Study Population

This analysis is based on data from ILI-002, a hospital-based prospective observational cohort study of ILI. The present analysis includes all participants enrolled in the ILI-002 study in whom 1 or 2 viruses were detected.

The ILI-002 study was carried out at 6 public hospitals, 5 of them located in Mexico City and 1 in San Luis Potosí. Participating hospitals included 2 general hospitals (1 located in Mexico City and 1 in San Luis Potosí), 2 tertiary care pediatric hospitals, and 2 tertiary care hospitals (1 of them dedicated to the treatment of respiratory disorders, whereas the other provides medical care in a wide range of medical specialties). Adults and children seeking medical attention with ILI, defined as a respiratory symptom (eg, cough, dyspnea) plus a systemic symptom (eg, fever, malaise), were invited to participate in the ILI-002 protocol (ClinicalTrials.gov identifier: NCT01418287). For those enrolled, a follow-up telephone or face-to-face interview was performed at 14+/- 3 days, and a visit happened 28+/-5 days after inclusion. The study protocol was approved by the ethics committee at all participating institutions, and all participants or guardians signed an informed consent or an assent form when pertinent.

Pathogen Detection

A nasopharyngeal swab for multiple PCR pathogen detection was obtained at enrollment. Samples were stored in transport media at 4°C at each site (for sites located in Mexico City) and sent daily to a central facility (Molecular Biology Laboratory, Infectious Diseases Department, Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, Mexico City) for testing. Samples obtained in San Luis Potosí (a site outside of Mexico City) were stored at –70°C and sent weekly to the central facility for pathogen detection testing.

Samples were tested with either the RespiFinder19 kit or the RespiFinder22 kit (PathoFinder BV, Maastricht, the Netherlands). The 19-pathogen PCR test can detect and differentiate 15 viruses (coronavirus [CoV] NL63, OC43, and 229E; HMPV, influenza A, influenza A H5N1, influenza B, parainfluenza virus [HPIV] types 1–4; RSV types A and B; rhinovirus [RV]/enterovirus; and adenovirus [HAdV]) and 4 bacteria (Bordetella pertussis, Chlamydophila pneumoniae, Legionella pneumophila, and Mycoplasma pneumoniae). The 22-pathogen assay added CoV HKU1, HBoV, and influenza A (H1N1)pdm09 while removing influenza A H5N1. As reported by the manufacturer, the analytical limit of detection of the assay varies between 5 and 50 copies per reaction for most targets. Samples that were tested originally with the RespiFinder19 kit were subsequently tested for HBoV detection with the use of virus-specific primers. In addition, all samples were tested by real-time RT-PCR for influenza A following the Centers for Disease Control and Prevention (CDC) protocol.

Study Variables

Participants hospitalized during the 28 days of the study follow-up were considered to have severe disease. Hospitalization was defined as participants who were admitted to the hospital or remained in the emergency departments for at least 24 hours.

Participants with a detected bacterial pathogen (Bordetella pertussis, Chlamydophila pneumoniae, Legionella pneumophila, and Mycoplasma pneumoniae), no virus, or >2 viruses were excluded. Comorbidities were defined as 1 of the following: chronic obstructive pulmonary disease, cardiovascular disease, diabetes mellitus, previous use of systemic steroids, obesity, overweight, and underweight.

Statistical Analysis

For this analysis, we grouped similar genera of viruses and examined the 8 most frequent groups of viruses isolated: influenza (A, A (H1N1)pdm09, and B grouped as influenza), HMPV, HPIV, RSV, RV, HAdV, CoV, and HBoV. Comparisons of baseline factors were made between hospitalization and nonhospitalization groups. All potential risk factors that were not categorical were grouped into categories. Chi-square statistics were used to make the univariate comparisons of the risk factors.

Logistic regression models were used to compare hospitalization between all pairs of viruses and the combinations of the 2 viruses. Combinations with <10 participants were not analyzed. Each logistic regression model included sex, age (grouped into 3 categories), days since symptom onset (grouped into 3 categories), and comorbidity (yes/no). Odds ratios and 95% confidence intervals were calculated.

