admitted to a study hospital, and known to be HIV seropositive before hospital admission for invasive pneumococcal disease. After identification of a potential case patient, study personnel contacted the attending physician to ask permission to enroll his or her patient into the study. If permission was granted, each case patient was informed about the study, and consenting patients were enrolled.
Consenting patients were administered a standardized questionnaire to collect information on risk factors for pneumococcal disease potentially relevant to the effectiveness of vaccine. Close exposure to a child younger than 10 years was defined based on patient assessment (ie, response to the survey question, "Did you have close exposure to a child younger than 10 years of age during the month before hospitalization?").
Two controls were sought for each case patient. Controls were identified from among HIV-infected persons 18 to 55 years old who had been hospitalized within 6 months (before or after) of the admission or discharge of the matching case patient and who had been known to be HIV infected before hospitalization. Controls were matched to case patients by hospital of admission and HIV stage based on CDC laboratory category (CD4 lymphocyte count of <0.20, 0.20-0.499, >0.50 ×109/L [<200, 200-499, ≥500 cells/mm3]) or, for cases without CD4 lymphocyte count within 3 months of admission, clinical stage. Clinical staging, when necessary, was determined from medical records that existed before the hospital admission of interest.9 When controls at clinical stage A could not be found for case patients at clinical stage A, controls at stage B with CD4 lymphocyte counts greater than 0.50 × 109/L were used. For case patients in clinical stage B, we were able to identify controls in the same clinical stage.
Potential controls were excluded if they had any evidence of pneumococcal disease during their hospital admission (including sterile-site isolates and radiographic evidence of pneumonia in conjunction with sputum Gram stain or culture containing obvious or apparent pneumococci). Potential controls with pneumonia without etiology identified as well as patients with a nonpneumococcal etiology were not excluded.
All S pneumoniae isolates were tested at the CDC for bile solubility and susceptibility to optochin. Isolates confirmed as pneumococci were then serotyped on the basis of capsular swelling with serotype-specific antisera (Quellung reaction)10; the antisera used were prepared at the CDC. Laboratory personnel performing serotyping did not know the vaccination status of case patients.
Study personnel collected names of all physicians who had cared for each case and control patient since 1988; they subsequently contacted those practitioners and requested a review of enrollees' records to determine if pneumococcal polysaccharide vaccine had been administered to those persons. If known, the vaccination status and (where appropriate) date(s) of vaccination were recorded; if unknown, the status was recorded as such.
Analysis was performed using SAS statistical software, version 6.12 (SAS Institute Inc, Cary, NC). Vaccine effectiveness (VE) was calculated by the formula VE = 1 − OR, where OR was the matched odds ratio.11 For 14 case patients, no matching control could be identified who had been admitted to the same institution within the 6-month period allowed by the study protocol; these cases were excluded from all analyses. Matched pairs were included in the analyses along with matched triplets. For vaccine effectiveness calculations, subjects with unknown vaccine status were excluded. For calculation of effectiveness against serotypes included in the 23-valent vaccine, cases without pneumococcal serotype results were excluded, as were those with nonvaccine serotypes (including those with vaccine-related serotypes, such as 23A).
We conducted matched univariate analysis and multivariable conditional logistic regression using the SAS procedure PHREG.12 After evaluation for interaction and collinearity, factors that were associated with risk in univariate analysis with a P value of less than .10 or were potential confounders based on previous studies were included in the multivariable models. Vaccination status was the independent variable included in all models and pneumococcal disease the dependent variable. We used stepwise (backward) elimination to determine variables that were independent predictors of invasive pneumococcal disease.
We identified 176 patients who met the case definition and had at least 1 matching control. Eighty-two percent of case patients were male, and their ages ranged from 20 to 54 years, with a mean age of 37.4 years. Of the 171 cases identified by race, 56 (33%) were white, 111 (65%) were black, and 2 (1%) were Asian/Pacific Islander; 9 (6%) of the 152 cases of known ethnicity were Hispanic (Table 1). Pneumococcal disease resulted in death for 11 case patients (6%).
For 25 cases, only a single matching control could be identified and enrolled, and 2 controls were enrolled for 151 cases. There were no statistically significant differences between cases and matched controls in age, sex, antiretroviral drug use at admission, or trimethoprim-sulfamethoxazole use at admission. However, cases had significantly higher CD4 lymphocyte counts than controls (Table 1) and were less likely to have an acquired immunodeficiency syndrome (AIDS)–defining illness (based on criteria for AIDS diagnosis in use at the onset of the study) at the time of enrollment (matched odds ratio [OR], 0.22; 95% confidence interval [CI], 0.13-0.39). Cases were more likely than matched controls to be black (matched OR, 2.1; 95% CI, 1.3-3.4) and to smoke cigarettes (OR, 1.7; 95% CI, 1.1-2.8). Case patients were more likely than controls to have close exposure to a child younger than 10 years (OR, 1.6; 95% CI, 1.1-2.4) (Table 1). A multivariable model found smoking, black race, and close contact with a child to be associated with invasive pneumococcal disease independent of vaccination status (Table 2).
