Laboratory-confirmed deaths caused by influenza A (H1N1)pdm09 in the Santa Catarina State, Brazil, 2009-2014.

INTRODUCTION
Little information is available on the evolution of influenza A (H1N1)pdm09-related mortality in Brazil.


METHODS
During the period 2009-2014, official epidemiological surveillance and mortality data were used to ascertain influenza-related deaths in the Santa Catarina State in Brazil.


RESULTS
The highest overall mortality rate (2.3 per 100,000) was observed in 2009. Pregnant women constituted the highest risk group (mortality rate 18.0 per 100,000). Following vaccination of high-risk groups, the rate reduced sharply but peaked again in 2012. It subsequently decreased, albeit more gradually.


CONCLUSIONS
Influenza A (H1N1)pdm09 mortality demonstrated a cyclic pattern, with two peaks followed by a gradual decrease.

This paper aimed to provide the fi rst account of the evolution of infl uenza A (H1N1)pdm09 mortality using RT-PCR confi rmed cases between 2009 and 2014 in the State of Santa Catarinathe state that registered the highest infl uenza -related mortality rate in Brazil.
The main data source for this analysis were the reports published by the Ministry of Health (9) (10) (11) (12) (13) (14) and by the official epidemiological surveillance system [Sistema Nacional de Agravos de Notifi cação (SINAN)] (15) . A (H1N1) pdm09 identifi cation was performed by RT-PCR in the Adolfo Lutz Institute laboratory (Rio de Janeiro, Brazil). The analysis focused on mortality as the main outcome measure since death is the most severe component of disease burden, and is also least likely to be missing from registers. Death caused by A (H1N1)pdm09 was identifi ed by SINAN. In addition, for all suspected infl uenza cases reported to SINAN with no update on disease evolution, a death certifi cate search was conducted. From the death certifi cate search, any death where A (H1N1) pdm09 was mentioned, whether listed as the main cause or not, was included in order not to miss deaths resulting from complications of infl uenza. Annual mortality rates (MR) caused by A (H1N1)pdm09 were calculated by calendar year, region of residence, age group (<5, 5-19, 20-39, 40-59, and ≥60 years), and for pregnant women. The Brazilian Institute for Geography and Statistics census data and population projections for the aforementioned age groups were used as MR denominators. The pregnant population was derived from the Ministry of Health's database [acronym Sistema de Informações sobre Nascidos Vivos (SINASC)]. Influenza A (H1N1)pdm09 vaccination coverage and percentage of PCR-investigated cases by calendar year were also analyzed. Poisson distribution was used to estimate the 95% confi dence intervals (CI) for the MR.   In the fi rst year of the infl uenza A (H1N1)pdm09 pandemic (2009), the percentage of PCR-investigated cases notifi ed to the Santa Catarina State Epidemiological Surveillance reached almost 90%, ranging between 83% in the Southern region and 95% in the Northeastern region. The following year, the percentage remained high (94-100%) in all except the Mountain region. In 2011, the lowest percentage was observed in the Highland region (69%) while other regions varied from 87-97%. In 2012, when a second peak of infl uenza A (H1N1)pdm09 occurred, the percentage of cases notifi ed ranged between 83% in the Far West to 99% in the Greater Florianopolis area, and remained at similar levels during next two years. In 2009, the highest MR per 100,000 was observed among pregnant women (18.0), followed by 20-59-year-olds (3.2), and those aged ≥60 years (2.0) ( Table 1). Although they initially had the highest MR, from 2010-2014 no infl uenza-related deaths were reported in pregnant women, not even during the second mortality peak in 2012. The age groups 40-59 and ≥60 years demonstrated a slower decline in MR after the second peak compared to other groups. In addition, no A (H1N1)pdm09 related deaths were registered among children aged <2 years during the period analyzed.  related deaths occurred in August. July and September added 16.7% and 21.5%, respectively. In 2012, June was the month with the highest death toll (63.5%), followed by May (21.6%) and June (13.5%).
Infl uenza vaccination coverage of the target population in Santa Catarina was just more than 50% in 2010 but climbed to 85.6% in 2011 and remained at 95-98% over the following three years. Among pregnant women, the coverage was 90.3% in 2010, dropped to 64.1% in 2011, and increased to 88.7% and 85.9% in 2012 and 2013, respectively. The coverage of children <2 years of age was reported as 100% for all years analyzed except for 2011, when it was slightly lower (91%). In 2010, only older persons with chronic diseases such as diabetes; hypertension; or serious heart, lung, or kidney diseases, were targeted for vaccination, so the overall coverage for this group was low (42.5%). However, universal vaccination against infl uenza A (H1N1)pdm09 was offered to this group after 2010, which increased coverage to 84.7%, 89.8%, and 93.2% in the subsequent years.
