Comparison of the Outcomes of Individuals With Medically Attended Influenza A and B Virus Infections Enrolled in 2 International Cohort Studies Over a 6-Year Period: 2009–2015

Abstract Background Outcome data from prospective follow-up studies comparing infections with different influenza virus types/subtypes are limited. Methods Demographic, clinical characteristics and follow-up outcomes for adults with laboratory-confirmed influenza A(H1N1)pdm09, A(H3N2), or B virus infections were compared in 2 prospective cohorts enrolled globally from 2009 through 2015. Logistic regression was used to compare outcomes among influenza virus type/subtypes. Results Of 3952 outpatients, 1290 (32.6%) had A(H1N1)pdm09 virus infection, 1857 (47.0%) had A(H3N2), and 805 (20.4%) had influenza B. Of 1398 inpatients, 641 (45.8%) had A(H1N1)pdm09, 532 (38.1%) had A(H3N2), and 225 (16.1%) had influenza B. Outpatients with A(H1N1)pdm09 were younger with fewer comorbidities and were more likely to be hospitalized during the 14-day follow-up (3.3%) than influenza B (2.2%) or A(H3N2) (0.7%; P < .0001). Hospitalized patients with A(H1N1)pdm09 (20.3%) were more likely to be enrolled from intensive care units (ICUs) than those with A(H3N2) (11.3%) or B (9.8%; P < .0001). However, 60-day follow-up of discharged inpatients showed no difference in disease progression (P = .32) or all-cause mortality (P = .30) among influenza types/subtypes. These findings were consistent after covariate adjustment, in sensitivity analyses, and for subgroups defined by age, enrollment location, and comorbidities. Conclusions Outpatients infected with influenza A(H1N1)pdm09 or influenza B were more likely to be hospitalized than those with A(H3N2). Hospitalized patients infected with A(H1N1)pdm09 were younger and more likely to have severe disease at study entry (measured by ICU enrollment), but did not have worse 60-day outcomes.

studies were limited by sample size; the outpatient studies only enrolled from 1 region.
To compare demographics, clinical characteristics and follow-up outcomes for adults infected with influenza A(H1N1) pdm09, A(H3N2), or B viruses, data from 2 prospective, concurrent, multicenter inpatient and outpatient cohorts were analyzed. The studies enrolled in northern and southern hemisphere sites from 2009 through 2015, with laboratory confirmation of influenza virus type/subtype, thus allowing comparison of infections over multiple seasons and locations.

MATERIALS AND METHODS
In 2009, the International Network for Strategic Initiatives in Global HIV Trials (INSIGHT) initiated 2 international observational cohort studies for adults aged ≥18 years with influenza A(H1N1)pdm09 virus infection: FLU002 for outpatients with influenza-like illness (ILI) and FLU003 for individuals hospitalized with complications and enrolled from either the general ward or the ICU [7,8]. Protocols were amended to include other influenza viruses (June 2011), and in October 2013 FLU003 was amended to remove the eligibility requirement of having severe or complicated influenza at enrollment. Some sites chose to enroll only outpatients or inpatients. Institutional review boards approved the protocols, and individual signed informed consent was obtained.
FLU002 outpatients had 14-day follow-up for death and hospitalization. FLU003 inpatients had a 60-day follow-up, with disease progression outcomes defined differently for general ward (death, hospitalization ≥28 days, ICU admission, or use of mechanical ventilation during follow-up) and ICU enrollments (death or hospitalization ≥28 days).

Statistical Analysis
Logistic regression was used to compare risk of disease progression outcomes among influenza type/subtypes. Crude and adjusted odds ratios (ORs) and 95% confidence intervals were derived using A(H1N1)pdm09 as the reference group. For FLU002 adjusted analyses, baseline covariates were age, gender, ethnicity, geographic region, symptom duration ≥6 days at enrollment, body mass index (BMI), smoking status, influenza vaccination in the season pre-enrollment (note that this study was not designed to assess vaccine efficacy), antiviral usage, and the presence of at least 1 of the following clinician-defined comorbidities: asthma/chronic obstructive pulmonary disease, immunosuppression, cardiovascular/liver/renal disease, or diabetes. Covariates significant in univariate analyses for FLU002 were retained for additional multivariable analysis in a parsimonious model. Covariates that remained significant in multivariable analysis were also considered for potential effect modification, ie, whether type/subtype differences varied by the significant predictor of outcome (for these, the interaction P value is cited).
FLU003 used the same covariates, along with the presence of at least 1 complication that initially defined enrollment eligibility. Crude and adjusted ORs were estimated for ward and ICU enrollments, separately and overall. For the latter, unadjusted and adjusted analyses were stratified according to ICU or general ward enrollment.
Sensitivity analyses were performed for both cohorts: 1 restricted to those enrolled post-April 2012 when all 3 viruses were co-circulating (to minimize confounding that might arise from combining analyses across seasons) and a second restricted to sites that enrolled at least 1 patient with each type/ subtype. For FLU003, a sensitivity analysis was carried out for baseline and outcome variables in which viruses were compared for the time pre/post the October 2013 protocol amendment.
Following a significant (P < .05) omnibus test for virus differences in disease progression outcomes, pairwise comparisons of viruses were considered significant if the (2-sided) P value was less than .05. To reduce type 1 error risk for type/subtype comparisons for baseline characteristics, a P value for the omnibus test of .01 was used to assess the significance (P < .01) of pairwise comparisons. Analyses used SAS v. 9.4 (Statistical Analysis Software, Cary NC).   Figure S2). One thousand one hundred eighty-six (84.8%) were enrolled in the general ward, and 212 (15.2%) were enrolled in the ICU.

