Risk factors for asthma among schoolchildren who participated in a case-control study in urban Uganda
Abstract
Data on asthma aetiology in Africa are scarce. We investigated the risk factors for asthma among schoolchildren (5–17 years) in urban Uganda. We conducted a case-control study, among 555 cases and 1115 controls. Asthma was diagnosed by study clinicians. The main risk factors for asthma were tertiary education for fathers (adjusted OR (95% CI); 2.32 (1.71–3.16)) and mothers (1.85 (1.38–2.48)); area of residence at birth, with children born in a small town or in the city having an increased asthma risk compared to schoolchildren born in rural areas (2.16 (1.60–2.92)) and (2.79 (1.79–4.35)), respectively; father’s and mother’s history of asthma; children’s own allergic conditions; atopy; and cooking on gas/electricity. In conclusion, asthma was associated with a strong rural-town-city risk gradient, higher parental socio-economic status and urbanicity. This work provides the basis for future studies to identify specific environmental/lifestyle factors responsible for increasing asthma risk among children in urban areas in LMICs.
eLife digest
Asthma is a chronic disease of the airways that leads to breathing difficulty and sometimes death: the condition affects about 235 million people worldwide, especially children. Scientists still do not know exactly what causes asthma, but studies in Europe and North America suggest that individuals born or raised in rural areas are less likely to be affected. However, few studies have examined asthma in African countries, where urbanization is often quickly increasing. Examining the factors associated with the disease as more people move to cities may provide new clues about how asthma emerges, and how to prevent it.
To this end, Mpairwe et al. conducted a study with over 1,670 schoolchildren in Uganda. Those born or raised in rural areas were least likely to have asthma, but the risk doubled among children from small towns, and tripled in those born or who grew up in the city. Children whose parents had a higher education and socioeconomic status had the highest asthma risk, but more work is required to understand why this is the case.
The study by Mpairwe et al. is the first step towards identifying environmental and lifestyle factors associated with increased asthma risk in Africa. Further studies may help scientists to understand how beginning life in a more urban area plays a role in the development of the disease.
Introduction
Asthma is estimated to affect more than 235 million people globally, and is the most common non-communicable condition among children (World Health Organisation, 2017). In Africa, the prevalence of asthma appears to be increasing (Asher et al., 2006; Addo-Yobo et al., 2007; van Gemert et al., 2011; Zar et al., 2007; Lawson et al., 2017), particularly in urban areas (Addo-Yobo et al., 2007; Morgan et al., 2018), but the causes of this increase are not fully understood. Moreover, asthma has various phenotypes which may have different aetiologies (Lötvall et al., 2011). Asthma risk factors appear to vary internationally, and to differ between high-income countries (HICs) and low-and-middle income countries (LMICs) (Douwes and Pearce, 2002).
There is little previously published data on asthma risk factors from Africa. The few studies reported have suggested that current residence in urban areas is associated with a higher risk of asthma than rural residence in Africa (Botha et al., 2019; Addo-Yobo et al., 2001) and other LMICs (Robinson et al., 2011; Gaviola et al., 2016). The association between helminth infections and asthma has been investigated in Africa and other LMICs, but the findings have been inconsistent across studies (Leonardi-Bee et al., 2006; Mpairwe and Amoah, 2019). Other risk factors for asthma in Africa and other LMICs, similar to those in HICs, include maternal smoking (Ayuk et al., 2018; Arrais et al., 2019), maternal history of asthma (Nantanda et al., 2013), childhood atopic sensitisation (Addo-Yobo et al., 2001; Nyembue et al., 2012) and history of allergy (Nantanda et al., 2013; Mehanna et al., 2018). Previous reports suggest no association between biomass fuels and asthma risk (Thacher et al., 2013; Oluwole et al., 2017a), but increased asthma symptoms (Oluwole et al., 2017a; Oluwole et al., 2017b). Unlike in HICs, higher parental education and socioeconomic status has been associated with asthma among children in Africa (Addo-Yobo et al., 2007; Nantanda et al., 2013; Wolff et al., 2012).
