Veterinarians who agreed to participate in the study were asked to enroll growing-pig farms (i.e., farms that house pigs 3–30 weeks of age) in the midwestern United States that were representative of modern swine production systems and that were owned by producers interested in participating in the study. Producers were allowed to withdraw from the study at any time.

From June 2009 through December 2011, participating farms were visited every month for 12–24 consecutive months. At each visit, the investigator would meet with the farm manager/owner to decide which pigs were to be sampled. If pigs were all in 1 age group, samples were collected from that group; if pigs were in >1 age group, samples would be collected from the age group closest to 10 weeks, the group most likely to yield the most influenza A virus–positive pigs, per previous reports (20).

A total of 30 nasal swab samples were collected at each visit, enabling us to be 95% confident of detecting at least 1 positive sample when influenza prevalence was at least 10%. Clinically healthy pigs were restrained by a snare, and a nasal swab (Starswab II, Starplex Scientific Inc., Etobicoke, Ontario, Canada) was inserted 2–3 inches into the back of each nostril while being rotated. Nasal swabs were labeled with a specific code containing the farm identification number, month, 2-letter state abbreviation, and nasal swab sample number.

During the visit, the age of the pigs and respiratory clinical signs (absence or presence of sneezing, coughing, and nasal secretion) among the group members were recorded. Nasal swabs and submission sheets were placed into a Styrofoam container with ice packs and shipped overnight to the laboratory for testing.

All nasal swab samples were tested at the virology department laboratory of St. Jude Children’s Research Hospital (Memphis, TN, USA). Nasal swab samples were initially screened for influenza A virus by real-time reverse transcription (RRT-PCR) selective for the matrix gene. Samples that were positive by RRT-PCR underwent further diagnostics for determination of subtype, including the influenza A(H1N1)pdm09 virus and swine H1 and H3 viruses (21–23).

A farm was considered positive for a given month if any of the 30 individually collected swab samples tested positive. Farm-level data, such as farm size, were analyzed by repeated measures logistic regression, and differences between farms were accounted for by including farm as a random effect and allowing for an autoregressive effect by month within the farm. Model building was performed by first screening independent variables through univariate analysis. Variables with a p value <0.25 were retained for the multivariable model. All selected variables were forced in the model including 2-way interactions and were sequentially removed if p value was >0.05.

We built 2 models. The first model assessed the relationships between farm status (positive vs. negative) for influenza virus and age, year, clinical signs, and season. The second model assessed the relationship between respiratory clinical signs (presence vs. absence) and subtype and season. Season was included in the models by categorizing the 4 seasons as follows: winter (January–March), spring (April–June), summer (July–September), and fall (October–December). Statistical procedures were performed in SAS 9.2 (SAS Institute Inc., Cary, NC, USA).

A total of 16,170 nasal swab samples from 540 groups of growing pigs from the remaining 32 farms were collected. From the total number of samples collected, 746 (4.6%) were positive for influenza A virus by RRT-PCR, and 178 viruses were isolated from these samples. At least 1 positive sample was detected in 117 (21.7%) groups of pigs; thus, these groups were classified as positive. Of the 32 farms, 29 (90.6%) had at least 1 positive group throughout the study. Of the 117 groups with RRT-PCR–positive results for influenza A virus, complete or partial subtype details were obtained for 99 (84.6%) pig groups (Figure 1). Influenza A virus infection with just 1 subtype (H1N1, H1N2, H3N2, or H1N1pdm09) was detected in 21, 19, 9, and 17 groups, respectively. Dual infections were detected in 10 groups, of which 8 concurrently harbored influenza virus subtypes H1N2 and H1N1pdm09, 1 harbored subtype H1N1 and an H3N-untypeable virus, and the remaining group harbored subtype H1N1 and an H1N-untypeable virus. Partial subtyping information was obtained for 16 pig groups, from which 11, 4 and 1 were infected with an H1N-untypeable, H3N-untypeable, and an H1N-untypeable with pandemic matrix gene virus subtype, respectively. In 18 groups, a subtype could not be defined through either RRT-PCR or sequencing. At most farms in our study, groups of pigs were identified as having multiple and different influenza A viruses detected throughout the surveillance period (Figure 1). Viruses were isolated from 178 swab samples that originated from 62 pig groups.

Of the positive groups, the mean and median numbers of samples positive for influenza A virus by RRT-PCR were 6.4 and 2, respectively. The numbers of positive samples ranged from 1 to 30; most groups (n = 48) had 1 positive sample (Figure 2). There were 15, 10, 4, and 2 groups that had 2, 3, 4, and 5 positive samples, respectively. A total of 13 groups had >20 positive samples, 2 had 29 positive samples, and 1 had 30 positive samples. Although most samples collected from these 13 groups were positive for influenza A virus by RRT-PCR, pigs in only 7 of these groups exhibited clinical signs of influenza-like illness.

The number of positive groups per farm ranged from 1 to 18. On average, 31% of the groups tested by farm throughout the program were classified as positive (Figure 3). Farms with no influenza A virus–positive results were monitored for ≈1 year but lacked consistency in the testing frequency and time intervals between tests.

The average age of the pigs in the 540 groups was 13.7 ± 5.7 weeks. Influenza virus was detected in pigs as young as 4 weeks and as old as 30–32 weeks of age. Mean age in positive groups was 12.4 ± 5.2 weeks and in negative groups was 13.9 ± 5.8 SD (Figure 4). However, age was not a statistically significant predictor of influenza A virus test status.

Respiratory clinical signs were observed in pigs in 187 (34.6%) of 540 groups. From these 187 groups that reportedly exhibited clinical signs, 43 (22.9%) were positive for influenza A virus. From the 353 groups that exhibited no clinical signs, 74 (20.9%) were positive for influenza virus (Table 1). Even when within-group prevalence of influenza A virus was high, such as in the 13 groups in which >20 samples were positive for influenza A virus, clinical signs of respiratory disease were low. Indeed, clinical signs of influenza-like illness were noted in only 7 of those 13 groups.

Throughout the study, the numbers (proportions) of groups sampled in each season were similar. In winter, 124 (23%) were sampled; in spring, 137 (25%); in summer, 150 (28%); and in fall, 129 (24%) (Table 2).

In the first model, univariate analysis of all variables (age, season, and year) except clinical signs yielded p<0.25. The multivariable model retained 2 variables: season and age. Odds of positive results for influenza A virus were 2 (95% CI 1.1–3.8) times and 1.9 (95% CI 1.0–3.5) times higher for groups of pigs tested during the spring and summer, respectively, than for groups tested during the fall. There was no association between the winter season and influenza virus detection. Age was not significantly (p = 0.09) associated with influenza A virus group status; however, the variable was left in the model because of confounding (Table 3).

In the second model, complete subtype information was available for 81 monthly test results. Because neither subtype nor season was significantly associated with presence of clinical signs at the univariate level (Table 4), no attempts were made to build a multivariate model.