Although annual influenza vaccination is recommended for persons who are infected with human immunodeficiency virus (HIV), data are limited regarding the epidemiology of influenza or the effectiveness of influenza vaccination in this population. We investigated a 1996 outbreak of infection with influenza A at a residential facility for persons with AIDS. We interviewed 118 residents and employees, reviewed 65 resident medical records, and collected serum samples for measurement of influenza antibody titers. After controlling for history of smoking, influenza vaccination, and resident or employee status, in a multivariate model, HIV infection was not statistically associated with influenza-like illness (ILI). Symptoms and duration of ILI were similar for most HIV-infected and HIV-uninfected persons. However, 8 (21.1%) of 38 HIV-infected persons with ILI (vs. none of 15 HIV-uninfected persons) were either hospitalized, evaluated in an emergency room, or had ILI lasting ⩾14 days (P = .06). Vaccination effectiveness (VE) was similar for HIV-infected and HIV-uninfected persons. Vaccination was most effective among HIV-infected persons with CD4 cell counts of >100 cells/µL (VE, 65%; 95% CI, 36%–81%) or HIV type 1 virus load of <30,000 copies/mL (VE, 52%; 95% CI, 11%–75%). Providers should continue to offer influenza vaccination to HIV-infected persons.
Between 650,000 and 900,000 persons in the United States are infected with HIV [1], and >320,000 persons have AIDS [2]. Influenza affects 7%–10% of the adult population annually, causing an average of 20,000 influenza-related deaths, primarily among high-risk populations [3]. Few studies address the incidence or severity of influenza in HIV-infected persons, although one investigation reported an increased rate of hospitalization and death attributable to influenza among HIV-infected women [4]. Despite limited data, the Advisory Committee on Immunization Practices considers HIV-infected persons to be at increased risk for influenza and recommends annual vaccination [3, 5–8].
The effectiveness of influenza vaccination in HIV-infected persons is not well established [9], and some HIV-infected patients have a diminished antibody response to influenza vaccination [10–17]. Moreover, some investigators have argued that vaccination could harm HIV-infected persons, because vaccine-induced antigenic stimulation might promote HIV replication [7, 18–20]. Some investigators have observed transient increases in HIV loads during the first 1–3 weeks after influenza vaccination [7, 18, 19, 21–23], but others have not [17, 24–27]. The importance of these increases is unknown, and changes in HIV viremia after natural influenza infection have not been well characterized [18, 28].
On 3 December 1996, the New York City Department of Health (NYCDOH) and the Centers for Disease Control and Prevention (CDC) were notified of an outbreak of infection with influenza A among residents and employees of a residential substance abuse treatment facility for persons with AIDS (facility A). We initiated an epidemiologic investigation, to determine whether HIV-infected persons, compared with HIV-uninfected persons, were at increased risk for developing influenza or more severe illness; whether influenza vaccine was effective; and whether amantadine therapy was effective and well tolerated in HIV-infected persons.
In late November 1996, an outbreak of acute respiratory illness was detected at facility A. On 29 November, 3 residents had influenza A diagnosed on the basis of rapid antigen tests; influenza A(H3N2) virus was cultured subsequently. On 30 November, amantadine therapy was recommended for all residents.
The 65 residents in facility A move freely through the building and eat in a common dining room; some spend time in the surrounding community. Comprehensive primary medical care is provided on site. Influenza vaccine was administered to most residents and employees during the autumn of 1996, but not to selected residents with advanced immunodeficiency or very low CD4+ cell counts. Residents and employees who were vaccinated before 7 November received a vaccine that was recalled subsequently by the manufacturer because of decreasing potency of the A(H3N2) component [29]. After the recall, the facility switched to a different manufacturer's vaccine but did not revaccinate persons who had received the recalled vaccine.
All residents and employees were invited to participate in the study. From 5 December to 12 December 1996, interviewers administered a standard questionnaire regarding respiratory illness, underlying illnesses, influenza vaccination history, and history of smoking. All residents' medical records were reviewed by use of a standardized data collection instrument. Supplementary data on influenza vaccination status, employee job descriptions, and numbers of hours worked per week were obtained from the facility. All participants were asked to provide paired serum samples for influenza antibody testing; those with respiratory symptoms were asked to provide serial nasopharyngeal swab specimens for viral culture and rapid antigen testing.
