The FLU 002 and FLU 003 protocols were approved by the institutional review boards (IRB) or institutional ethics committees (IEC) at the University of Minnesota and at each of the participating clinical sites worldwide (see Appendix S1). Written documentation of IRB/IEC approval to each site Principal Investigator was a required element in the site registration process that preceded site activation as a study center. Copies of these approval letters are filed with the central coordinating center at the University of Minnesota. All patients (or proxy) gave signed informed consent prior to enrollment.

Enrolled outpatients with ILI were followed for 14 days for hospitalization or death. Henceforth for FLU 002 patients, this composite outcome is referred to as “disease progression”. At 14 days the resolution of symptoms was also assessed.

Enrolled hospitalized patients were followed for 60 days. For general ward patients, outcomes assessed included death, ICU admission and/or mechanical ventilation, or prolonged hospitalization; the latter was defined as an inpatient stay exceeding 28 days of the 60-day follow-up period, not necessarily consecutively. For patients enrolled after ICU admission, death or prolonged hospitalization for >28 days were the primary outcomes. For FLU 003 patients, this composite outcome, stratified according to whether patients were enrolled from a general ward or ICU, is referred to as “disease progression”. Length of hospitalization, resolution of symptoms, and resumption of normal activities were assessed at 28 and 60 days after enrollment.

In both studies, respiratory (nasal and oropharyngeal) swabs were collected at enrollment for influenza testing. The combined respiratory sample was sent to one of two central laboratories for influenza testing (SAIC Frederick, Inc, Maryland or Advanced BioMedical Laboratories, New Jersey) by reverse transcription polymerase chain reaction (RT-PCR) assay using specific primers and probes for detection of influenza A, (seasonal H1, H1N1pdm09, H3), and B viruses. In FLU 003, a local RT-PCR test result was required either prior to enrollment (for confirmed diagnoses) or at the time of enrollment (for suspected diagnoses). Initially, local RT-PCR test results were only recorded as influenza A positive or negative; after the first year, influenza A subtyping results were recorded. We assessed the discordance of local and central RT-PCR results. Results are shown in Appendix S2 with a rationale for inclusion of patients in each A(H1N1)pdm09 virus-infected cohort.

Outpatients enrolled with A(H1N1)pdm09 confirmed at the central laboratories are included in the FLU 002 cohort. The FLU 003 hospitalized cohort includes patients with A(H1N1)pdm09 virus infection confirmed at a central laboratory and patients who tested positive for influenza A by a local laboratory and negative for influenza A at a central laboratory during the initial 6 months of enrollment when A(H1N1)pdm09 virus was highly prevalent and the results of local RT-PCR testing did not record the influenza A subtype (see Appendix S2).

In a random subsample of 333 patients with A(H1N1)pdm09 virus infection, a tandem multiplex PCR (AusDiagnostics, Sydney Australia) was performed on upper respiratory specimens to estimate the prevalence of potential co-pathogens in each study [8]. These laboratory analyses were performed at the Centre for Infectious Diseases and Microbiology Laboratory Services, Westmead Hospital, Westmead, New South Wales, Australia.

Descriptive statistics were used to describe the characteristics of patients enrolled in the two cohort studies. Cross-sectional comparisons of patients in the two studies were performed to assess factors potentially contributing to disease severity: odds ratios (ORs) (hospitalized patients versus outpatients) and 95% confidence intervals (CIs) are cited. Unadjusted (univariable) and adjusted (multivariable) ORs are cited. Similar analyses were done for the subsample of patients for whom tandem multiplex PCR for other pathogens was performed.

The percentage of patients developing disease progression during follow-up was computed for each study. In addition, cumulative mortality for patients in FLU 003 is summarized with Kaplan-Meier plots. For these analyses, follow-up was censored at the end of follow-up (60 days) or the date of last contact (e.g., discharge or day 28) for those who did not complete the full follow-up. Logistic regression was used to study baseline predictors of disease progression and mortality. Prognostic factors for disease progression were determined separately for the two studies and for pooled data from the two studies. Unadjusted and adjusted ORs are cited along with 95% CIs and p-values. In expanded models, an interaction term (covariate x study) was included in the logistic model to assess whether associations with disease progression differed for FLU 002 and FLU 003.