We undertook a case-control study among schoolchildren in an urban area in Uganda, to investigate the main risk factors for asthma and the patterns of allergic sensitisation.
Results
Reference characteristics of participating schools and participant flow
We enrolled participants from 55 schools (32 primary, 23 secondary). Of the 6385 children initially identified from the pre-screening exercise, we were unable to contact 4550 parents/guardians in time for them to attend the parents’ meeting; most of these children were in the boarding section (Figure 1). Of the 1835 who attended the meeting, 97% provided written informed consent for their child to participate in the study. We screened 1779 participants and of these, 77 who had initially reported breathing problems either did not have an asthma diagnosis or did not have asthma symptoms in the last 12 months and were excluded. We enrolled 562 children with and 1140 without asthma, but excluded thirty-two with incomplete data (Figure 1). At enrolment, 477 asthma cases successfully performed the spirometry, and only three of these had FEV1 values less than 80% of predicted values.
Early life risk factors for asthma
Participants had mean age 11 years (range 5–17 years); children with asthma were slightly older, and more likely to have parents with a tertiary education and a reported history of asthma (Table 1). Compared to children born in rural Uganda, children born in any town in Uganda or in the city had an increased risk of asthma [adjusted OR (95% confidence interval (CI)) 2.16 (1.60–2.92)] and [2.79 (1.79–4.35)], respectively. The same pattern was observed for the area where the child spent most of their early life (0–5 years) (Table 1). There were no differences in reported exposure to farm animals, or to cigarette smoke during pregnancy. Children with asthma were less likely to have a BCG scar [0.67 (0.51–0.89)] (Table 1), but the TST response (induration >10 mm) at enrolment was similar between cases (15.6%) and controls (14.4%) [1.03 (0.67–1.58)]. There was no statistical evidence of interaction between parental education and the children’s area of residence in early life, nor interaction between age and any of the asthma risk factors.
Early life risk factors for asthma among schoolchildren enrolled in a case-control study between 2015–17 (N = 1,670)
Risk factors | Asthma cases N = 555 | Non-asthma controls N = 1115 | Adj. OR (95% CI)* |
---|---|---|---|
Age, years, Mean (Range) | 11.42 (3.26) | 10.98 (3.05) | 1.07 (1.03–1.11) |
Sex: Girls (933) | 294 (53.0) | 639 (57.3) | 0.79 (0.63–0.99) |
Mother’s history of asthma [m = 164] | |||
Yes (102) | 61 (12.0) | 41 (4.1) | 3.05 (1.97–4.71) |
Father’s history of asthma [m = 175] | |||
Yes (85) | 65 (12.8) | 20 (2.0) | 6.64 (3.90–11.30) |
Fathers’ highest education level | |||
None/Primary (437) | 99 (17.8) | 338 (30.3) | 1 |
Secondary (594) | 193 (34.8) | 401 (36.0) | 1.61 (1.20–2.16) |
Tertiary (639) | 263 (47.4) | 376 (33.7) | 2.32 (1.71–3.16)† |
Mothers’ highest education attained [m = 4] | |||
None/Primary (577) | 158 (28.6) | 419 (37.6) | 1 |
Secondary (601) | 185 (33.4) | 416 (37.4) | 1.11 (0.85–1.44) |
Tertiary (488) | 210 (38.0) | 278 (25.0) | 1.85 (1.38–2.48)† |
Child’s residence at birth | |||
Rural (352) | 74 (13.3) | 278 (24.9) | 1 |
Town (1,169) | 412 (74.2) | 757 (67.9) | 2.16 (1.60–2.92) |
City (149) | 69 (12.5) | 80 (7.2) | 2.79 (1.79–4.35)† |
Child’s residence for most of 0–5 years | |||
Rural (337) | 72 (13.0) | 265 (23.8) | 1 |