Vaccination status could be verified for 162 cases (92%) and 305 controls (93%). Forty-one cases (25%) and 112 controls (37%) had received the 23-valent pneumococcal vaccine before hospitalization (OR, 0.59; 95% CI, 0.38-0.91) (Table 1). Among the 11 fatal cases, 5 were immunized. The earliest onset of illness for vaccinated cases was 53 days following immunization.
Stepwise multivariable analysis, incorporating race, smoking, household contact with children, and CD4 category (<0.10, 0.10-0.199, ≥0.20 × 109/L) into a conditional logistic regression model, estimated a vaccine effectiveness of 49% (95% CI, 12%-70%) (Table 3). When an interaction between race and vaccination status (P = .06) was included in the model, vaccine effectiveness was 76% (95% CI, 35%-91%) for white persons and 24% (95% CI, −50% to 61%) for black persons. Inclusion in the model of the interval between HIV diagnosis and pneumococcal disease (or hospital admission for controls) did not affect the results.
Among all control subjects of known vaccination status, black persons (29%) were less likely to be vaccinated than white persons (45%) (χ2 = 8.5, P = .004). There were no significant differences between whites and blacks in trimethoprim-sulfamethoxazole use or antiretroviral drug use; however, whites were more likely than blacks to have AIDS at enrollment.
Sterile-site isolates were available for serotyping from 153 case patients; the remaining 23 case patients were diagnosed as having pneumococcal disease by the isolation of pneumococci from sterile sites at their admitting institutions, but the isolates were unavailable for confirmation and serotyping. For 124 patients (81%), isolates were of serotypes contained in the 23-valent pneumococcal vaccine; 11 others (7%) were vaccine-related serotypes (eg, serotype 23F is vaccine specific and serotype 23A is vaccine related). When the multivariable model was limited to cases with isolates of serotypes included in the 23-valent vaccine and their controls, 91 matched sets were available for analysis. Adjusted vaccine effectiveness controlling for all 4 variables (race, smoking, household contact with children, and CD4 cell count) was 39% (95% CI, −18% to 68%). Analysis including interaction terms for race and vaccination status found a vaccine effectiveness of 74% (95% CI, 8%-93%) for white persons and 8% (95% CI, −112% to 60%) for black persons.
We explored several factors to understand the apparent differences in vaccine effectiveness between white and black persons. Information on CD4 cell count at the time of vaccination was not available. Data on the interval between HIV diagnosis and hospitalization were used to evaluate the duration of recognized HIV infection in black and white patients. Among cases, whites had a longer interval (median, 1641 days) between diagnosis of HIV and hospitalization compared with black patients (median, 1034 days) (P = .01). Among all immunized study participants, the interval between diagnosis of HIV infection and pneumococcal vaccination was greater for white patients (median, 974 days) compared with black patients (median, 413 days) (χ2 = 3.05, P = .08). Presence of sickle cell characteristics did not influence vaccine effectiveness calculations: 2 cases and 5 control patients reported sickle cell anemia; no cases and 2 control patients reported sickle cell trait.
Similar proportions of blacks (9%) and whites (7.5%) received more than 1 dose of pneumococcal vaccine since 1988. For comparison with use patterns for another vaccine given to adults, a similar proportion of whites (70%) and blacks (62%) had received influenza vaccine in the previous year. Among cases, 87% of blacks and 89% of whites were infected with pneumococcal serotypes that were included or related to those within the 23-valent vaccine.
The markedly increased rate of pneumococcal disease among HIV-infected persons and the documented safety, effectiveness, and low cost of 23-valent pneumococcal polysaccharide vaccine for the prevention of invasive disease among immunocompetent persons have been cited as justifications for routine use of the vaccine in HIV-infected persons.3,6 However, the strength of the recommendation is low. The Advisory Committee on Immunization Practices assigned a category C: "Effectiveness of vaccination is not proven, but the high risk for disease and the potential benefits and the safety of the vaccine justify vaccination." Evidence of transient increases in HIV load following pneumococcal immunization, although not shown to have adverse clinical impact,13 may have contributed to reluctance to use this vaccine in HIV-infected patients by some physicians in the absence of demonstrated effectiveness.