FLU002 Outpatient Cohort Baseline Characteristics
The median age of outpatients with A(H1N1)pdm09 was 35 years, compared with 39 years for those with A(H3N2) (P < .0001) and 42 years with B (P < .0001; P < .0001 also for A(H3N2) vs B). Most outpatients were enrolled within 3 days of ILI onset. This varied significantly among types/subtypes (P < .0001): More influenza B-infected outpatients were enrolled 4-5 days or ≥6 days from symptom onset than A(H1N1)pdm09 (P < .0001) or A(H3N2) (P < .0001) infections. Comorbidities were common, and more frequent in influenza B virus infection than A(H1N1)pdm09 (P = .0009) or A(H3N2) (P = .01). This difference was largely due to a greater percentage of immunosuppressed outpatients with influenza B (P = .0002) ( Table 1). Age differences among type/subtypes persisted (P < .0001) after adjustment for other factors in Table 1; differences in the prevalence of comorbidities were reduced (P = .09 after adjustment). Tables 1 and Supplementary Table S1 summarize baseline characteristics for the 1398 hospitalized patients, which were similar  for ward (1186) and ICU (212) enrollments. The median time from hospitalization to enrollment was 2 days for ward and 3 days for ICU enrollments. ICU patients included those admitted directly or general ward transfers (74/212, 34.9%). Compared with FLU002 outpatients, FLU003 inpatients were older with more underlying comorbidities (Table 1). Among hospitalizations, those with A(H1N1)pdm09 were younger (49 years) than A(H3N2) (67 years; P < .0001) or influenza B (59 years; P < .0001). The time from symptom onset to hospitalization and enrollment varied significantly among types/subtypes (P < .0001), with more A(H3N2) patients enrolled within 6 days of symptoms onset than with A(H1N1)pdm09 (P < .0001) or B (P = .0003). Differences in age among type/subtype persisted (P < .0001) after adjustment for other factors in Table 1.

FLU003 Inpatient Cohort Baseline Characteristics
Percentages of hospitalized patients with comorbidities varied by virus (P < .0001). More patients with A(H3N2) (P < .0001) or B (P = .007) had comorbidities than those with A(H1N1)pdm09. This partially reflected age differences; after age adjustment, a significant difference remained in the percentage with comorbidities between A(H1N1)pdm09 and A(H3N2) (P = .0009), but not between A(H1N1)pdm09 and B (P = .36).
With adjustment for all of the factors in Table 1, the difference among type/subtypes in the prevalence with comorbidities was no longer significant (P = .16).
Overall, and among hospital ward-enrolled patients, the percentage with complications varied among types/subtypes (P < .0001): More inpatients with A(H1N1)pdm09 had complications compared with those with A(H3N2) (P < .0001) or B (P < .0001) (Supplementary Table S2). In part, the greater percentage of A(H1N1)pdm09 inpatients with complications arose because the initial A(H1N1)pdm09-focused protocol required at least 1 complication to be present at enrollment. When this requirement was removed in October 2013, there were no significant differences in percentages of patients with complications by virus type/subtype. Suspected bacterial pneumonia was more common among all patients with A(H1N1)pdm09 virus infection (P = .001) (Supplementary Table S2).
Of covariates considered in adjusted analyses, older age (P < .01), the presence of ≥1 comorbidities (P = .002), and continent of enrollment (Australia/Asia, North America, Europe, South America; P < .0001) were significantly related to death or hospitalization in both univariate and fully adjusted analyses. For each of the subgroups defined by these baseline factors, outpatients with A(H1N1)pdm09 were more likely to be hospitalized. Among older outpatients, the difference among type/ subtypes was reduced, leading to a significant age interaction (P = .03) (Table S3).

FLU003 Inpatient Cohort Day-60 Composite Outcomes
Day 60 disease progression outcomes were known for 1291 (92.3%) inpatients (Supplementary Figure S2): The percentage with a composite outcome did not vary by virus (Table 4). This was observed overall (P = .32) and for ICU (P = .80) or ward (P = .39) enrollments. With considerations of the components of the composite outcome, there was no difference in the number of patients transferred to the ICU from the ward (P = .24) while there was a trend toward more patients with A(H1N1) pdm09 having an extended (≥28 days) hospitalization (P = .07); the percentage dying within 60 days was 7.1% for A(H1N1) pdm09, 5.0% for A(H3N2), and 7.3% for B (P = .30). Of the 82 deaths, 64 (78.0%) occurred during the enrollment hospitalization. Mortality was higher in ICU (24.3%) than general ward enrollments (3.0%; P < .001).