Town (1,247) | 434 (78.2) | 813 (72.9) | 1.96 (1.44–2.66) |
City (86) | 49 (8.8) | 37 (3.3) | 3.74 (2.19–6.39)† |
Maternal regular contact with farm animals during pregnancy [m = 252] | |||
Yes (500) | 152 (30.2) | 348 (38.0) | 0.82 (0.64–1.06) |
Maternal cigarette smoking during pregnancy [m = 302] | |||
Yes (36) | 13 (2.7) | 23 (2.6) | 1.07 (0.51–2.22) |
Breast feeding duration [m = 402] | |||
<1 year (328) | 114 (25.3) | 214 (26.2) | 1 |
>1 year (940) | 337 (74.7) | 603 (73.8) | 1.12 (0.85–1.49) |
Allergic rhinitis ever | |||
Yes (862) | 429 (77.3) | 433 (38.8) | 5.53 (4.31–7.09) |
Allergic conjunctivitis ever | |||
Yes (766) | 361 (65.0) | 405 (36.3) | 3.12 (2.50–3.90) |
Eczema ever | |||
Yes (264) | 126 (22.7) | 138 (12.4) | 2.08 (1.56–2.77) |
Urticarial rash ever | |||
Yes (569) | 222 (40.0) | 347 (31.1) | 1.42 (1.13–1.79) |
Presence of a BCG scar on child’s arm at enrolment [m = 83] | |||
Yes (1,315) | 413 (78.4) | 902 (85.1) | 0.67 (0.51–0.89) |
-
Adj. OR = adjusted odds ratio; CI = confidence interval; N = number; m = missing; number (%) in 2nd and 3rd column unless indicated otherwise.
*adjusted for child’s age, sex, residence at birth and father’s education level.
-
†Test for trend p-value<0.0001. For parental history of allergy, 6.6% mothers and 3.4% fathers responded by telephone.
Current features of asthma cases versus controls
Asthma cases were more likely to report a high frequency of ‘trucks passing on the street near their home’ [2.28 (1.52–3.43)]; to come from homes that used electricity/gas for indoor cooking [1.58 (1.16–2.17)]; and to report having used de-worming medication more than twice in the last 12 months [2.18 (1.62–2.93)] than controls (Table 2). There was weak evidence for an inverse association between asthma and infection with any helminths species [0.75 (0.53–1.07)] (Table 2); overall this was not statistically significant as shown by the fact that the confidence interval included the null value of 1, but it was significant for T. trichiura [0.33 (0.13–0.89)] (Supplementary file 1a). Children with asthma were more likely to report a history of allergic diseases such as allergic rhinitis, conjunctivitis, eczema and urticarial rashes (Table 1) and to have these conditions currently (Table 2). The prevalence of current exposure to cigarette smoke (in the household) was similar among cases and controls, although only about 11% were exposed (Table 2).
Current risk factors and features of schoolchildren enrolled in an asthma case-control study between 2015–17 (N = 1,670).
Current features | Asthma cases N = 555 | Non-asthma controls N = 1115 | Adj. OR (95% CI)* |
---|---|---|---|
Frequency of ‘Trucks’ passing on street near child’s home | |||
Rarely (931) | 264 (47.6) | 667 (59.8) | 1 |
Frequently (618) | 232 (41.8) | 386 (34.6) | 1.60 (1.27–2.01) |
Almost all the time (121) | 59 (10.6) | 62 (5.6) | 2.28 (1.52–3.43)† |
Cooking fuel most frequently used indoor | |||
No indoor cooking (390) | 120 (21.6) | 270 (24.2) | 1 |
Charcoal stove (890) | 262 (47.2) | 628 (56.3) | 1.04 (0.79–1.37) |
Gas/Electricity (390) | 173 (31.2) | 217 (19.5) | 1.58 (1.16–2.17) |
Child’s reported regular physical activity levels as recommended by WHO [m = 1] | |||
Yes (999) | 300 (54.0) | 699 (62.7) | 0.71 (0.56–0.90) |
Current exposure to cigarette smoke by a household member§ | |||
Yes (187) | 68 (12.2) | 119 (10.7) | 1.26 (0.90–1.76) |
Use of de-worming medication in last 12 months [m = 1] | |||
None (570) | 151 (27.2) | 419 (37.6) | 1 |