Hospitalization of Single Viral Diagnosis

The adjusted odds ratios (ORs) for hospitalization rates between each of the viruses included in the study are shown in Figure 1. Participants infected with HBoV were more likely to be hospitalized than those infected with influenza, CoV, HPIV, and RV (Figure 1A). Those with HAdV, HMPV, and RSV had similar but not statistically significant results. Participants infected with RSV were more likely to be hospitalized compared with cases of influenza, CoV, RV, HPIV, HAdV, and, to a lesser extent (not statistically significant), HMPV (Figure 1B). Participants with HMPV were more likely to be hospitalized compared with cases of influenza, CoV, and RV, but were not statistically significantly different compared with HPIV and HAdV (Figure 1C). HPIV cases were more likely to be hospitalized compared with cases of influenza, CoV, and, to lesser extent, HAdV (as assessed by point estimates) (Figure 1D), although none of these comparisons were statistically significant. Participants with HAdV were more likely to be hospitalized compared with cases of influenza and CoV (as assessed by point estimates), although neither of these comparisons reached statistical significance (Figure 1E). Participants with RV were more likely to be hospitalized as compared with CoV cases (Figure 1F). Although not statistically significant, CoV cases were less likely to be hospitalized as compared with those with influenza (Figure 1G). Figure 1H shows comparisons with influenza cases (already described).

Hospitalization of Dual Viral Diagnosis

When 2 viruses were isolated, those with combinations of RSV+HPIV, CoV+HMPV, and CoV+RSV were less likely to be hospitalized than participants infected with individual viruses (Figure 2A–C). The point estimates for individual viruses in HMPV and RV demonstrated a higher likelihood of these patients being hospitalized than those with combinations, but all confidence intervals included 1 (Figure 2D). Participants with HBoV+RV were more likely to be hospitalized than those with RV, but were hospitalized as frequently as those with HBoV alone (Figure 2J). The point estimate for severity in CoV+RV was greater than that for individuals for CoV or RV alone (Figure 2K), but the confidence interval included 1. The confidence intervals for all other combinations included 1, but the point estimates indicated that some of the combinations could be more severe than 1 of the single agents (Figure 2L–R).

DISCUSSION

Reports of the impact of multiple viral infections have been more frequent with the availability of PCR assays that detect multiple pathogens [9–15]. Most reports analyze data by grouping all combinations and compare this group with different single viruses. This has resulted in variable interpretations. ILI-002 is a large study of those with ILI in which a multipathogen PCR assay was performed on samples from all participants. The size of this cohort allowed there to be a sufficient number of participants with various virus combinations to perform separate analyses for some virus combinations and examine whether virus combinations increase severity over individual viruses. Our results highlight the importance of carrying out these separate analyses.

Our data demonstrate that the severity of diseases was higher with specific viruses (eg, HBoV, RSV, and HMPV). However, in no combination was the dual infection significantly worse than in both of the individual viruses. Furthermore, as a class effect, it does not appear that infection with >1 virus increases severity of disease. Many of the confidence intervals for the combinations include no difference for each of the comparisons with the individual viruses. Based solely on point estimates, which may change with increasing numbers of participants, there appear to be some patterns that indicate a leading or governing effect of 1 virus in the combination. An example of this is the combination of CoV+RSV (Figure 2C, panel C); CoV cases and CoV+RSV cases were less severe than cases of only RSV. Moreover, the severity of CoV+RSV cases was no different than that of CoV cases. This could indicate that CoV is the leading/governing agent in the combination. Similarly, HBoV may be governing RV in the combination of these 2 viruses (panel J), RV governing CoV (panel K), influenza governing HBoV (panel N), and RV governing RSV (panel P). The confidence intervals around the combinations are often large and include 1, so these interpretations are not conclusive and could differ if larger numbers of participants were studied. However, the pattern observed in the point estimates is suggestive of the governing effect described above and warrants further study.

One explanation for the governing results is that virological testing was done only once in each participant, which does not distinguish sequential infections with 2 viruses from simultaneous infection with both viruses. It is possible that these results could reflect sequential infections rather than simultaneous infection and that symptoms (and hospitalization) could be the result of only 1 of the 2 viruses. The sample collection time with respect to the course of illness would then be an important factor. In a study carried out in participants with ILI, viral co-infections were detected more frequently in samples obtained during the first 2 days from symptom onset compared with those obtained after 3–7 days. Thus, it is plausible that detection of co-infections might be the result of prolonged shedding of 1 virus with a subsequent infection with the second virus. Also, it could reflect a reduced ability of a second virus to replicate due to an already initiated host response and the production of interferon as a result of an initial viral infection. Interference of 1 virus with another virus has been shown to occur in vitro, and epidemiological studies suggest that circulation of 1 virus might affect circulation of another virus in communities.