Definitions. A case of influenza-like illness (ILI) was defined as a documented temperature of ⩾37.8°C (⩾100°F), among residents, or a reported temperature ⩾37.8°C, among employees, and presence of either cough or sore throat occurring from 13 November to 5 December 1996. Participants with ⩾1 respiratory symptom during this period (i.e., cough, sore throat, or rhinitis) and laboratory confirmation of influenza A infection were also considered to be case patients. Symptoms were defined by patient report or by complaints documented in the medical record. A 4-fold increase in antibody titer (by either hemagglutinin inhibition testing or complement fixation) or isolation of influenza A from a nasopharyngeal swab specimen were considered to be laboratory confirmation of influenza infection.
Laboratory investigation. Rapid antigen testing, virus isolation, and subtyping were done at the virology laboratory at Montefiore Medical Center. Virus isolates were forwarded to the CDC for further strain characterization. Paired blood samples were obtained on 5–6 December, and again on 18–20 December. Serum samples were tested by use of complement fixation for influenza A and B viruses, respiratory syncytial virus, adenovirus, parainfluenza, and mycoplasma at New York State Department of Health (NYSDOH), and by use of hemagglutinin inhibition tests for antibody to 1996–1997 influenza vaccine strains—A/Nanchang/933/95(H3N2), A/Texas/36/91(H1N1), and B/Harbin/07/94—and the outbreak strain, A/New York/83/97(H3N2), at CDC. Serum samples were considered to be acute if they were drawn within 7 days of developing ILI and convalescent if they were drawn 2–6 weeks later.
Plasma HIV-1 RNA levels, measured by means of quantitative RT-PCR assay (Amplicor HIV-1 Monitor kit; Roche Molecular Systems) [30] in specimens obtained from 23 October to 5 December 1996, were obtained from residents' medical records. The most recent CD4+ lymphocyte counts were recorded from the medical record for residents and were self-reported by HIV-infected employees.
Statistical methods. Bivariate analyses were done by means of either Mantel-Haenszel χ2 test or Fisher's exact test. P ⩽ .05 was considered to be statistically significant. We constructed a multivariable logistic regression model that included influenza vaccination status and variables significantly associated with ILI. Hypotheses associated with the logistic regression were tested by use of 1-tailed P values calculated with exact methods [31]. Geometric mean antibody titers and plasma HIV-1 RNA levels were analyzed by use of a nonparametric Wilcoxon rank sum test. Persons who had received an influenza vaccination <2 weeks before onset of ILI were excluded from analyses with vaccination as a variable. Vaccine effectiveness was calculated as (ARUV − ARV)/ARUV (where ARUV and ARV are the attack rates among unvaccinated and vaccinated persons, respectively). All analyses were done by use of Epi Info, version 6.0 (CDC) [32] or SAS, version 6.12 (SAS) [33].
Participation and description of study population. Fifty (77%) of 65 residents and 68 (60%) of 113 employees were interviewed. The age and sex distributions of residents and participating employees were similar. The proportion of residents who were current smokers (88%) was significantly higher than the proportion among employees (22%; P < .01). All residents had AIDS, as defined by CDC criteria [34]. Eight (12%) employees reported being infected with HIV, of whom 5 had AIDS (table 1). Seventy-three participants provided paired serum samples, and 15 provided nasopharyngeal specimens for influenza testing.
Demographic and clinical characteristics of residents and employees involved in an outbreak of infection with influenza A at facility A, New York City, 1996.
Predominant risk factors for HIV infection among residents were injection drug use (65%) and heterosexual sex (25%). Most residents (92%) had CD4+ counts of <300 cells/µL (median, 149 cells/µL; range, 6–437 cells/µL). More than 70% of residents had started receiving antiretroviral therapy ⩾6 months before the outbreak of infection, but changes in antiretroviral therapy were made frequently during the 6 months that immediately preceded the influenza outbreak. At the time of the outbreak, 85% of residents were receiving reverse-transcriptase inhibitors, 45% were receiving protease inhibitors, and 39% were receiving 3 anti-HIV medications.
Description of the outbreak of infection. The outbreak of infection began on 13 November and peaked on 29 November (figure 1). On 30 November, influenza A was diagnosed, and amantadine therapy was instituted among residents. No new cases occurred after 2 December. Thirty-four (52%) of 65 residents and 19 (28%) of 68 employees developed ILI.
