Our results revealed that: the younger people were amongst the most affected; that certain regions of the city bore a disproportionately high risk for a pandemic burden compared with other regions and that the occupation and working environment of individuals were important A(H1N1)pdm09 infection risk predictors for Djibouti. We therefore report for the first time, the epidemiological characteristics of pandemic influenza A(H1N1)pdm09 in this nation, which is located in the WHO Eastern Mediterranean Region (EMR). EMR is comprised of 22 countries, including Djibouti (see introduction) and has little available epidemiological information for A(H1N1)pdm09, particularly relating to morbidity (incidence), mortality (fatalities) and sero-prevalence (sero-surveillance). The EMR is in the northern hemisphere. According to recent studies, the global pandemic occurred in two major waves in both the southern and northern hemispheres
. In many parts of the world, the first wave occurred between May and November 2009, followed by a gap and overlap by an ascending second wave that ran from October 2009 to February 2010
[4, 26]. Once introduced, the A(H1N1)pdm09 virus spread rapidly across the country for 15 to 20 weeks, thereafter transmission was sustained at a much lower intensity
. This strain had the ability to predominate and replace other seasonal strains like the H3N2 and seasonal H1N1, with few exceptions.
As documented in temperate countries
, in EMR the two waves seem to have followed and peaked in cold times during winter, but their peak magnitude varied between countries. The first pandemic wave is presumed to have begun with the reports of the first cases in the region, which were in Kuwait in May 2009 and in the United Arab Emirates (UAE) in July 2009
[28, 29], before spreading to other member states. Once infection was introduced to these two nations it gradually progressed to allow community transmission, peaking in August 2009 for UAE, and in October 2009 for Kuwait
[28, 29]. The second wave was more pronounced and better documented than the first, probably due to more awareness of the infection. It begun in late October 2009, peaked in December 2009 and January 2010 and then declined from February to March 2010
, with a lag in Afghanistan, Iraq and Oman
[3, 26]. This corroborates the sentinel data in Egypt and Morocco
, which confirmed that the peak of A(H1N1)pdm09 incidence occurred between November 2009 and January 2010. Similarly, to Djibouti’s neighbour, Ethiopia, which reported its first two cases in June 2009 (first wave), followed by a lapse until early 2010 (second wave), when more cases were noticed
. Although this scenario was acceptable to the regional WHO EMR office, it should be considered carefully since countries with functional surveillance systems were likely to detect and report the pandemic circulation earlier than those without these surveillance systems.
The EMRs official statistics of 24th October 2009, indicated 17,150 incidences and 111 fatalities
; and later, as of 30th April 2010, reported 1,059 fatality cases
. Amongst the 1,059 fatalities reported, Egypt (277), Iran (147), Saudi Arabia (128), Syria (138) and Morocco (64) had the highest proportion in 20 of the 22 member states involved
. At the country level, limited Information was available from Djibouti, with only nine laboratory confirmed cases and zero deaths cited in both reports. Unpublished influenza and influenza like illness data suggest that Djibouti city might have experienced a lagged second wave with a steady increase from January to June and a sharp decrease in July (Dr. Ammar Abdo Ahmed, personal communication). This is, however, based on a quite limited number of cases (n = 278). We therefore observed that it was difficult to propose an estimate of the actual epidemic impact, because the different patterns reported from neighbouring countries (e.g., Ethiopia
) indicated that incident cases might have been underreported. We attempted to provide additional epidemiological information through this seroprevalence study of the Djiboutian population, which happened in the winter of 2010 (November 2010 to February 2011).
