The impact of the 2009 influenza pandemic on the seasonality of human respiratory syncytial virus: A systematic analysis

Background Several local studies showed that the 2009 influenza pandemic delayed the RSV season. However, no global‐level analyses are available on the possible impact of the 2009 influenza pandemic on the RSV season. Objectives We aim to understand the impact of the 2009 influenza pandemic on the RSV season. Methods We compiled data from published literature (through a systematic review), online reports/datasets and previously published data on global RSV seasonality and conducted a global‐level systematic analysis on the impact of the 2009 influenza pandemic on RSV seasonality. Results We included 354 seasons of 45 unique sites, from 26 countries. Globally, the influenza pandemic delayed the onset of the first RSV season by 0.58 months on average (95% CI: 0.42, 0.73; maximum delay: 2.5 months) and the onset of the second RSV season by a lesser extent (0.25 months; 95% CI: 0.12, 0.39; maximum delay: 3.4 months); no delayed onset was observed for the third RSV season. The delayed onset was most pronounced in the northern temperate, followed by the southern temperate, and was least pronounced in the tropics. Conclusions The 2009 influenza pandemic delayed the RSV onset on average by 0.58 months and up to 2.5 months. This suggests evidence of viral interference as well as the impact of public health measures and has important implications for preparedness for RSV season during the ongoing COVID‐19 pandemic and future pandemics.


| INTRODUCTION
Respiratory syncytial virus (RSV) is the most common pathogen identified in young children with acute lower respiratory infections 1,2 and poses a major burden on hospital beds during the peak of RSV transmission. RSV activity has clear seasonality in most parts of the world; RSV season, usually defined as a particular period of time with high RSV activity above a certain threshold, typically starts in late autumn and early winter in temperate regions and lasts for about 5 months. 3 4 In August 2010, WHO declared the end of the influenza pandemic. 5 Several local reports from China, [6][7][8] France, 9 Germany 10 and Israel 11,12 showed that the 2009 influenza pandemic delayed the RSV season by several weeks, whereas a study from Spain did not observe any differences in RSV season between the pandemic and the prepandemic period (a summary of these reports is available in Table S1). However, no global-level analyses are available. In the present study, we complied data from published literatures, online reports/datasets and previously published data on global RSV seasonality 3

| Data source
We collected RSV seasonality results (e.g. onset, offset, peak and duration of RSV season) and RSV seasonality data (e.g. weekly or monthly counts of laboratory-confirmed RSV infection) from the literature via a systematic literature review, online datasets/reports and previously published data on global RSV seasonality. 3 The following eligibility criteria were applied for the selection process.

| Inclusion criteria
• Studies reporting laboratory-confirmed incidence data of human infection of RSV.
• RSV seasonality results or RSV seasonality data should be extractable for the pandemic (1st RSV season) period as defined above, plus at least 1 year in pre-or post-pandemic period.
• Studies should be able to test RSV year-round (e.g. not just during influenza seasons) and should report at least 25 positive RSV cases per year. 3 • For studies that reported RSV seasonality data, the data should be made available at least on a monthly basis.

| Exclusion criteria
• Studies reporting respiratory infections only among those with special medical conditions (e.g. patients with chronic obstructive pulmonary disease or patients infected with human immunodeficiency virus).
• Studies only reporting nosocomial infections.

| Systematic literature review
A systematic literature review (PROSPERO registry number: CRD42021239011) was conducted. We searched three databases, Medline (Ovid), Embase (Ovid) and Global Health (Ovid) for any publications between 2009 and 2020 that potentially fulfilled the selection criteria above. The detailed search strategy can be found in Text S1.
Publications in any languages were considered for eligibility. The literature search and screening (including title and abstract screening and full-text screening) were conducted by two reviewers Y. L. and T. M., independently, with inconsistencies resolved through discussion among the review team. For data extraction, we used a standard data extraction form, modified based on our previous global seasonality of respiratory viruses study. 3 The data extraction form collected information on study sites, period, subjects, case definition, clinical specimens, RSV testing method, RSV seasonality results (including onset, offset, peak and duration as per reported in the literature) and RSV seasonality data (e.g. weekly or monthly counts of RSV positives). The data extraction was conducted independently by T. M. and jointly by X. W., F. d. W. and J. M. Where any inconsistencies occurred, a final decision was made by Y. L.

| Additional data
We extracted RSV activity data from three online datasets/reports from the FluWatch programme in Canada, 13  We also included RSV activity data from our previously published review on global RSV seasonality. 3

| Quality assessment
For each included record, two reviewers (X. W. and T. M.) conducted quality assessment independently using a modified questionnaire based on our previous study. 3 Briefly, the questionnaire comprised three brief questions regarding data representativeness, diagnostic practices and timely reporting; for each question, each study was rated from A (very good) to D (bad). Studies with any 'D' ratings were excluded from the analysis and studies with any 'C' ratings were excluded from the sensitivity analysis that is described in the next section. Details of the questionnaire are available in Text S2.