DISCUSSION
In these 2 large international cohort studies of adults with laboratory-confirmed influenza, outpatients with A(H1N1)pdm09 were more likely to be hospitalized within 14 days than those with A(H3N2) or influenza B. Although hospitalized adults with A(H1N1)pdm09 were more likely to have severe influenza as measured by the percentage of ICU enrollments, differences in disease progression or mortality at 60-day follow-up were not detected. Notably, both outpatients and hospitalized patients with A(H3N2) or B had more comorbidities than those with A(H1N1)pdm09, and patients with A(H1N1)pdm09 were younger.
Among outpatients, the 14-day hospitalization risk was lowest for those with influenza A(H3N2). The percentage of outpatients with A(H3N2) subsequently hospitalized was 0.7%, compared with 3.4% and 2.2% for A(H1N1)pdm09 and B, respectively. The lower A(H3N2) hospitalization risk persisted after adjustment for age and other covariates, when analyses were done over influenza seasons with substantial virus co-circulation and for sites that enrolled at least 1 patient of each type/ subtype. The finding was also consistent for subgroups defined by age, continent of enrollment, and presence of comorbidities. Our results differ from studies that examined hospitalization within 30 days among enrollments in vaccine effectiveness studies in Wisconsin [5,6]. The differences may relate to variations in sample size and statistical power, the study time periods, inclusion of children, outcome definitions, or unmeasured confounders. In Belongia et al. [6], 6/150 (4.0%) adults with A(H1N1)pdm09 were hospitalized within 30 days of onset (contrasting with 14-day follow-up in FLU002) vs 17/377 (4.5%) for A(H3N2). However, outpatients with A(H1N1) pdm09 were enrolled in 2009, and those with A(H3N2) were enrolled in 2007-2008 (influenza B patients were not analyzed) [6], whereas we enrolled adults continuously from 2009 through 2015.
In FLU003, similar to US Centers for Disease Control and Prevention (CDC) studies [2,3], we found that inpatients with A(H1N1)pdm09 were more likely to be admitted to the ICU. Although more A(H1N1)pdm09 inpatients were enrolled from the ICU (and the percentage transferring to the ICU from the general ward was higher), this difference was not significant. If ICU enrollment and transfer from the ward are considered together, A(H1N1)pdm09 virus infection appeared to be associated with more severe disease.
The observation of increased severity of hospitalization with A(H1N1)pdm09 virus infection was not supported by the 60-day follow-up. Disease progression did not vary among types/subtypes irrespective of general ward or ICU enrollment. Furthermore, in neither the CDC studies [2,6] nor FLU003 was there evidence of mortality differences among types/subtypes. In Chaves et al. [2], the percentages of patients dying before hospital discharge were 3.7%, 3.9%, and 4.1%, for A(H1N1)pdm09, A(H3N2), and influenza B, respectively. In FLU003 inpatients, the corresponding percentages for death within 60 days were 7.1% (A(H1N1)pdm09), 5.0% (A(H3N2)), and 7.3% (B). Two other studies have reported mortality by virus type/subtype. Loubet et al. [4] reported that the percentage of patients dying during hospitalization did not vary by type/subtype (A(H1N1) pdm09 2%, A(H3N2) 4%, influenza B 4%), but the total number of deaths (n = 20) was much smaller than in FLU003 (n = 82) or Chaves et al. (n = 109) [2]. Similarly, in Hong Kong mortality was compared for adults hospitalized with influenza A vs B in 2007-2008 and found not to differ: 21/414 (5%) with A(H3N2) vs 13/215 (6%) with influenza B [9]. It is uncertain why outpatients with A(H1N1)pdm09 were more likely to be hospitalized, but hospitalized A(H1N1)pdm09-infected patients did not have worse 60-day outcomes (despite being more likely to be enrolled from the ICU). One explanation may be that younger adults hospitalized with A(H1N1)pdm09 virus infection, although less likely to have comorbidities, were more likely to be admitted to the ICU if complications were present. Strengths of our study include concurrent, prospective enrollment of both outpatients and inpatients from multiple countries in both hemispheres over 6 years. A defined follow-up period for assessing outcomes was used with a high proportion of follow-up evaluation, and common protocols were used for sample collection and laboratory testing. Limitations include adults-only enrollment and a relatively small number of disease progression outcomes in the outpatient cohort, thereby limiting power, an issue identified for many studies [10].
In summary, the findings from these global follow-up studies illustrate the benefit of planned prospective observational studies to complement influenza surveillance approaches. Such studies are critical for assessing predictors of disease progression including the changing severity among types/subtypes over time.