Acute lower respiratory infection in the developing world

Keywords::

• acute lower respiratory infection
• acute respiratory infection
• Africa
• developing countries
• Haemophilus influenzae type b
• Hib
• pneumococcus
• pneumonia
• respiratory syncytial virus
• RSV
• Streptococcus pneumoniae

Pneumonia and acute lower respiratory infection (ALRI) are leading causes of childhood mortality globally. In wealthy countries, it is usually of little importance for public health policy if studies are conservative and underestimate the disease burden. This situation occurs because the cost of hospitalization and other interventions is high enough to provide an economic justification for prevention measures (notably immunization) at relatively low burden levels. Similarly, overestimation of the disease burden often means that what was considered to be an extremely efficient investment is only just highly efficient.

The situation in resource-poor settings is different. In most situations vaccine is not cost saving but rather cost effective, in that the cost per disability adjusted life year saved is less than three-times the per capita gross domestic product (a metric used by the WHO [101]). Several reasons account for why the true economic efficiency of vaccines may be lower in developed countries. Many patients with severe disease do not present for medical care and thus do not incur a cost to the medical care system. For patients that present for medical care, diagnostic (e.g., laboratory and radiology) and therapeutic (e.g., oxygen, antibiotics and mechanical ventilation) interventions may be minimal. Finally, the cost of labor is low in relation to potential interventions since the former is valued within the country’s economic framework while prices for drugs and vaccines are based on the economy of the countries where they are manufactured, which is usually in developed countries.

Moreover, in resource-poor settings, the cost to government health budgets from interventions such as introducing new vaccines can be high in relation to the total health budget. Consequently, a decision to introduce a vaccine to prevent ALRI often results in underfunding other competing healthcare priorities such as malaria or tuberculosis prevention and control.

For these reasons, it is critical for developing countries to have an accurate, rather than a conservative or exuberant, disease burden estimate. In the last several years, summary reviews have been published on all-cause ALRI [1,2], as well as pneumococcal [3], Haemophilus influenzae type b (Hib) [4] and respiratory syncytial virus (RSV) [5] disease burden among children as well as the influenza epidemiology in Africa [6]. What do these studies suggest is the incidence of ALRI, etiology-specific ALRI, and ALRI mortality in the developing world?

Reported burden of disease

All-cause ALRI

All-cause ALRI incidence estimates assist with prioritizing broad-based strategies such as oxygen and antibiotic therapy, case identification and referral, and hospital bed projections. Two reviews have been published recently that evaluated community-based longitudinal studies of ALRI among children <5 years of age [1,2]. In total, 28 studies were included leading to an annual estimated incidence of clinical pneumonia in the developing world of 0.29 episodes per child-year, implying a global 150.7 million new cases per year of which 11–20 million are severe enough to require hospitalization.

Hib & pneumococcus

The most thorough reviews and estimates of Hib and pneumococcal pneumonia incidence were those conducted by Watt et al.[4] and O’Brien et al.[3]. As most Hib and pneumococcal pneumonia is not associated with bacteremia, microbiological documentation requires lung punctures or transtracheal aspirates, neither of which is practically (or ethically) feasible in developing countries. An alternative approach is to use vaccine as a probe to determine disease incidence. This method employs a clinical trial design (ideally, a randomized blinded clinical trial) of a vaccine with known efficacy to determine disease incidence. Vaccine-preventable disease incidence is calculated as incidence in the control group minus incidence in the vaccinated group, and can be applied to any outcome of interest, such as clinical pneumonia, severe pneumonia, radiologically confirmed pneumonia, and others. The final outcome does not give true incidence but rather the incidence of disease preventable by vaccine, which is a product of incidence, vaccine efficacy, vaccine coverage and indirect protection.

As randomized vaccine probe studies are enormously expensive, only four have been conducted for the Hib vaccine (in The Gambia, Indonesia, Bangladesh and Chile), and only the first of these two used a fully randomized blinded trial design. Watt et al.[4] estimated the incidence rate of Hib pneumonia as 1304 per 100,000 per year, with a case fatality ratio (CFR) of 4%, equating to 7.9 million episodes of Hib pneumonia and 0.29 million deaths annually based on the year 2000 birth cohort.

For pneumococcus, four vaccine trials using a randomized design were conducted: in the USA, The Gambia, South Africa and The Philippines [3]. The summary estimated incidence of pneumococcal pneumonia was 2228 per 100,000 per year among children less than 5 years of age, with a CFR of 5%, equating to 13.8 million episodes of pneumococcal pneumonia and 0.74 million deaths annually based on the year 2000 birth cohort.

