Review
Seasonal influenza epidemiology in sub-Saharan Africa: a systematic review Bradford D Gessner, Nahoko Shindo, Sylvie Briand
Acute respiratory infection (ARI) is a leading cause of mortality worldwide, of which influenza is an important cause that can be prevented with vaccination. We did a systematic review of research published from 1980 to 2009 on seasonal influenza epidemiology in sub-Saharan Africa to identify data strengths and weaknesses that might affect policy decisions, to assess the state of knowledge on influenza disease burden, and to ascertain unique features of influenza epidemiology in the region. We assessed 1203 papers, reviewed 104, and included 49 articles. 1–25% of outpatient ARI visits were caused by influenza (11 studies; mean 9·5%; median 10%), whereas 0·6–15·6% of children admitted to hospital for ARI had influenza identified (15 studies; mean 6·6%; median 6·3%). Influenza was highly seasonal in southern Africa. Other data were often absent, particularly direct measurement of influenza incidence rates for all ages, within different patient settings (outpatient, inpatient, community), and for all countries. Data from sub-Saharan Africa are insufficient to allow most countries to prioritise strategies for influenza prevention and control. Key data gaps include incidence and case-fatality ratios for all ages, the contribution of influenza towards admission of adults to hospital for ARI, representative seasonality data, economic burden, and the interaction of influenza with prevalent disorders in Africa, such as malaria, HIV, and malnutrition.
Introduction
Methods
Pneumonia is the leading cause of paediatric mortality worldwide1,2 and infections such as Streptococcus pneumoniae, Haemophilus influenzae type b, and respiratory syncytial virus are well recognised contributors to global disease burden.3–7 Recent studies have documented the substantial role of influenza in respiratory infection morbidity and mortality at all ages,8–17 although results from other studies have cast doubt on some of the reports from elderly patients.18,19 Modelling analyses have estimated yearly excess mortality attributable to influenza in the USA to be about 30 000 between 1976 and 2007, depending on the year and underlying assigned cause.16 Other studies have documented the effect of influenza on work11,20 and school absenteeism11,21,22 and on the economic costs of influenza.23–27 Influenza also contributes to acute respiratory infection (ARI) burden by increasing the risk of bacterial pneumonia, particularly that caused by pneumococci.28–31 About 1·1 billion people live in sub-Saharan Africa, an area that has high population growth rates, low life expectancy, and many of the world’s poorest countries.32 Despite the abundance of information from many areas of the world, little is known about influenza epidemiology in sub-Saharan Africa,33 and no systematic review has been published. We aimed to help identify data gaps that might affect policy decisions (such as routine vaccine use and target groups and the need for antiviral drugs), to assess the state of knowledge on seasonal influenza epidemiology in sub-Saharan Africa, and to ascertain unique features of influenza epidemiology in the region. We concentrated on sub-Saharan Africa because epidemiological, socioeconomic, and vaccine policy factors in northern Africa are likely to be substantially different. We focused on seasonal human influenza because data on pandemic or avian and animal influenza are unlikely to greatly inform routine yearly influenza prevention and treatment decisions.