Height and weight data, used to determine body mass index (BMI), were available for 91.0% of those enrolled. Date of onset of symptoms for ILI and smoking prevalence data were available for 98.7% and 99.2% of enrolled patients respectively. Other baseline covariate data were present for all patients. To minimize bias and increase power for multiple regression analyses that require complete covariate information for each patient, multiple imputation was used to predict values that were substituted for the missing data. The imputation was done in an iterative manner using the baseline covariate data available. The regression coefficients from five rounds of imputation were used to obtain the ORs. The imputation had little effect on the univariable analyses, therefore summary statistics from these analyses are based on the observed data. In a sensitivity analysis, a complete case analysis was performed and adjusted ORs were estimated for all of the baseline variables excluding BMI. Estimates similar to those based on multiple imputation were obtained (data not shown).

All statistical tests are two-tailed and p-values less than 0.05 were considered to indicate statistical significance. Statistical analyses were performed using SAS (Version 9.3).

In FLU 002, outpatients with A(H1N1)pdm09 virus infection were enrolled by 53 sites in 15 countries (see Acknowledgements for number enrolled by country). Asian sites enrolled 20.3% of patients; 4.1% of patients were from Australia; 46.1% from Europe; 8.0% from South America; and 21.5% from the United States. The median age of enrolled outpatients with A(H1N1)pdm09 virus infection was 30 years; those enrolled in the first year (2009–2010) had a median age that was 6 years younger than in subsequent periods (29 versus 35 years; p<0.001 for difference) (Table 1). Fifty-two percent of patients were female; 1.9% had a BMI of ≥40 kg/m²; 21% reported smoking; and 2% of the women aged ≤45 years of age were pregnant at the time of enrollment or within the previous two weeks. Median time from the onset of symptoms to enrollment was two days; for 75% of patients this time was three days or less. Fifty-five patients (9.3%) had HIV infection or other immune dysfunction; 50 of the 55 patients had HIV infection, reflecting the fact that many of the infectious disease clinics participating in FLU 002 cared for patients with HIV infection. Fifteen (2.5%) patients were prescribed influenza antivirals (all oseltamivir) in the 14 days prior to enrollment. On the day of enrollment, 28% of patients were prescribed antiviral treatment (data not shown).

Disease progression status at day 14 was available for 572 (96.9%) of enrolled patients in FLU 002. Twenty-nine patients (5.1%; 95% CI: 3.4–7.2%) experienced disease progression during the 14-day follow-up period; 28 (4.9%) required hospitalization and one patient died (Table 2). Of the 28 patients initially enrolled as outpatients who were subsequently hospitalized, 12 (42.9%) were admitted to the hospital later on the same day as study enrollment.

One hundred and five outpatients (18.3%; 95% CI 15.2 to 21.7%) with A(H1N1)pdm09 virus infection reported that their symptoms had not resolved by day 14; the percentage who died, were hospitalized, or continued to report symptoms at day 14 was 22.2% (95% CI: 18.9 to 25.8%).

In FLU 003, hospitalized patients with A(H1N1)pdm09 virus infection were enrolled at 56 sites in 16 countries; sites in 15 of these countries also enrolled patients in FLU 002 (enrollment by country is given in Acknowledgments). Asian sites enrolled 7.1% of patients; 10.5% of patients were from Australia; 70.4% from Europe; 2.0% from South America; and 10.0% from the United States. Fifty-five percent were enrolled between October 2009 and September 2010 (Table 3). Three hundred and seven (78.3%) of the 392 A(H1N1)pdm09 patients were enrolled from a general hospital ward and 85 (21.7%) were enrolled from an ICU. The median age of hospitalized patients with A(H1N1)pdm09 was 48 years; those enrolled in the first calendar year of enrollment had a median age that was seven years younger (44 versus 51 years; p = 0.001 for difference) than in subsequent years. This age difference was evident both for patients enrolled from the general ward and from the ICU. Fifty-one percent of patients were female; 11% were Asian, 4% were black, and 85% were white/other; the median BMI was 26 kg/m²; 5.3% had a BMI of ≥40 kg/m²; 30% reported smoking; and 25% of the women aged ≤45 years were pregnant. Fifty-three patients (13.5%) had HIV infection or other immune dysfunction; 14 of the 53 patients had HIV infection. Median time from the onset of symptoms to enrollment was five days for patients enrolled in the general ward and 10 days for patients enrolled from an ICU. Eighteen patients (4.7%) developed ILI symptoms after being hospitalized for some other condition; the median (IQR) time between admission and ILI symptom onset was 8 days (IQR: 5–18). Excluding the patients who likely acquired A(H1N1)pdm09 virus infection in the hospital, the median time from admission to enrollment was two days for patients enrolled from a general ward and 5 days for patients enrolled while in an ICU.