RSV & influenza

Nair et al. reviewed published and unpublished data on RSV acute respiratory infection incidence. The overall incidence was estimated as 68 per 1000 per year and the incidence of severe RSV ALRI was estimated as 5–35 per 1000 per year [5], accounting for approximately 3.4 million cases based on the year 2005 birth cohort. The review found insufficient data to estimate global RSV mortality, but suggested a range of 0.066 to 0.2 million.

No review of the global childhood influenza burden has yet been published. A recent review of influenza epidemiology in Africa demonstrated that influenza is often identified among children hospitalized with ALRI, but the causal association between ALRI and influenza was not clear, as some studies showed that the proportion of children identified with influenza was the same as that among community controls. The review’s primary conclusion was that existing data are insufficient to guide public health policy in the region [6].

Distribution of burden

Children in all countries are at risk of ALRI. However, severe ALRI cases and ALRI-related deaths are unevenly distributed. For example, the distribution of overall ALRI deaths is predicted to be highly concentrated in a few countries with large birth cohorts or high ALRI mortality incidences [1,2]. Approximately 20% of deaths are estimated to occur in India alone; 43% in India, Nigeria, the Democratic Republic of Congo and Ethiopia; and 70% in 15 countries of which 10 are in Africa, three in the Indian subcontinent and two (China and Afghanistan) in Asia.

For pneumococcus, ten countries were modeled to account for 66% of all pneumococcal cases (of which 96% were pneumonia) including India (27%), China (12%), Nigeria (5%), Pakistan (5%), Bangladesh (4%), Indonesia (3%), Ethiopia (3%), the Democratic Republic of Congo (3%), Kenya (2%) and the Philippines (2%) [3]. Many of these countries are included in this list primarily because of the large birth cohort size. The distribution of pneumococcal deaths is different. Among countries with a pneumococcal mortality incidence >300 per 100,000 per year, all but one (Afghanistan) were located in Africa, giving African countries a disproportionately higher share of deaths relative to cases. The effect of this is that although India had both the highest estimated number of ALRI cases and deaths, the next three countries with the greatest number of ALRI deaths were African, and included Nigeria, Ethiopia, and the Democratic Republic of Congo. Data on Hib were approximately similar. All countries with a modeled Hib mortality rate of at least 200 per 100,000 per year were African except Afghanistan; Nigeria, Ethiopia and the Democratic Republic of the Congo followed India as the countries with the greatest predicted number of Hib deaths. For RSV, over 91% of deaths were estimated to occur in developing countries [5].

Instability in estimates

While the numbers previously reported are easy to communicate, they hide considerable uncertainty in their precision. For overall ALRI rates, the individual studies on which the global burden is determined are not geographically representative or homogeneous, even within regions. In Africa, for example, five studies were identified, four of which were from West Africa and one from Kenya. The ALRI incidence varied from 0.06 to 1.3 episodes per child-year, a variation of over 20-fold; two studies occurred in Nigeria, where the reported incidences varied from 0.08 (among measles-immunized children in Ibadan) to 1.3 episodes per child-year (among all children in Ilorin). In Asia, incidence varied from 0.07 to 2.45 episodes per child-year (a 35-fold variation) and among the four studies from India incidence varied from 0.07 to 0.54 per child-year. In Central and South America incidence varied from 0.04 to 1.8 episodes per child-year (a 45-fold variation).

The estimate of all-cause severe ALRI incidence is more problematic. It was based on a total of six studies, two of which were not included in the estimate of overall incidence because the study period was less than one year. Representation was limited to West Africa, the Indian subcontinent and Syria, and proportions of ALRI that were severe were calculated from the studies and then applied to global case counts. Moreover, the case definitions for both numerators and denominators varied across studies. For example, severe pneumonia was defined as lobar consolidation [7], hospitalization due to pneumonia as defined by a clinician [8], a respiratory rate of at least 70 plus intercostal retractions [9], a respiratory rate of at least 50 plus intercostal retractions [10], increased respiratory rate, stridor or intercostal retractions [11], and failure to eat or drink, cyanosis or intercostal retractions [12]. Not surprisingly, the proportion of ALRI that was severe varied considerably.

Numerous possible reasons exist for these variations. Factors that may lead to true differences in burden include the underlying health of the study population, such as nutritional status and HIV prevalence; natural and vaccine-induced population immune status to common ALRI etiologies; immunity to etiologies that may promote severe ALRI such as measles; the epidemiological context in which studies were conducted including rainfall, elevation, Saharan winds, temperature, indoor smoke exposure, and others; and the presence of novel infecting agents during the study period, for example, the emergence of a pandemic influenza strain. Sources of bias that may lead to variations in incidence estimates include differences in case definitions, the education and training of clinicians, study duration, the population under investigation, such as age distribution, and the frequency of subject assessment for the presence of ALRI.