Once (724) | 231 (41.6) | 493 (44.3) | 1.37 (1.06–1.77) |
>Twice (375) | 173 (31.2) | 202 (18.1) | 2.18 (1.62–2.93)† |
Used albendazole in last 12 months [m = 18] | |||
Yes (892) | 352 (63.9) | 540 (49.0) | 1.76 (1.40–2.22) |
Used praziquantel in last 12 months [m = 55] | |||
Yes (393) | 113 (21.0) | 280 (26.0) | 0.84 (0.63–1.11) |
Infection with any helminths at enrolment‡ [m = 127] | |||
Yes (206) | 53 (10.3) | 153 (14.9) | 0.75 (0.53–1.07) |
Allergic rhinitis in last 12 months [m = 1] | |||
Yes (215) | 119 (21.4) | 96 (8.6) | 2.91 (2.13–3.97) |
Allergic conjunctivitis in last 12 months | |||
Yes (177) | 86 (15.5) | 91 (8.2) | 2.10 (1.51–2.92) |
Eczema in last 12 months | |||
Yes (60) | 41 (7.4) | 19 (1.7) | 4.52 (2.54–8.06) |
Urticarial rash in last 12 months | |||
Yes (36) | 18 (3.2) | 18 (1.6) | 1.92 (0.96–3.84) |
-
Adj. OR = adjusted odds ratio; CI = confidence interval; N = number; m = missing; WHO = World Health Organisation. Number (%) in 2nd and 3rd column. *adjusted for child’s age, sex, residence at birth and father’s education level. †Test for trend p-value<0.0001. ‡Helminths infections included mainly Schistosoma mansoni, Trichuris trichuria, Hookworm, and Ascaris lumbricoides; this variable was additionally adjusted for reported worm treatment in the last 12 months. §This included the adolescents’ own smoking history.
Atopy and asthma
Children with asthma were more likely to have allergic sensitisation: SPT positive to at least one of seven whole allergen extracts [2.40 (1.92–3.00)]; and elevated asIgE levels to any of three whole allergen extracts [2.45 (1.53–3.91)], and higher total IgE (Table 3). The most common allergens were dust mites and cockroach. Asthma cases were more likely to have elevated FENO levels [2.57 (2.01–3.29)] (Table 3).
Atopy among schoolchildren enrolled in an asthma case-control between 2015–17 (N = 1,635).
Atopy characteristics | Asthma cases N = 546 | Non-asthma controls N = 1089 | Adj. OR (95% CI)‡ |
---|---|---|---|
Any positive skin prick test to any of 7 allergens | |||
Yes (653) | 302 (55.3) | 351 (32.2) | 2.40 (1.92–3.00) |
Blomia tropicalis (dust mite) SPT | |||
Positive (478) | 241 (44.1) | 237 (21.8) | 2.51 (1.98–3.19) |
Dermatophagoides mix (dust mite) SPT | |||
Positive (535) | 262 (48.0) | 273 (25.1) | 2.46 (1.95–3.10) |
German cockroach SPT | |||
Positive (354) | 150 (27.5) | 204 (18.7) | 1.59 (1.23–2.06) |
Peanut SPT [m = 4] | |||
Positive (57) | 27 (5.0) | 30 (2.8) | 1.73 (0.99–3.05) |
Cat SPT | |||
Positive (22) | 22 (4.0) | 23 (2.1) | 1.94 (1.02–3.69) |
Pollen mix (weeds) SPT [m = 2] | |||
Positive (43) | 22 (4.0) | 21 (1.9) | 2.14 (1.11–4.11) |
Mould mix SPT | |||
Positive (19) | 8 (1.5) | 11 (1.0) | 1.40 (0.52–3.76) |
Fractional exhaled nitric oxide (FENO) [m = 42] | |||
Elevated (>35 ppb) (378) | 199 (36.7) | 179 (17.0) | 2.57 (2.01–3.29) |
Any positive allergen-specific IgE¥ (of 3 allergens, N = 392) | |||
Atopic level (249) | 143 (72.6) | 106 (54.4) | 2.45 (1.53–3.91) |
Dermatophagoides-IgE | |||
Atopic level (186) | 116 (58.6) | 70 (35.9) | 2.32 (1.49–3.63) |
Cockroach IgE | |||
Atopic level (198) | 110 (55.8) | 88 (45.1) | 1.69 (1.08–2.63) |
Peanut IgE | |||
Atopic level (63) | 39 (19.7) | 24 (12.3) | 1.92 (1.05–3.50) |
Total IgE: Mean (SD) kU/L | 930.25 (1246.60) | 640.88 (1146.61) | 1.00024 (1.00005–1.00043) |
-
N = number; Adj. OR = adjusted odds ratio; CI = confidence interval; SD = standard deviation; m = missing; ppb = parts per billion; number (%) shown in 2nd and 3rd column unless indicated otherwise.