Cases of influenza-like illness (ILI) among residents and employees of facility A, by date of onset, New York City, November 1996 (n = 52). Cases of ILI were defined as either temperature of ⩾37.8°C (⩾100°F) and cough or sore throat or as ⩾1 respiratory symptom (i.e., cough, sore throat, or rhinitis) and laboratory confirmation of infection with influenza A by means of culture or serologic tests. One case had an unknown date of onset.
Overall, 73 participants were tested for influenza, and 22 (30%) had laboratory-confirmed influenza A infection; 21 cases were confirmed by a 4-fold increase in titer (12 by means of hemagglutinin inhibition only, 4 by means of complement fixation only, and 5 by means of both methods), and 1 was confirmed by a positive culture result only. Thirty-one tested participants had ILI, and 20 (65%) of them tested positive for influenza. Two (10%) serologically confirmed participants were residents (i.e., they had AIDS) with no ILI symptoms. Nasopharyngeal specimens obtained from 3 residents (including 2 who had 4-fold increases in hemagglutinin inhibition titer) had positive results of a rapid antigen test and yielded influenza A(H3N2) isolates that were antigenically similar to the H3N2 component of the 1996–1997 vaccine, A/Nanchang/939/95. No other respiratory pathogen was implicated by complement fixation tests or virus culture as a potential cause for the outbreak of infection.
Clinical characteristics and severity of ILI. Symptoms among participants with ILI were similar, regardless of HIV status (table 2). However, 8 HIV-infected persons (21.1%) with ILI (vs. none of the 15 HIV-uninfected participants with ILI) required hospitalization, required evaluation in an emergency room, or experienced prolonged illness (P = .06). All 8 HIV-infected persons with more severe ILI were current smokers, and 3 (38%) had asthma or diabetes. The median CD4+ cell count (191 cells/µL; range, 25–394 cells/µL) and HIV-1 RNA level (geometric mean titer, 7295 copies/mL; range, 200–47,510 copies/mL) in HIV-infected participants with more severe ILI did not differ significantly from those with milder ILI.
Clinical characteristics of persons with influenza-like illness (ILI), according to HIV status, in an outbreak of infection with influenza A at facility A, New York City, 1996.
Risk factors for ILI. In a combined analysis of residents and employees, current cigarette smoking, HIV infection, and resident status were significantly associated with ILI (table 3). When employees were analyzed separately, neither HIV-AIDS nor smoking were significantly associated with ILI. Among residents, only smoking was associated with ILI. The median CD4+ cell count, the mean preoutbreak HIV-1 RNA level (measured during the fall of that year), and the median number of previous opportunistic illnesses among residents who developed ILI were similar to those who did not.
Attack rates of influenza-like illness (ILI) among persons involved in an outbreak of infection with influenza A in New York City in 1996, according to selected risk factors.
Multivariate logistic regression indicated that current smoking was the strongest predictor of ILI in the combined study population (adjusted OR, 9.11; 95% CI, 3.18–26.10), followed by lack of influenza vaccination (adjusted OR, 2.00; P = .06). HIV infection and status as employee or resident were not independently associated with ILI (adjusted OR, 1.76 and 1.23; P = .37 and P = .54, respectively).
Effectiveness of influenza vaccination. Influenza vaccination was effective in preventing ILI in HIV-infected and HIV-uninfected persons (27% vs. 39%; table 4). Vaccine was 32% effective for prevention of symptomatic laboratory-confirmed influenza among HIV-infected persons. In the multivariate analyses, vaccination was significantly protective only among smokers, regardless of HIV status (adjusted OR, 0.28; 95% CI, 0.10–0.81). The number of HIV-infected persons in this study was too small to confirm that influenza vaccine was protective against hospitalization and emergency room evaluation (22% among unvaccinated persons vs. 5% among vaccinated persons; RR, 4.4; P = .153) or against illness with a duration of ⩾14 days (28% among unvaccinated persons vs. 5% among vaccinated persons; RR, 5.6; P = .078). The median durations of illness and fever were lower among HIV-infected vaccinated persons than they were among unvaccinated persons, although the difference was not significant (median duration of illness, 6.0 vs. 8.5 days; P = .053; median duration of fever, 2.0 vs. 3.0 days; P = .138).