This study had several limitations. Firstly, late sampling did not enable distinction between individuals infected during the very first wave in summer 2009 and those infected during the second wave (autumn 2009 and during 2010). The general assumption was that the infection occurred between the winter of 2009 (November 2009 to February 2010) and that of 2010 (November 2010 to February 2011), with the exception of subjects over 65 years old who could have been exposed to pandemic antigenically related strains before
[5, 6]. As a result, the study could only provide a rough cumulative estimate of the influenza A(H1N1)pdm09 burden during years 2009–2010. Secondly, our study potentially underestimated the prevalence of pandemic influenza since the cohort data collection ended one month before the end of the winter season. However this underestimation was unlikely to have any significant effect on epidemiological characteristics such as age, spatial distribution and risk factors. Thirdly, in the absence of pre-pandemic samples, sero-prevalence data were used as a surrogate for seronconversion data to estimate the number of infected cases. Of note is that this approach has been validated to be a robust estimation of seroconversion rate as is evident from previous studies
[15, 16]. Fourthly, the opportunistic study of individuals recruited from the 2009 Hajj Pilgrim and community health (social) workers databases cannot claim to provide a representative picture of the Djiboutian population. Hypothetically, we might have compromised the participation of the middle class households since the Hajj Pilgrimage to Mecca is costly and only those amongst the high-income bracket can afford to do it, and, conversely, it is the majority of low-income persons who depends on the community health (social) workers service. These two groups are therefore likely to have different exposures to infection; and international travel and participation in the Hajj could have elevated the risk to pandemic influenza to those who were involved
. However, with regard to influenza epidemiology, such a bias was expected to have a limited impact on the estimate of the global number of cases in the general population
In our study, which happened soon after the second pandemic wave (November 2010- February 2011), we had a 29.1% overall seroprevalence and with most occurring amongst the young, 35.1% in 0–4 years old and 34.5% in 5–19 years old. These observations are consistent with the recent meta-analysis estimation of ten studies from nine countries (of Asia, Africa, Europe and North America)
, which indicate that the overall prevalence, in the same period, was 32% (95% CI 26-39%) and more occurrences were in 5–19 year olds at 47% and in 0–4 year olds at 36%, as was the case of Djibouti
. We also compared the prevalence (34.4%) amongst young Djiboutian <20 years with two other countries, Laos
 (20.8%) and Mali
 (19.5%) that applied the study protocol. A brief overview of the EMRs status, based on the first 500 reported cases, indicates more young people (29 years) were disproportionately affected in the first wave and accounting for up to 59.6% of all incidences
. However, precise comparison with findings from other regions is hampered by the limited information reported
, and variations in the occurrence of the first and second waves
[35, 36]. Our findings suggest that the country had a comparable higher disease burden than many of the other documented countries. This therefore usefully complements data from the Djibouti Ministry of health, the regional health agencies, the WHO and the CDC records.
In the risk assessment for A(H1N1)pdm09 infection, our results indicated that the seropositivity was significantly associated with individuals’ age (<25 years), occupation (student), place of residence (District 4) and the living environment (living near to a riverbank). As mentioned earlier, people under 25 years old were immunologically naïve and vulnerable to infection with the new virus
[5, 6]. Amongst the elderly, cross-reactive antibodies against the 2009 pandemic influenza virus are likely to have originated from infections caused by the 1918 Spanish influenza virus and antigenically related descendants as suggested by haemagglutinin sequence comparison
. Similarly, students may be at specific risk of infection because of their age class and as a result of their close social interactions within the school environment that potentiates rapid spread of ILI
. By residential location, we observed that District 4 had the highest risk for infection compared to those residing elsewhere. This region is geographically separated from the three others, districts 1, 2 and 3, by a seasonal river. This population is largely made up of refugees or illegal migrants fleeing the civil unrest in neighbouring countries, living in close proximity and have high material deprivation. Such characteristics in part contributed to their underrepresentation (and that of the district) in our cohort, due to the very restrictive household enrollment protocol applied. In spite of this, the overall picture of seropositivity distribution by age group and risk factors was unlikely to be biased. Because, the age group proportions of the district was not significantly different from those of the three other districts. Whilst low wealth index alone was not associated with seropositivity (see below), it is possible that specific characteristics of this vulnerable population may represent an increased risk of infection
. Besides District 4, individuals living close to the River Balbala valley banks (see Figure
4, comprised of some residents from District 2, District 3 and District 4), had a three times greater risk of infection than those living further away. We have no specific explanation for this observation.
Of note is the observation of significant protection against the A(H1N1)pdm09 infection by those people having a low wealth index, large household size, being working class and working in open air spaces. A possible explanation is that the Djiboutian people belonging to a low wealth index class could have had limited exposure to the new pandemic strain, hence the lower incidence
, a situation mostly associated with relative social isolation compared to other groups such as the working class who would benefit from better living standards, access to medical treatment and information. Similarly, people working in areas which are well ventilated would have limited chance of sustained air contamination from infectious aerosols. According to Kieffer et al., the lower risk amongst households with more members was attributed to the overcrowding, which results in reduced times that one member spends face-to-face with another, thereby decreasing exposure times between individuals, and thus limiting the infection risk