Search strategy and selection criteria
www.thelancet.com/infection Vol 11 March 2011
Lancet Infect Dis 2011; 11: 223–35 Agence de Médecine Préventive, Paris, France (B D Gessner MD); and Global Influenza Programme, WHO, Geneva, Switzerland (N Shindo MD, S Briand MD) Correspondence to: Dr Nahoko Shindo, WHO Global Influenza Programme, Health, Security, and Environment, 20 Avenue Appia, CH-1211, Geneva 27, Switzerland
[email protected]
We searched the National Library of Medicine through PubMed for (“influenza” AND “Africa”) OR (“Africa” AND (“pneumonia” OR “acute respiratory infection”)) OR (“influenza” AND each individual sub-Saharan African country); sub-Saharan African countries were defined by the UN Educational, Scientific and Cultural Organization34 with the addition of Reunion. The search was limited to studies of people, studies published in English, French, or Portuguese, and studies published from Jan 1, 1980, to Dec 31, 2009. We selected the starting date of 1980 because of the generally increasing quality of studies in later years, including more consistent and robust testing, the paucity of articles related to seasonal influenza (versus pandemic or historical accounts of influenza), and the increased difficulty in obtaining journal articles from older publications. We searched references of identified articles for additional articles, and we reviewed abstracts and titles and selected studies if we thought they included some aspect of influenza epidemiology. This approach included papers the focus of which was seasonal influenza epidemiology and studies of vaccine, pneumonia in general (in case influenza testing was included), and other viral respiratory illness. We excluded studies that reported results from outside sub-Saharan Africa, that only had data on avian or pandemic influenza, that only had data on unrelated diseases (including animal influenza), that were limited to historical information or generic worldwide summaries (except to identify original data references), that duplicated results in other more comprehensive reports, or that were limited to immunology, drug resistance, or other nonepidemiological factors. We excluded WHO reports of influenza activity published in the Weekly Epidemiologic Report35–41 because data were not consistently published in this format. WHO routinely enters data received through 223
Review
For more on FluNet see http://www.who.int/flunet
the Global Influenza Surveillance Network into the FluNet system, including data published in the Weekly Epidemiologic Reports. We used this system to assess seasonality in the three countries that consistently reported data (Madagascar, Senegal, and South Africa). However, because this system is based on four levels of influenza activity, it could not be used for other purposes. Because we aimed to catalogue and summarise all existing data for this systematic review, we did not use quality screens for studies. The one exception was that we excluded studies that reported seroprevalence data based on a single serological test rather than paired specimens, because this method does not usually provide information on current disease.
Data abstraction and synthesis
See Online for webappendix
BDG abstracted data directly into a structured Excel database. Partly because data-quality screens were not used, we did not validate article selection and data abstraction. We sought data for incidence, proportion of disease caused by different causes, seasonality, underlying illness, concurrent illness, age, outbreak context, and case-fatality ratio. Because of the different methods used in the studies, we did not use summary measures apart from for the proportions associated with influenza among paediatric cases of ARI admitted to hospital and outpatients with ARIs. For these outcomes, summary measures were limited to ranges, means, and medians. Data synthesis mainly consisted of reporting the key findings of individual studies. The main risks of bias in comparison and synthesis of studies were different inclusion criteria and variation in tests used. Another risk of bias was the different times of the year during studies—given the known or potential
1203 articles identified and titles and abstracts screened for eligibility 1187 from database search 16 from other sources
104 potentially relevant articles screened for full review
49 articles included
Figure 1: Flow chart for study selection
224
1099 articles excluded due to lack of relevance for one or more of the following reasons: Influenza not main topic Review or historical article Article focused on avian or animal influenza Article not based in sub-Saharan Africa Article not an epidemiology study Articles on pandemic influenza
55 articles excluded for one or more of the following reasons: Article was worldwide summary from WHO or US CDC Article on pneumonia epidemiology without influenza testing Article not an epidemiology study Review or historical article Article repeated more comprehensive data Article only had a single sample serology
seasonality of influenza, unadjusted analysis might underestimate or overestimate results. To address this issue, when including studies with fewer than 12 months of surveillance in summary calculations of proportions, we assumed that no influenza occurred outside the surveyed months and that the average number of admissions to hospital was the same for surveyed and non-surveyed months. The value for these studies was then calculated as the reported proportion divided by 12 months per year divided by the number of months of surveillance. For studies of at least 12 months, we used reported results. We did subanalyses of studies with durations that were multiples of 12 months, but no conclusions changed. The primary outcomes were incidence of influenza, proportion of individuals treated in hospital or as outpatients for ARI caused by influenza, influenza casefatality ratios in individuals admitted to hospital and during outbreaks, relative frequency of influenza identified in association with ARI compared with that of other viruses, influenza seasonality and risk groups, contribution of influenza outbreaks, the effect of underlying illness on influenza risk, and the association between influenza and other pathogens or disorders. No detailed protocol was written and the review is not registered; instead the authors met and agreed upon the methods and goals as delineated in formal terms of reference.