Studies of the effectiveness of pneumococcal vaccine in HIV-infected patients have produced conflicting results. A case-control study conducted in Baltimore, Md, suggested an effectiveness of 78% for prevention of pneumococcal bacteremia among patients with CD4 cell counts greater than 0.20 × 109/L but did not find effectiveness among persons with lower CD4 cell counts.14 A recent prospective, randomized, placebo-controlled study15 conducted in Uganda found no efficacy in preventing invasive pneumococcal disease or pneumonia. Given substantial differences in the epidemiology and treatment of HIV infection and epidemiology of pneumococcal disease (including serotype distribution) in sub-Saharan Africa when compared with industrialized settings, data from Uganda would appear to have limited relevance to discussions on vaccine policy for industrialized countries. A cohort study conducted among subjects in CDC's Adult/Adolescent Spectrum of HIV Disease Project16 found that subjects with CD4 cell counts greater than 0.20 × 109/L at the time of pneumococcal vaccination had a 41% (95% CI, 10%-61%) decreased risk of pneumococcal pneumonia compared with those who were unvaccinated.17 That study also found a 25% decreased risk of nonspecific pneumonia among vaccinated persons (95% CI, 16%-33%). Likewise, immunologic data are not conclusive. Clearly, HIV-infected patients (particularly late in the course of immunosuppression) have reduced total and functional antibody responses to pneumococcal vaccine2; however, immunologic protection is complex, and it is not possible to define the degree of protection elicited by immunization with existing immunologic methods in the absence of well-defined correlates of protection.
This study provides additional support for a conclusion that 23-valent pneumococcal vaccine protects against invasive pneumococcal disease among HIV-infected adults in the United States. We found a vaccine effectiveness of 49% (95% CI, 12%-70%) against invasive pneumococcal disease among all patients after adjusting for other factors. Postlicensure case-control and indirect cohort studies in immunocompetent patients have demonstrated 56% to 81% effectiveness of the vaccine in preventing pneumococcal bacteremia.4,5,18,19 The point estimate for effectiveness in this group of HIV-infected patients was not substantially lower than that shown in immunocompetent patients.
Neither our study nor the previous investigations that estimated pneumococcal polysaccharide vaccine effectiveness among HIV-infected persons14- 16 were conducted among people who benefited from highly active antiretroviral therapy. Our study enrolled subjects during 1992 to 1995, and the majority of case and control subjects were not taking antiretroviral therapy. In the United States, use of triple combination antiretroviral therapy among HIV-infected persons increased markedly after 1996.16 Our findings may therefore underestimate the performance of pneumococcal vaccine in patients who are receiving newer antiretroviral therapy.
We cannot explain the disparity we observed between white and black patients. Systematic differences between cases and controls in access to health care would bias vaccine effectiveness calculations if one group was more likely to be evaluated in a clinic setting (where vaccine might be offered) or to receive prophylactic antibiotics (such as trimethoprim-sulfamethoxazole) or antiretroviral therapy (which could provide nonspecific protection against bacterial infections by improving immunologic status). However, none of these circumstances appears to have occurred in this study. We matched controls to cases by hospital of admission to partially control for access to preventive health services. We chose controls among hospitalized rather than outpatient populations to minimize the possibility that control patients would systematically be more likely to be offered preventive health services. We restricted enrollment to patients whose HIV status was known before hospitalization to ensure that all cases and controls had an indication for vaccination before study enrollment. However, white patients were more likely than blacks to have AIDS at the time of hospitalization, as well as a longer interval from HIV diagnosis to enrollment in this study. These factors may have resulted in a greater opportunity to be immunized for whites. The shorter interval between HIV diagnosis and hospitalization (ie, study enrollment) among black patients may reflect delayed diagnosis of HIV or more rapid disease progression among blacks; either factor could potentially result in blacks receiving pneumococcal vaccine at a time of more advanced immunocompromise, which could translate to poorer immunologic responses. Given the small sample sizes and the wide CIs, it is not possible to know whether findings from the study represent actual differences in vaccine effectiveness for the 2 groups. However, the finding most likely reflects advanced immunocompromise among blacks in our study and underscores the continued need to improve access to HIV therapeutic and preventive services among minority populations. This is particularly relevant, since black persons have an increased risk for invasive pneumococcal infection compared with whites,20 although the basis for this difference is not known.
