Table 5

Biomass fuel use and ALRI in children under 5 in developing countries

Study Design Case definition Exposure Confounding adjusted Comments OR (95% CI)
Rural South Africa (1980) Natal (Kossove)58 Case control, 0–12 months, 132 cases, 18 controlsOutpatient cases: Wheezing, bronchiolitis & ALRI; Clinical + x ray. Controls: Non-respiratory problemsAsked: “Does the child stay in the smoke?” Prevalence = 33%Routine data collection:
• number of siblings
• economic status
Examined, not adjusted
Only 63% of 123x rayed had pneumonic changes. Control group was small. Exposure assessment was vague.4.8 (1.7 to 13.6)
Rural Nepal (1984–85) Kathmandu Valley (Pandeyet al)53 Cohort, 0–23 months, 780 (study 1), 455 (study 2)Two-weekly home visits: ARI grades I–IV (Goroka) BreathlessnessAsked mothers for average hours per day the child near fireplace. In study 1, same team asked about exposure and ARI > possible bias 77% exposed over 1 hourSince homes were “homogeneous” confounding not taken into accountDose response relationship found Exposure assessment not validated2.2 (1.6 to 3.0)
Rural Gambia (1987–88) Basse (Campbellet al)54 Cohort, 0–11 months, 280 Weekly surveillance. Mother's history of “difficulty with breathing” over subsequent 3 month periodReported carriage of child on the mother's back Prevalence = 37% Adjusted for
• birth interval
• parental ETS
• crowding
• socioeconomic score
• nutritional indicators
• vaccination status
• no. of health centre visits
• ethnic group
• maternal education
• other
Father's ETS only other significant factor. Cautious about interpretation, ability to deal with confounding, and to establish causation where exposure and incidence high2.8 (1.3 to 6.1)
Urban, Argentina (1984–87) Buenos Aires (Cerqueiro et al)67 Case-control, 0–59 months Cases: 516 inpatients; 153 outpatients, Controls: 669Three hospitals: Cases: ALRI within previous 12 days Controls: well baby clinic or vaccination, matched by age, sex, nutritional status, socioeconomic level, date of visit, and residence.Interview with mother: Household heating by charcoal; heating with any fuel; bottled gas for cookingNone, but success of matching verified. Multivariate analysis “currently underway”No data available re charcoal heating in outpatient households. Chimney smoke nearby found to be associated (OR 2.5–2.7) with ARLI in both kinds of patients. ETS not significant for either9.9 (1.8 to 31.4) for charcoal heat for inpatients 1.6 (1.3 to 2.0) for any heating fuel in inpatients
2.2 (1.2 to 3.9) for gas cooking in outpatients
Rural Zimbabwe (?) Marondera (Collingset al)57 Case control, 0–35 months, 244 cases, 500 controls Hospital: Cases: Hosp ALRI, clinical and x ray. Controls: Local well baby clinic (a) Questionnaire on cooking/exposure to woodsmoke (b) COHb (all) (c) TSP (2 h during cooking): 20 ALRI and 20 AURI cases 73% exposed to open fireQuestionnaire:
• maternal ETS
• overcrowding
• housing conditions
• school age sibs
• paternal occupation not adjusted
Confounding: only difference was number of school age sibs, but not adjusted. COHb not different between ALRI and AURI. TSP means: ALRI (n=18) 1915 μg/m3 AURI (n=15) 546 μg/m3 2.2 (1.4 to 3.3)
Rural Gambia (?) Upper River Division (Armstrong and Campbell)56 Cohort, 0–59 months, 500 (approx.) Weekly home visits: ALRI clinical andx rayQuestionnaire: Carriage on mother's back while cookingQuestionnaire:
• parental ETS
• crowding
• socioeconomic index
• number of siblings
• sharing bedroom
• vitamin A intake
• no. of wives
• no. of clinic visits
Adjusted in MLR
Boy/girl difference could be due to greater exposure. Report carriage on back quite a distinct behaviour so should define the two groups fairly clearly with low level of misclassificationApproach (1) (all episodes) M: 0.5 (0.2 to 1.2) F: 1.9 (1.0 to 3.9)
Approach (2) (1st episode) M: 0.5 (0.2 to 1.3) F: 6.0 (1.1 to 34.2)
Urban Nigeria (1985–86) Ibadan (Johnson and Aderele)62 Case control, n=103+103, 0–59 monthsCases: Hospitalised for ALRI (croup, bronchiolitis, pneumonia, empyema thoracis) based on clinical,x ray, and biolab workup. Controls: infant welfare clinic, age and sex matched, no respiratory diseaseInterview
Type of cooking fuel used at home (wood, kerosene, gas)
NoneAge, nutritional status, ETS, crowding, and location of cooking area also not significant. NS
Urban Nigeria (1985–86) Ibadan (Johnson and Aderele)62 Case fatality, n=103, 0–59 monthsCases: Death in hospital among ALRI patients (see above)Interview
Type of cooking fuel used at home (79 = kerosene, gas = 5, wood = 16, other = 3)
NoneOverall case fatality rate = 7.8%. 5 of 8 deaths were from wood burning homes; one additional death had partial exposure to wood smoke. Poor nutrition (1.8×), low income (1.5×), low maternal literacy (2.1×) were more frequent in wood burning homes. ETS rates were similar. Yet paternal income, maternal education, household crowding, ETS not related to case fatality rate12.2 (p<0.0005) for those exposed to wood smoke compared with those exposed to kerosene and gas