Similar to global ALRI incidences, variation in incidences between studies existed for pneumococcus and Hib. This variation was less than for all-cause ALRI, likely due in part to the lower number of studies documenting Hib and pneumococcal incidence.

The RSV ALRI burden review documented that the reported annual severe RSV ALRI incidence among children <5 years of age varied by study/site almost 20-fold from two to over 50 per 1000. With respect to RSV mortality, the authors concluded ‘national, regional, and global RSV mortality also varies widely from year to year’. In addition to wide variations across studies, a similar degree of variation in incidence may be seen at a single site over different years; for example, over 3 years on one island in Indonesia, our study team estimated that seasonal incidence by district and year varied approximately tenfold [13].

Methodological limitations

Several concerns exist specific to estimates of Hib and pneumococcal pneumonia incidence. First, some studies – including the two main reviews – model incidence estimates, to some extent, from percentage reductions in outcomes, instead of absolute incidence reductions. This issue is problematic primarily because the proportion prevented will vary depending on the incidence of other diseases. For example, during a major influenza epidemic, the proportion of pneumonia prevented by Hib vaccine is likely to decline. Likewise, if population immunity to RSV immunity is high, the proportion of pneumonia prevented by pneumococcal conjugate vaccine is likely to rise.

A second example occurs with estimates of pneumococcal mortality (which is due primarily to pneumococcal pneumonia). The sole robust data point for this outcome derives from the landmark study in The Gambia. In this study, 9-valent pneumococcal conjugate vaccine reduced all-cause mortality by 16%, with an absolute rate reduction of 4.8 from 30 to 25 per 1000 child-years [14]. In The Gambia these figures are equivalent. But how should one apply them elsewhere? In a setting where mortality is 150 per 1000 child-years, will a vaccine produce a new mortality rate of 145 or 126 per 1000 child-years?

Second, because Hib and pneumococcal pneumonia incidence have been estimated only from vaccine probe studies, by definition they do not apply to children younger than the age at which a vaccine is first administered – usually 6 weeks or 2 months. To the extent that a single dose of vaccine has limited effectiveness, this age may be increased to the age at which dose two is administered, usually 10 weeks or 3 months. One also must consider that – in the developing country setting in which most vaccine probes were conducted – immunization is often administered later than the target age. This is a large issue particularly for mortality as almost half of infant deaths occur during the neonatal period (the first 28 days of life) and the percentage occurring during the first 3–4 months will be even higher. If causes of death were identical at different infant ages, the indirect effects of Hib and pneumococcal conjugate vaccines would allow extrapolating protection from older to younger infants. However, causes of mortality differ between the neonatal period (when preterm birth, congenital anomalies, and perinatal complications are common) versus the postneonatal period [15]. In summary percentage reductions in outcomes apply only to children between the ages of 3–4 months and 2 years (the age at which follow-up stopped in most studies) and should not be extrapolated to other age groups.

Third is the use in models of radiographically confirmed pneumonia as a proxy for all severe or hospitalized pneumonias. For the clinical trials of Hib and pneumococcal vaccines that used a vaccine probe approach, the definition of radiographically confirmed pneumonia was standardized through a process organized by the World Health Organization in an effort to facilitate comparison of epidemiological results between study sites [16]. The precise definition is lengthy, but can be summarized as an obvious fluffy (i.e., alveolar) opacity, or pleural effusion, as agreed upon by two out of three independent readers. This outcome was not developed to design public health policy and for at least two reasons may not be appropriate: it constitutes only a fraction of all severe pneumonia (˜20%) and does not appear to predict severity [17].

As an illustration, in the Hib vaccine probe studies, the proportion of all severe or hospitalized pneumonia prevented by vaccine was 4-7% (18). However, the number frequently communicated is the 20% reduction in radiologically confirmed pneumonia reported in The Gambian and some other trials. In the systematic review of Hib disease, the Hib severe or hospitalized pneumonia incidence [4] was approximately five- to six-fold higher than the vaccine-preventable severe pneumonia annual incidences of 200–300 per 100,000 children <2 years of age reported (or calculated) from the original vaccine probe studies [18]. This discrepancy could have resulted from failure to detect the majority of cases. However, at least in Indonesia our study team implemented measures to achieve high case ascertainment, such as inclusion of hospitalized and nonhospitalized persons, use of village aides to identify and refer cases, and removal of financial and other barriers [19].