Overall, most previous studies have shown similar severity of single infections when compared with mixed infections. However, some studies have shown significant differences between specific combinations of viruses and single viruses. For example, in cases of co-infection with RSV+RV and RSV+HBoV, the illness appeared to be more severe than in cases of RSV or HBoV infection alone. Our results for RSV+RV follow the same pattern, but we did not have sufficient data to study the combination RSV+HBoV.

Our analysis showed that participants with detection of only influenza virus were less likely to require hospitalization than participants in whom other viruses were detected. This was observed despite the fact that the study included the 2013–2014 winter season, when a severe wave of influenza A(H1N1)pdm09 was registered in Mexico. This result could be derived from inclusion of all influenza subtypes in the analysis. The lower hospitalization rate in participants with influenza virus infection compared with those with other viruses may also be explained by influenza vaccination. Since 2009, influenza vaccination coverage in Mexico has been high, and influenza vaccination has been reported to reduce influenza hospitalizations. Unfortunately, it was not possible to obtain detailed data regarding influenza vaccination status for study participants to assess this.

We also found that for single virus comparisons, HBoV, RSV, and HMPV were associated with severe infections. The RSV and HMPV finding is consistent with other reports that show these viruses to be leading causes of LRTI in children and adults [28–30]. In contrast, the role of HBoV as a cause of severe infections is less well established. Because HBoV infection is very common and frequently found in asymptomatic participants, the pathogenic role of this virus has been questioned [31, 32]. Children with higher viral loads tend to have more severe infections, longer hospitalization duration [34, 35], and are found to have co-infection by other viruses [36, 37] less frequently than those with lower viral loads. However, in children in daycare, viral load had no apparent association with severity of illness, which also might reflect challenges in reproducibly measuring viral load in secretions. In all, these studies suggest that HBoV is frequently present in children as an asymptomatic or chronic infection with low viral loads, whereas some infections in which a high viral load is present may be associated with severe infections requiring hospitalization. Our study did not measure viral load, so it is unclear if our HBoV participants represented a sample of high–viral load participants, as we did not include asymptomatic participants with HBoV.

The strengths of this study include the large numbers of participants and the consistent baseline testing and follow-up. Limitations include the definition of severity, as hospitalization was a surrogate for actual patient health status. Patients may be hospitalized for causes other than their respiratory illness, such as worsening of comorbidities, observation in high-risk participants, or social reasons. Further studies should be done using direct measures of patient health status such as intensity and duration of participants’ symptoms or complications (eg, pneumonia). We have developed a symptom scale for influenza (FLU-PRO) as part of the study ILI-002 that could be used in future studies.

The presence of chronic underlying conditions is an important factor that is associated with risk of developing severe respiratory infections. In addition, it is possible that chronic conditions may increase the risk of acquiring infections by multiple pathogens. A recent study reported that patients with coinfection caused by 2 or 3 different influenza virus strains were more likely to have underlying cardiovascular disorders than those with single influenza infections; however, no differences were observed in the prevalence of other underlying conditions. Although some other studies have found a higher prevalence of chronic disorders in patients with multiple viral pathogens, these appear to be limited to specific conditions, and no differences have been observed for other disorders [15, 40]. In addition, many studies have not found an association between the presence of chronic disorders and detection of multiple viruses [9, 14, 21, 24]. The main objective of our study was to determine if codetection of 2 viruses was associated with worse outcome, and we did not analyze which factors may have led to acquisition of ≥2 viruses; nevertheless, our analysis included the presence of chronic conditions as a covariate, in order to account for potential confounding.

In a previous analysis of ILI-002 limited to children <5 years of age, the severity of single viruses was compared, and the results were similar to our findings for single viruses in the present study. The majority of studies have focused on children, and the effect of mixed viral infections in adults is less clear. One of the strengths of this analysis is that our study population included both pediatric and adult symptomatic participants. Although some comparisons resulted in estimates with wide confidence intervals, our interpretation based on point estimates revealed several patterns and could be hypothesis-generating for future studies.

CONCLUSIONS

Our results suggest that, in general, having >1 virus detected by PCR on a respiratory sample does not increase the severity of disease from ILI. To assess the differences in hospitalization rates of mixed respiratory infections, it is necessary to carry out analyses between specific combinations of viruses. When 2 viruses are detected, it appears that the clinical severity of a respiratory infection, as defined by hospitalization, may be associated with 1 of these agents. Future studies with a larger number of participants designed to distinguish between sequential and simultaneous infection and using direct measures of patient health status should be of help in defining the role of each virus during the evolution of an acute respiratory episode.