Rural Tanzania (1986–87) Bagamoyo District (Mtango et al)60 Case-control
Cases: ALRI deaths = 154
Other deaths = 456
Controls = 1160 0–59 months
Cases: Verbal autopsy certified by physician of all deaths in period. Controls: Multistage sampling (40 of 76 villages). Children with ALRI were excludedHousehold interview;
• Child sleeps in room where cooking is done
• Cook with wood
Village, age, questionnaire respondent, maternal education, parity, water source, child eating habit, whether mother alone decides treatment.About 95% of all groups cook with wood. No tendency to be different distances from road. Perhaps confusion of ALRI with other diseases (e.g. measles). Water not from tap had OR = 11.9 (5.5 to 25.7). Models with all deaths, pneumonia deaths, and non-pneumonia deaths all had same significant risk factors. No difference in source of treatment by location where child sleeps. Maternal education, religion, crowding, and ETS, not significantAll deaths: 2.8 (1.8 to 4.3) for sleeping in room with cooking. 4.3 for pneumonia only. 2.4 for other deaths
Rural Gambia Upper River Division (de Francisco et al)61 Case-control
Cases: 129 ALRI deaths
Controls: 144 other deaths
270 live controls
0–23 months
Cases: Verbal autopsy confirmed by 2 of 3 physicians. Controls: Matched by age, sex, ethnic group, season of death, and geographic areaIndoor air pollution index based on location and type of stove, carrying of child while cooking, and parental ETS (details not provided)Cases vs. live controls: Adjusted for significant factors in univariate analysis: socioeconomic score, crowding, parental ETS, and nutrition indicators plus maternal education. No significant factors for cases vs. dead controls.Only other significant risk factor remaining after multiple conditional logistic regression was whether child ever visited welfare clinic OR = 0.14 (0.06 to 0.36) Misclassification of ALRI deaths (e.g. confusion with malaria) is possible reason for lack of significant difference between cases and dead controls.5.2 (1.7 to 15.9) for cases vs. live controls
Urban Brazil (1990) Porto Alegre (Victoraet al)64 Case control, 0–23 months, 510 cases, 510 controlsCases: ALRI admitted to hospital, clinical and X-ray. Controls: Age matched, neighborhoodTrained field worker interview:
• Any source of indoor smoke (open fires, woodstoves, fireplaces)
• Usually in kitchen while cooking
• cigarettes smoked
• housing quality
• other children in hh
• income/education
• day centre attendance
• history of respiratory illness
• (other)
Hierarchical model/MLR
Only 6% of children exposed to indoor smoke. Urban population with relatively good access to health care. Not representative of other settings in developing countriesIndoor smoke: 1.1 (0.61 to 1.98)
Usually in the kitchen: 0.97 (0.75 to 1.26)
Urban and rural India (1991) South Kerala-Trivandrum (Shah et al)63 Case control, 2–60 months, 400 totalHospital: Cases: Admitted for severe/very severe ARI (WHO definition). Controls: Outpatients with non-severe ARIHistory taken, including
• type of stove, with “smokeless” category
• outdoor pollution
• smokers in house
• number of siblings
• house characteristics
• socioeconomic conditions
• education
• birth weight etc.
Adjusted in MLR
This is a study of the risk factors for increased severity, as the controls have ARI (non-severe). On MLR, only age, sharing a bedroom, and immunisation were significant. Exposure assessment was vague and invalidated“Smokeless” stove: 0.82 (0.46 to 1.43).
Rural Gambia (1989–1991) Upper River Division (O'Dempsey et al)55 Prospective case-control, n=80+159, 0–59 monthsAttending clinic. Cases: if high respiratory rate, transported to Medical Research Council where physician diagnosed pneumonia after lab tests and x ray. Controls: selected randomly from neighbourhood of cases, matched by ageHousehold questionnaire: Mother carries child while cookingAdjusted for mother's income, ETS, child's weight slope, recent illness, and significant illness in last six months.No effect of bednets, crowding, wealth, parental education, paternal occupation, age of weaning, and nutritional status. ETS OR = 3.0 (1.1 to 8.1). Aetiological (preventive) fraction for eliminating maternal carriage while cooking = 39%; for eliminating ETS in house = 31%. May be reverse causality, i.e. sick children being more likely to be carried.2.5 (1.0 to 6.6)
  • This list is confined to quantitative studies that have used internationally standardised criteria for diagnosing ALRI. There are additional studies that have noted a relationship with various respiratory symptoms including cough, runny nose, noisy respiration, and sore throat—for example, the study in Lucknow, India by Awasthiet al 68 which is discussed in the text.