The optimal way to evaluate vaccine efficacy is by means of randomized, blinded, prospective studies. However, for ethical reasons, such studies are not often possible to conduct in the postlicensure period, particularly once vaccine is recommended for use in the population that would be studied. Even if such a study is ethically acceptable, it is often not affordable or feasible to conduct a prospective study large enough to answer the question, especially when evaluating a subpopulation. Case-control studies are an affordable alternative. Linked to a system for case surveillance and with a case definition highly specific for disease (such as invasive pneumococcal disease), enough cases can be found (particularly if disease incidence is high) to provide sufficient study power. The risk is that this method will introduce bias. We did not find evidence that cases and controls in this study were unbalanced regarding crucial factors that would have an impact on the ascertainment of vaccination status. However, controls were more immunosuppressed than cases, as evidenced by the higher proportion of control patients with CD4 lymphocyte counts less than 0.10 × 109/L (<100 cells/mm3) or with a diagnosis of AIDS at the time of hospitalization. Despite our efforts to adjust for this disparity in the multivariable analyses, residual confounding may have occurred.
This study also confirmed that smoking cigarettes is a risk factor for invasive pneumococcal disease. Cigarette smoking has been associated with chronic respiratory disease and a variety of infections acquired through the respiratory route, including legionellosis21 and meningococcal disease.22 Close contact with young children was also identified as a risk factor for invasive pneumococcal disease. A substantial proportion (30%-100%) of young children are colonized with S pneumoniae, with the highest colonization rates occurring among children attending day care.23,24 Young children may then transmit pneumococci to susceptible adults. A recent study of the epidemiology of invasive pneumococcal disease among nonimmunocompromised, nonelderly adults found both smoking and household exposure to children who participate in day care to be strong independent risk factors for the development of invasive pneumococcal disease.25
We could not evaluate the role of CD4 cell count or stage of HIV infection in increasing the risk of pneumococcal disease in this study, because stage was a matching criterion; however, other studies14,26 have suggested that the risk of pneumococcal disease is proportional to degree of immunosuppression among HIV-infected patients. Although we did not find differences in the point estimates of vaccine effectiveness for patients with low and relatively high CD4 cell counts at the time of study enrollment, this study was not designed to evaluate this question and did not have sufficient power to do so. By restricting our controls to patients hospitalized for conditions other than pneumococcal infection but known to have HIV before hospitalization, we may have biased selection to controls who were more immunosuppressed, because pneumococcal infection is one of the earliest manifestations of HIV infection.2 This effect could have biased our findings. For instance, an imbalance in the study such that controls were more immunosuppressed than cases may have led to controls having greater opportunity to be vaccinated (bias toward finding effectiveness). We did not have information regarding the CD4 cell count or other immunologic parameters at the time of immunization among vaccinated cases and controls. Nonetheless, patients with low CD4 cell counts have low total and functional antibody responses to 23-valent pneumococcal vaccine.2
The paradoxical finding that lower CD4 cell counts appeared to be protective against invasive pneumococcal disease in our study bears explanation. This finding is likely an artifact of the method of control selection, which restricted enrollment to subjects who were hospitalized for conditions other than pneumococcal disease and had HIV infection diagnosed before admission. This design was selected to ensure that both case and control patients had an indication for pneumococcal vaccine and therefore opportunity to be immunized before hospitalization and to avoid overmatching for preventive services that might be associated with identifying controls based on outpatient source of care. We attempted to control for CD4 cell count on admission, but the lowest category (<0.20 × 109/L) included persons with a range of immunocompromise. Invasive pneumococcal disease can occur throughout HIV infection; in contrast, control patients tend to have opportunistic infections that present only at advanced stages of immunosuppression.
The findings of this study support continued use of pneumococcal vaccine among HIV-infected patients in the United States as recommended by the Advisory Committee on Immunization Practices. Methods to improve vaccine-induced protection, such as vaccination early in the course of HIV infection or following improvement of immune status resulting from antiretroviral therapy, should be considered and explored. It remains to be seen whether alternative vaccination strategies utilizing pneumococcal capsular polysaccharide protein-conjugate vaccines given sequentially with 23-valent pneumococcal vaccine will offer increased protection for this population27 or whether 23-valent vaccine alone will provide higher effectiveness when given following immune reconstitution to patients who benefit from highly active antiretroviral therapy.
Accepted for publication April 5, 2000.
We thank Gretchen Rothrock, MPH, Wendy Baughman, MSPH, and Georgia Jackson, BSN, MPH, for careful collection of data, Laura Conn, MPH, for data management, and Richard Facklam, PhD, for serotyping the pneumococcal isolates.
Reprints: Robert F. Breiman, MD, Centers for Disease Control and Prevention, 1600 Clifton Rd, MS C-12, Atlanta, GA 30333 (e-mail: firstname.lastname@example.org).