As would be expected, by most measures of disease severity assessed (medical history, complications defining eligibility, and other complications) patients enrolled in the ICU had more severe illness than those enrolled from the general ward. Exceptions were a history of asthma/chronic obstructive pulmonary disease (COPD), cardiovascular disease (CVD), liver or renal disease, and exacerbations of other co-morbidities which were more common among patients enrolled from a general ward than those enrolled from an ICU.

Two hundred and fifty-eight patients (65.8%) reported taking antivirals for influenza in the 14 days prior to enrollment; 256 were taking oseltamivir and 5 were taking zanamivir (3 following a course of oseltamivir). For patients taking an antiviral before enrollment, 46.6% reported starting antiviral treatment within 3 days of the onset of ILI symptoms; the median time between symptom onset and starting antiviral treatment was four days (IQR: 2–7).

Disease progression status was known at day 60 for 370 (94.4%) patients enrolled in FLU 003 (Figure 2). During the 60-day follow-up period, 80 (21.6%; 95% CI: 17.5 to 26.2%) patients developed disease progression; for those enrolled in the general ward and ICU, 37 (12.8%; 95% CI: 9.2 to 17.2%) and 43 (53.1%; 95% CI: 41.7 to 64.3%) patients experienced disease progression, respectively (Table 4).

Thirty-two patients (8.7%; 95% CI: 6.1 to 12.1%) died during the 60-day follow-up period. Twenty seven of these 32 patients died before discharge from the hospital at which they were enrolled. Figure 3 shows Kaplan-Meier plots for all-cause mortality for those enrolled in the general ward and the ICU. Cumulative mortality at 14, 28 and 60 days for those enrolled from a general ward were 2.3, 2.7, and 3.7%; for those enrolled from an ICU, these percentages were 9.4, 19.2, and 25.6%, respectively (95% CIs are given in the legend of Figure 3).

The number of days hospitalized since the time of enrollment, taking into account re-admissions (49 patients had at least one re-admission), was 5 days (IQR 2–12); for general ward patients the median number was 4 days (IQR 1–8) and for those enrolled from the ICU the median number was 15 days (IQR 8–32). For the estimation of these medians, deaths were assigned a worst-case time of 60 days.

At 28 days of follow-up among 289 surviving patients who had been discharged and attended the follow-up visit, 25.3% (95% CI: 20.3 to 30.7%) indicated that influenza symptoms had not resolved; 38.5% (95% CI; 30.3 to 46.7%) of patients had not resumed normal activities. At 60 days of follow-up among 292 surviving patients who had been discharged and attended the follow-up visit, 14.7% of patients (95% CI: 10.7 to 19.3%) indicated that symptoms had not resolved; 24.3% (95% CI: 17.4 to 32.2%) indicated that they had not resumed normal activities.

Table 5 summarizes the differences between FLU 002 and FLU 003 patients. In multivariable analyses, compared to outpatients, hospitalized patients were older, more likely to be female, have a history of asthma or COPD, and a history of CVD, liver or renal disease, and based on linear trend, have greater BMI and a longer duration of symptoms (p<0.05 for all). In addition, in the first year significantly fewer hospitalized patients were enrolled.

We also assessed whether pregnant women were more likely to be enrolled in FLU 003 than FLU 002. Among women aged ≤45 years, there were more pregnant women in FLU 003 than in FLU 002 (see Tables 1 and 2) (univariable OR = 16.0; 95% CI: 5.9 to 43.1). After covariate adjustment, this OR was 32.5 (95% CI: 8.9 to 118.6).