Attack rates (AR), by subgroup, for influenza-like illness (ILI) among unvaccinated (UV) and vaccinated (V) persons and vaccine effectiveness among persons involved in an outbreak of infection with influenza A in New York City, 1996.
Vaccine was effective in preventing ILI among HIV-infected participants with CD4+ cell counts of >100 cells/µL and among residents who had preoutbreak HIV-1 RNA levels of <30,000 copies/mL (table 5). The subgroup with both CD4+ cell counts of >100 cells/µL and virus loads of <30,000 copies/mL had the highest vaccine effectiveness.
Attack rates and vaccine effectiveness, by CD4+ cell count and HIV-1 load, for influenza-like illness among unvaccinated (UV) and vaccinated (V) HIV-infected persons involved in an outbreak of infection with influenza A, New York City, 1996.
Tolerance and therapeutic effectiveness of amantadine. Of the 54 residents who started prophylactic (n = 27) or therapeutic (n = 27) amantadine, 41 (76%) completed a 14-day course. Residents who took amantadine for treatment of ILI within 48 h after onset of illness had a slightly shorter duration of fever and illness, compared with residents who did not receive amantadine or residents who started it >48 h after the onset of illness (median duration of fever, 2 vs. 3 days; P = .059; median duration of illness, 6.0 vs. 7.5 days; P = .074).
Among persons who received prophylactic amantadine, 9 (33%) reported ⩾1 CNS complaint (most commonly confusion, insomnia, or depression) and 4 (15%) reported minor gastrointestinal symptoms. One person with a preexisting seizure disorder and a subtherapeutic phenytoin level received amantadine for ILI and had a generalized seizure on the second day of therapy.
This investigation suggests that there is an increased risk of prolonged illness, emergency room visits, and hospitalization from influenza among HIV-infected persons. Most HIV-infected and HIV-uninfected persons in this cohort had similar symptoms and duration of ILI. However, 13% of HIV-infected persons with ILI had prolonged illness, compared with none of the ill HIV-uninfected persons. Similarly, 8% of HIV-infected persons required emergency room evaluation, and 5% were hospitalized for ILI, in contrast to none of the HIV-uninfected persons. The percentage of HIV-infected persons requiring hospitalization was ∼100-fold higher than would have been expected among healthy persons of the same age [35–37].
Our findings are consistent with published reports that suggest that influenza may be more severe among HIV-infected persons [38–45], as with other immunocompromised persons [44, 46–50]. A population-based review of Medicaid data indicated that HIV-infected women were at increased risk for hospitalization and death from cardiopulmonary disease during influenza season [4].
No large epidemiologic studies have addressed the question of whether the incidence of influenza or ILI is increased in HIV-infected persons. Investigators of a 1988 outbreak of infection with influenza A(H1N1) at a similar facility [51] found that none of 7 HIV-positive residents developed ILI, compared with 14 (37%) of 38 HIV-negative residents, and concluded that HIV infection does not increase the risk of ILI.
During this outbreak of infection, we did not observe a significant increase in the incidence of ILI among HIV-infected persons. We also did not find a direct relationship between the degree of immunosuppression among HIV-infected persons, as measured by CD4+ cell count and plasma HIV-1 RNA levels, and the incidence of ILI. Although the risk of ILI was higher in HIV-infected persons than in HIV-uninfected persons by means of univariate analyses, in multivariate analyses, current smoking was the strongest risk factor for ILI, as observed in other investigations [52–56], and accounted for the apparent increase in risk for ILI among HIV-infected participants. We observed increased vaccine effectiveness among smokers. Our findings suggest that smokers should be targeted for influenza vaccination.
We found that influenza vaccination was effective in preventing illness among persons with HIV infection and AIDS. Among HIV-infected persons with ILI, influenza vaccination also appeared to reduce the durations of illness and fever, as well as the risk of visiting an emergency room or being hospitalized. Although the latter findings were not statistically significant, they are noteworthy because they are similar to vaccination effects observed in other high-risk populations, (e.g., elderly persons) for whom influenza vaccination is more effective for the prevention of pneumonia, hospitalizations, and death (50%–80%) than ILI (30%–40%) [3].