Fourth, the total of projected cases from the three reviewed causes of pediatric ALRI (Hib, pneumococcus and RSV) appears to exceed the number of all pediatric ALRI cases. As mentioned previously, the overall summary estimate of severe ALRI is 11–20 million cases per year. Within the three etiologies for which a systematic review has been published, there are estimated to be 14 million cases due to pneumococcus, 8 million due to Hib, and 3 million due to RSV, for a total of 25 million cases, or 5 million more than the high-end estimate for all-severe ALRI. In addition to etiologies with systematic reviews, pneumonia may be caused by influenza, human metapneumovirus, parainfluenza, adenovirus, Klebsiella sp., nontypable H. influenzae, Mycobacterium tuberculosis, Chlamydia pneumoniae and others. Even accounting for an enlarging birth cohort, it appears that either the global ALRI specific estimates are too low, the etiology specific estimates too high, or etiologies occur together.

Data on ALRI deaths are more consistent, with an estimated total case count of 2 million, of which 0.54–0.81 were reported as resulting from pneumococcus, 0.21 to 0.43 million from Hib, and 0.066–0.20 frpm RSV leaving 0.56–1.2 million due to other etiologies. The observation that potentially over half of global ALRI deaths result from etiologies other than pneumococcus, Hib and RSV should prompt interest in conducting studies to further define etiology-specific mortality incidences. This observation also emphasizes the usefulness of etiology nonspecific measures to reduce mortality, such as handwashing [20], HIV prevention and treatment in endemic areas [21], increased access to care, and provision of basic interventions such as antibiotics and oxygen [22].

Summary, conclusions & recommendations

The global burden of pediatric ALRI is high. Regardless of the uncertainty surrounding absolute incidence estimates and case counts, all published reviews (overall and etiology-specific) are consistent in reporting that the greatest burden of disease remains in countries with large birth cohorts or high incidences of severe disease and death. Remarkably, of the 194 countries globally, approximately 15 experience the majority of severe ALRI cases and deaths.

The great expense of the vaccine probe studies necessary to determine Hib and pneumococcal pneumonia incidence has limited the number of data points available. The great variation in overall ALRI incidences reported in different studies adds more uncertainty. Even for vaccine probe studies, methodological issues greatly affect the interpretation of disease burden.

Overall, there is a need for better ALRI burden data and to communicate accurately the data that exist. This need not delay introduction of interventions, particularly in sites where, despite imprecision, burden is known to be high. In this spirit, I propose the following conclusions and recommendations:

• • Global estimates of ALRI burden are imprecise and likely to vary substantially by month, year and location;

• • While global estimates are useful to motivate international efforts (such as vaccine development or financing), decisions on implementation occur at a national level. Consequently, there is a need for better data to be collected in a systematic way across sites, so that decision-makers can understand the absolute and relative burden of ALRI in their countries;

• • For vaccine probe studies, absolute rate reductions rather than proportion reductions should be considered for modeling as the former may be less affected by changes in nontargeted etiologies;

• • Radiologically confirmed pneumonia should not be used in models as a surrogate for severe or hospitalized pneumonia. Severe or hospitalized pneumonia is a more appropriate outcome, as this outcome is likely to be of most importance to decision-makers;

• • Communication on the likely impact of vaccine should reflect the best data available, including addressing nuances in study design. For example, Hib vaccine is likely to prevent 4–7% of severe or hospitalized ALRI and achieve a rate reduction in documented cases of at least 200–300 per 100,000 children per year and potentially substantially more. Similarly, serotype 1 containing pneumococcal conjugate vaccine reduced mortality among children aged at least 6 weeks by 16% and five per 1000 child-years in The Gambia; may achieve one or the other of these results in other areas of Western Africa; and will have an unknown impact on postneonatal mortality in other regions;

• • International organizations with a global viewpoint should target the 10–15 countries with the greatest burden of ALRI deaths, including a particular focus on India and Nigeria. Interventions should include broad-based measures such as handwashing, increased access to care, provision of appropriate clinical care, and HIV prevention and treatment; and etiology-specific prevention measures, such as pneumococcal and Hib immunization;

• • Ongoing data are needed on the proportion of children with severe ALRI who present for modern medical care, on influenza and RSV-specific mortality by region, and on the contribution of other etiologies to pediatric ALRI and ALRI mortality, including in the neonatal period.

Financial & competing interests disclosure

Bradford D Gessner works for AMP, which receives unrestricted grant support from sanofi-aventis and research support from sanofi-aventis, Pfizer, Merck and GlaxoSmithKline. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.