Results We screened the titles and abstracts of 1203 published articles (figure 1). 49 are included in this report.42–90 Studies varied substantially on the basis of the number of years of assessment, age-groups included, number of influenza isolates obtained, testing methods implemented, and clinical case definitions used (table 1). From 1980 to 2009 there was a modest increase in the number of studies published each year. Although studies originated from 14 countries, only South Africa and Madagascar produced more than five included studies, and seven countries produced only one study each (figure 2, webappendix p 1). Seven countries had data on seasonality reported over at least 12 consecutive months. However, for studies in three countries, data included only nine (Kenya48), 15 (Nigeria62), and 14 (The Gambia86) influenza cases. Additionally, data from the Nigerian paper had inconsistencies between data reported in the abstract and the main results. Data from a second study from The Gambia86 were limited to young children admitted to hospital. For the remaining four countries, data were reported on outpatient influenza over surveillance periods of 2 years to 13 years (figure 3).52,54,55,68,70,73,89 Seasonality was strong in southern Africa (ie, Zambia, Madagascar, and South Africa) and weak in Senegal, which is close to the equator. In addition to these data from peer-reviewed publications, Senegal, Madagascar, and South Africa consistently reported influenza activity www.thelancet.com/infection Vol 11 March 2011
Review
to WHO during 2000–2009 and these same patterns were noted (figure 4; webappendix p 2).91 Eight papers that reported seven different laboratoryconfirmed influenza outbreaks were identified—one in South Africa72 and one in Zambia90 in school-aged children, four in Madagascar (two of which reported the same outbreak),50,53,56,58 one in the Democratic Republic of the Congo,45 and one in Reunion,66 among people of all ages (table 1). The proportion of samples from which influenza was identified ranged from 6% to 95% (mean 41%; median 28%). Three studies reported clinical influenza attack rates that ranged from 34% to 67% (mean 49%; median 47%).45,56,58 In the Democratic Republic of the Congo45 and Madagascar,56,58 attack rates were reported from a few communities known to be experiencing high Tested specimens (n)
rates of disease. For example, in Madagascar, the attack rate was 67% in 750 residents of a small village, but the capital city did not report unusual ARI activity. In the Democratic Republic of the Congo, the attack rate was 47% in 2629 surveyed residents of a single district, which was hypothesised to be a small fraction of the total extent of the outbreak. The residential college outbreak in South Africa involved an attack rate of 34%.72 Case-fatality ratios in people with confirmed influenza were not reported in any of the studies evaluating outbreaks. Two studies reported case-fatality ratios in clinically defined cases. In the Democratic Republic of the Congo,45 case-fatality ratio estimates ranged from 0·8% (six of 792) to 1·4% (18 of 1245) depending on the evaluation design, with higher case-fatality ratios in children younger than 5 years (3·5%) and adults older than 65 years (3·2%).
Months of follow-up; dates With influenza virus (n)
Age range
Clinical setting
Test method
Clinical presentation of tested patients
Outcomes included
Influenza type (n)
Burkina Faso, Bobo-Dioulasso42
NA
NA
7; March 2003–June 2003; February 2006–March 2006
4–29 years
Community
NA
Sore throat, rhinitis
CI
NA
Chad, N’Djamena43
124
0
4; February 1988–May 1988
All ages
Hospital (cases), community (controls)
Culture
Meningitis (cases)
CI
NA
Côte d’Ivoire, Abidjan44
211
30
24; January 2003–December 2004
General population
Outpatient
Culture, ELISA
ILI
OP, A
A (H3), 21 A (H1), 1 B, 8
Democratic Republic of the Congo, Bosobolo District45
6
4
2; November 2002–December 2002 All ages
Outpatient
ELISA, PCR, serology
ILI
O
A (H3N2), 4
Ethiopia, Addis Ababa46
104
7
24; 1987–89
Hospital
Serology
CAP
IA
A, 4 B, 3
Kenya, Nairobi47
600
248
Undefined
Kenya, Nairobi48
822
9
Kenya, Coast Province49
228
Madagascar, Antananarivo50
14–75 years
8; January 2005–August 2005
All ages
Outpatient
IFA
24; November 1981–October 1982