Only one randomized controlled trial has addressed the effectiveness of influenza vaccination in preventing illness in HIV-infected persons [57]. This study found that vaccination protected against respiratory illness and laboratory-confirmed symptomatic influenza. Some, but not all, previous investigations have shown diminished antibody response to vaccination among persons with HIV infection and AIDS, compared with HIV-uninfected vaccinees [10–12, 15–17, 58–60].
In this outbreak, vaccination was more effective in preventing ILI among HIV-infected persons with CD4+ cell counts of >100 cells/µL and among those with an HIV-1 RNA load of <30,000 copies/mL, which is consistent with previous studies that have reported that serologic response to influenza vaccine correlates directly with CD4+ cell count in HIV-infected hosts [10, 22] and with CD4 : CD8 ratio [17] and inversely with stage of disease [11, 12, 15].
Data on the impact of vaccination on HIV load have been conflicting [7, 8, 17–19, 21, 24–28]. Some investigators have reported increases in HIV RNA levels after influenza vaccination, but most have reported a transient effect, with HIV loads generally peaking within 1–2 weeks of vaccination. The clinical significance of such findings is unclear. Natural infection with influenza may also trigger increases in HIV load [18, 20, 28].
To our knowledge, our investigation is the only study that evaluates the use of amantadine therapy in persons with HIV or AIDS. Although the outbreak of infection may have been waning by the time amantadine was offered to residents, it came to an abrupt halt 2 days after amantadine prophylaxis was started. Antiviral effectiveness in preventing illness in individual patients with HIV infection was difficult to evaluate but appeared to reduce illness severity when taken early. Prophylactic amantadine was associated with a relatively high rate of incidence of minor side effects (37%), but no unusual serious adverse effects were observed.
This outbreak of infection provided an opportunity to evaluate many unanswered questions regarding influenza in HIV-infected persons, but the small study population and the retrospective nature of the investigation were limitations. The clinical case definition for ILI is nonspecific, and not all cases of ILI were laboratory confirmed. Therefore, some respiratory illnesses may have been misclassified. However, no other respiratory pathogen was implicated in the outbreak, and misclassification of respiratory illnesses as influenza would have biased our vaccine effectiveness estimates toward the null. Furthermore, our data may have underestimated the effectiveness of vaccination among HIV-infected persons, because the vaccine that was used was recalled voluntarily during the autumn of 1996 for decreasing potency of the H3N2 component. Finally, almost all the HIV-infected patients in the study had AIDS, CD4+ cell counts of <300 cells/µL, and access to good medical care, and so our findings may not be generalizable to all persons with HIV infection.
The incidence of influenza was not increased among persons with HIV infection in this study, but a substantial proportion of persons had prolonged illness or required hospitalization. Influenza vaccination prevented illness among those persons with CD4+ cell counts of >100 cells/µL or HIV-1 loads of <30,000 copies/mL and likely reduced illness severity among HIV-infected persons. Although larger prospective studies of HIV and influenza are needed, we urge clinicians to continue to offer influenza vaccination annually to persons with HIV infection.
We thank the residents and employees of Project Samaritan AIDS Services Incorporated (PSASI) for participating in this study and the following people who contributed to the investigation: John Cavallero (PSASI, Bronx, New York); Marcelle Layton, Steve Cato, Andrea Young, Anne Labowitz, Jose Poy, and Denise Kenney (Communicable Disease Program, NYCDOH), and Igbal Poshni and Alex Ramon (Bureau of Laboratories, NYCDOH); Nancy Cox, Hector Izurieta, Lynnette Brammer, and Nancy Arden (Influenza Branch, CDC), Jon Kaplan, Louisa Chapman, and Tom Folks (Division of Viral and Rickettsial Diseases, National Center for Infectious Disease, CDC), Laura Fehrs (State Branch, Division of Field Epidemiology, CDC), and Meade Morgan (Division of Public Health Surveillance and Informatics, Epidemiology Program Office, CDC, Atlanta); Ilya Spiglan (Montefiore Medical Center, New York); and Bruce McReedy (LabCorps, Research Triangle Park, New Jersey).
Appropriate informed consent was obtained, and clinical research was conducted in accordance with guidelines for human experimentation, as specified by the US Department of Health and Human Services. The research protocol was approved by institutional review boards at the New York City Department of Health and at the Centers for Disease Control and Prevention.
IDSA Members: For your free access to this journal, log in via the IDSA members area.
Open access options for authors visit Oxford Open
This journal enables compliance with the NIH Public Access Policy