| Data analysis
For those studies/reports/datasets that had RSV seasonality data (e.g. weekly/monthly counts of RSV positives), we determined the RSV seasonality results using the following approach: we first divided the timeline into 12-month intervals so that each interval had a complete RSV season; for each of the weeks/months per interval, we then calculated the annual cumulative proportion (ACP), which ranged 0-1.
The ACP of the last week/month of the interval should be 1. Based on ACP, RSV onset was defined as the week/month with ACP being 0.1, and RSV offset was defined as the week/month with ACP being 0.9. Linear interpolation was applied to allow for non-integer results for RSV onset/offset (e.g. RSV onset could be month 1.2 or week 5.5 rather than month 1 or week 6). RSV duration was defined as the difference between RSV onset and offset. RSV peak was defined as the week/month with the highest RSV counts in each 12-month interval.
For those studies that reported RSV seasonality results (e.g. onset, offset, peak and duration of RSV season), we used the extracted results for our data analysis. If studies had both RSV seasonality results and RSV seasonality data, we prioritised the inclusion of RSV seasonality data in our main analysis and prioritised the inclusion of RSV seasonality results in our sensitivity analysis. Where available, we calculated the time interval between onset and peak as an additional measure of interest.
Our primary outcome of interest, determined a priori, was the difference in RSV onset between the first RSV season in the pandemic and RSV seasons during the inter-pandemic period (i.e. pre-pandemic and post-pandemic). Secondary outcomes of interest included the difference in RSV offset, peak, duration and onsetto-peak interval between the first RSV season in the pandemic and the inter-pandemic period. The same comparisons as described above were repeated between the second RSV season since the pandemic and the inter-pandemic period. An ad hoc analysis was also conducted to compare the difference in RSV onset between the third RSV season since the pandemic and the inter-pandemic period.
Subgroup analysis that separated pre-and post-pandemic periods was also conducted. Based on the quality assessment results, we conducted a sensitivity analysis that excluded studies with 'C' ratings in any of the questions. We also conducted a sensitivity analysis that excluded studies reporting less than five RSV seasons. As an exploratory analysis, we compared the time length required to reach different levels of ACP between the pandemic and inter-pandemic periods, by using seasonality data. This would help examine the impact of the influenza pandemic on RSV activity over the complete course of one RSV season.
All data analyses and visualisations were conducted using the R software (version 3.6.2).

| RESULTS
As shown in Figure 1, we screened 1792 records by title and abstract and 259 records by full-text, which led to the inclusion of 32 studies from the literature review. In addition, we included three more records from online datasets/reports and eight more records from previously published data on global RSV seasonality, 3 Table S2.
Overall, we found that the influenza pandemic delayed the onset of the first RSV season by 0.58 months (95% CI: 0.42, 0.73) on average, with a maximum delay of 2.5 months. By comparison, the influenza pandemic delayed the onset of the second RSV season by a lesser extent, which was 0.25 months (95% CI: 0.12, 0.39), with a maximum delay of 3.4 months. RSV seasons during the pandemic were found to be shorter, and the interval between onset and peak was also shorter ( Table 1). Similar findings were observed in the sensitivity analyses that prioritised the inclusion of seasonality results (Table S3); excluded studies with any 'C' ratings (Table S4); and excluded studies with less than five RSV seasons (Table S5). The above findings did not change substantively when only including the pre-pandemic period or when only including the post-pandemic period. In the ad hoc analysis that assessed the third RSV season (since the influenza pandemic), we did not observe a statistically significant delay in the RSV season onset (95% CI: À0.33, 0.04).
Some regional variations in the effects of the influenza pandemic on RSV seasonality were noted, as shown in Figure 2 and Figures S1-S4. For the first RSV season of the pandemic, the delay in RSV onset and peak was more pronounced in the northern temperate, whereas the delay in RSV offset was more pronounced in the southern temperate; no statistically significant findings were observed in the tropics where RSV activity was seasonal ( Table 2). For the second RSV season of the pandemic, interestingly, we found that the RSV season ended earlier in the southern temperate, opposite from what was observed in the northern temperate; no statistically significant findings were observed in the tropics ( Table 2).
The results of our exploratory analysis using only the RSV seasonality data confirmed the regional variations observed above but provided more details (Figure 3): in the northern temperate, the delay in observing the same level of cumulative RSV activity was most pronounced at the beginning of both the first and second RSV seasons since the pandemic, but the delay became less pronounced over the course of the season, and there was almost no delay at the season offset; in the southern temperate, a pronounced delay (0.38 months, 95% CI: 0.14, 0.61) was observed around the offset for the first RSV season and an advanced RSV epidemic (À0.29 months, 95% CI: À0.63, 0.04) was observed around the offset for the second RSV season, though not being statistically significant.

| DISCUSSION
In this study, we highlighted a globally averaged delay of 0.58 months and a maximum delay of 2.5 months in the onset of the first RSV season in the influenza pandemic compared with the inter-pandemic period. The delayed onset was most pronounced in the northern temperate, followed by the southern temperate, and was least pronounced in the tropics. The second RSV season was impacted to a lesser extent, and the third RSV season was not impacted.   tropics, however, we did not observe any statistically significant differences between RSV seasons in the pandemic and in the interpandemic period, although the point estimates indicated that the RSV season might be delayed. This could be due to the lack of statistical power because the majority of the data were from temperate regions.
This could also be due to the fact that the timing of RSV season was more varied in the tropics than the temperate regions. 3 Our study has several strengths. First, we went beyond published literatures that reported the impact of the influenza pandemic on RSV seasonality (as summarised in Table S1), by compiling both RSV seasonality results and RSV seasonality data from various sources.
This allowed us to analyse the best available data and helps reduce publication bias that tended to favour statistically significant results.