Figure 4 gives the frequency distribution of the number of days between the development of A(H1N1)pdm09-related symptoms and enrollment for patients in FLU 002 and FLU 003. This graphical depiction illustrates the longer period of time between symptom onset and enrollment for patients in FLU 003. Also, for those in FLU 003 for whom central laboratory RT-PCR results were negative, but with positive results for A(H1N1)pdm09 virus infection by a local laboratory, this time was even longer than for those with centrally confirmed A(H1N1)pdm09 virus infection in FLU 003(median time between illness onset and enrollment for these patients was 10 days; IQR: 6–15). Overall, there was a median of 2 (IQR: 1–4) days between local and central swab collection (see Appendix S2).

The prevalence of other co-pathogens was compared for a subsample of respiratory specimens for 235 patients in FLU 002 and 98 patients in FLU 003 (bottom of Table 5). With the exception of S. aureus, which was more common in FLU 002 than FLU 003 in univariable analysis but not in multivariable analyses, the prevalence of potential co-pathogens in the upper respiratory tract did not differ significantly between patients in the two studies.

Table 6 summarizes the association of baseline characteristics with disease progression in pooled analyses of data for FLU 002 and FLU 003 patients. The same baseline characteristics considered in the cross-sectional comparisons in Table 5 are summarized. In the unadjusted analysis, in addition to enrollment in the ICU, older age (median 48 years vs. 35 years), longer duration of symptoms (≥6 days versus <4 days), diabetes, history of CVD, renal or liver disease, and immunosuppression were significantly associated with disease progression. In multivariable analysis, enrollment in the ICU (OR 12.1, 95% CI 5.6 to 26.4; p<0.001), age (OR = 1.22 per 10 years older, 95% CI: 1.02 to 1.45; p = 0.03), duration of symptoms (≥6 days versus <4 days, OR 2.66, 95% CI 1.36 to 5.20; p = 0.004), and immunosuppression (OR 2.20, 95% CI 1.17 to 4.13; p = 0.01) were associated with disease progression.

An analysis was performed for female patients aged ≤45 years with A(H1N1)pdm09 virus infection to investigate whether pregnancy was associated with an increased risk of disease progression. For this cohort of 336 women, among whom 29 developed disease progression, the unadjusted OR for disease progression associated with pregnancy was 4.09 (95% CI: 1.57 to 10.6; p = 0.004). With covariate adjustment, this OR was reduced and no longer significantly greater than one (OR = 1.61, 95% CI:0.42 to 6.19).

Separate analyses were carried out for patients in each study (data not shown). With few exceptions, the multivariable analyses for each study were consistent with the pooled results. In both studies, there was an increased risk of progression associated with symptoms for 6 or more versus <4 days (ORs 2.54 and 2.85 for FLU 002 and FLU 003) and immunosuppression (ORs 4.04 and 1.99). Older age was not associated with progression in FLU 002 (OR = 0.95; p = 0.80) and was associated with an increased risk of progression in FLU 003 (OR = 1.27; p = 0.02); however, the difference in the ORs was not significant (p = 0.76). Asthma or COPD was associated with a non-significant increased risk of progression in outpatients (OR = 2.22; p = 0.21) and a significant reduced risk of progression in hospitalized patients (OR = 0.35; p = 0.01) (p = 0.005 for difference in ORs). Among women aged ≤45 years, pregnancy was associated with an increased risk of progression in FLU 002 (OR = 30.1; p = 0.015) and was not associated with disease progression in FLU 003 (OR = 0.88; p = 0.89) (p = 0.07 for difference in ORs). In outpatients, there was an increased risk of progression for those enrolled during the first year (OR = 12.3; p = 0.02); this was not evident for inpatients (OR = 0.83; p = 0.57) (p = 0.06 for differences in ORs). The associations of other baseline factors considered with disease progression did not differ for FLU 002 and FLU 003 patients.

We also examined predictors of mortality during the 60-day follow-up in patients enrolled in FLU 003 (Table 7). In univariable analyses in addition to enrollment in the ICU, Asian race, duration of symptoms ≥6 days, and a history of diabetes were associated with an increased risk of death. In multivariable analyses, Asian race (p = 0.01) and duration of symptoms (p = 0.03) remained significant predictors. There was also evidence of a higher risk of death for those with immunosuppression (p = 0.03) and for those enrolled in the initial calendar period of enrollment (p = 0.01).