Infl uenza A (H1N1)pdm09 in Santa Catarina showed two peaks: the fi rst one in 2009, when the pandemic fi rst reached Brazil, and the second in 2012. The second mortality peak (MR 1.2 per 100,000) was about half the fi rst, probably due to vaccination of high-risk groups (pregnant women, people with severe chronic diseases, children <2 years, and those aged ≥60 years). Further, after 2009, no subsequent infl uenza-related deaths were registered among pregnant women. A recent report also showed high vulnerability of pregnant women in the beginning of the pandemic in Brazil, when the case-fatality reached almost 7% (3) . Contrary to the widespread reports on low A (H1N1)pdm09 mortality among older-compared with younger-age groups in Brazil (1) (2) (4) (5) (6) (7) , the present study found the opposite result in Santa Catarina, although the difference was not statistically signifi cant. High mortality rates among older people were reported in Hong Kong and Japan, possibly due to influenza A(H3N2) which also circulated in those countries and which is known to cause signifi cant mortality (14) . In line with these results, the present study found 18 PCR-confi rmed deaths from A (H3N2) in 2009 and almost seven deaths per year in the period 2011-2014 in the general population.
In 2009, the A (H1N1)pdm09 mortality rate per 100,000 population based on laboratory-confirmed cases in Santa Catarina was twice the Brazilian average of 1.1 (2) and was even greater compared with mortality rates in the Ceará State in the Northeast (0.9) and in the São Paulo State in the Southeast of Brazil (0.8) during winter (June-September) (6) . This result is in line with a comprehensive analysis of the A (H1N1)pdm09 spatial distribution in Brazil that found strong evidence for progressively milder disease at lower latitudes (2) , likely due to warmer winters. In 2012, the second highest infl uenza mortality year, winter began earlier and was much colder than usual. This may explain why in that year the highest proportion of A (H1N1) pdm09 deaths occurred in June, two months earlier than in 2009 when the month of August accrued the most deaths. Regional variations of A (H1N1)pdm09 mortality rate in Santa Catarina were more pronounced in 2009 compared to 2012. All regions except the Northeastern region registered a decrease in infl uenza mortality, although no signifi cant differences in A (H1N1)pdm09 vaccination coverage were observed across the regions during the period analyzed. Some regions displayed a faster decline of the MR after the second peak whereas the others presented a more gradual decline (Figure 2). Infl uenza mitigation efforts should be analyzed and improved when necessary in order to further reduce the related mortality.
The present study showed that the vital statistics data based on death certifi cates underestimated infl uenza mortality in SC by >70% during the peak years of the A (H1N1)pdm09 epidemic, thereby missing a signifi cant increase in excess mortality caused by infl uenza, independent of pneumonia-related deaths. Vital statistics in Brazil are based on the underlying cause of death which gives more weight to chronic diseases as opposed to secondary infections such as infl uenza, thus underestimating the latter as a cause of death. Statistical modeling of the worldwide A (H1N1)pdm09 data indicated up to a 20-fold higher mortality than offi cially reported in 2009. In Brazil, this study estimated a likely MR of about 3.5 (the ratio 3.5:1 means that 3.5/4.5 deaths were missed, i.e. 78%), consistent with the >70% underreporting found in the present study. Better communication between epidemiological surveillance and vital statistics would greatly improve the accuracy of data on causes of death.
Limitations of the present study include possible correlation between case severity and probability of being registered by epidemiological surveillance, e.g. patients requiring hospital care. Although the choice of mortality as the main outcome measure and stratifi cation by region of residence partially accounted for these factors, their residual infl uence remains unknown. Further, the high percentage of laboratory-confi rmed cases does not rule out selection bias as areas with lower sensitivity to detect infl uenza coincide with areas in which the capacity of the health care system to diagnose and treat infl uenza A (H1N1)pdm09 is low, therefore introducing bias in the MR estimates. Also, the high sensitivity of RT-PCR has likely been achieved at the expense of underestimation of the real MR in the population, and should be treated as its lower bound. Strengths of the study include high coverage of laboratory-based outcome confi rmation, with excellent sensitivity and specifi city of RT-PCR in diagnosing infl uenza A (H1N1)pdm09; active search for additional outcomes in the death certifi cate database; and comprehensive geographical coverage of the Santa Catarina State.
The A (H1N1)pdm09 pandemic interrupted more than a decade-long downward trend in respiratory disease mortality in Latin America (15) . The mitigation of the pandemic in the Southern Brazilian states with the largest disease burden in the country was largely successful but regionally heterogeneous, leaving some areas with great susceptibility to disease recrudescence. Continuous improvements in epidemiological surveillance, vaccination coverage, and treatment must be made to further reduce the disease burden in Brazil.