*adjusted for child’s age, sex, residence at birth and father’s education level. Skin prick test using whole allergen extracts.
-
¥allergen-specific IgE levels were obtained using ImmunoCAP, on a random sample of 200 cases and 200 controls; standard cut-off for allergic sensitisation (>0.35 allergen-specific kilo units per litre (kUA/L)).
Assessment of different combinations of risk factors for asthma
We investigated the relative importance of area of residence at birth versus the first years of life, on the asthma risk, by looking at children who migrated between rural and urban areas during these two periods. Children born and raised in rural areas had the lowest risk (the reference group); children born and raised in urban areas had the highest risk [2.55 (1.80–3.61)], children born in the urban area who migrated and spent most of 0–5 years in rural areas still had an increased risk of asthma [2.11 (1.14–3.91)], unlike children who were born in rural areas but migrated and spent most of 0–5 years in urban areas (Table 4).
Residence in early life as a risk factor for asthma among schoolchildren enrolled in a case-control study between 2015–17 (N = 1,670)
Child’s residence in early life | Asthma cases N = 555 | Non-asthma controls N = 1115 | Adj. OR (95% CI)* |
---|---|---|---|
Born and spent first 5 years in rural (265) | 50 (9.0) | 215 (19.3) | 1 |
Born in rural, spent first 5 years in urban (87) | 24 (4.3) | 63 (5.6) | 1.54 (0.86–2.75) |
Born in urban, spent first 5 years in rural (72) | 22 (4.0) | 50 (4.5) | 2.11 (1.14–3.91) |
Born and spent first 5 years in urban (1,246) | 459 (82.7) | 787 (70.6) | 2.55 (1.80–3.61) |
-
N = number; Adj. OR = adjusted odds ratio; CI = confidence interval; number (%) shown in 2nd and 3rd column.
*Adjusted for age, sex, and father’s education level.
We investigated the combined effects of the child’s area of residence at birth and parental history of allergic disease. Even among children with no parental history of allergic disease, compared to being born in a rural area, being born in a town [2.15 (1.35–3.44)] or in the city [3.33 (1.61–6.90)] was associated with an increased risk of asthma, Table 5. Compared to the same reference group, children with a parental history of allergic disease had a higher risk of asthma that increased steadily from among children born in rural areas [2.78 (1.58–4.89)], in a town [5.16 (3.25–8.17)] and in the city [5.54 (2.99–10.26)] (Table 5).
Combined effects of residence at birth and parental history of allergic disease as risk factors for asthma among schoolchildren in a case-control study from 2015 to 17 (N = 1,532).
Child’s residence at birth | Parental history of allergy | Asthma cases N = 526 | Non-asthmatic controls N = 1006 | Adj. OR (95% CI)* |
---|---|---|---|---|
Rural | − | 27 (5.1) | 158 (15.7) | 1 |
Town | − | 137 (26.1) | 383 (38.1) | 2.15 (1.35–3.44) |
City | − | 20 (3.8) | 31 (3.1) | 3.33 (1.61–6.90) |
Rural | + | 42 (8.0) | 85 (8.4) | 2.78 (1.58–4.89) |
Town | + | 253 (48.1) | 306 (30.4) | 5.16 (3.25–8.17) |
City | + | 47 (8.9) | 43 (4.3) | 5.54 (2.99–10.26) |
-
Adj. OR = adjusted odds ratio; CI = confidence interval; N = number; ‘−' refers to no history of parental allergy, ‘+' refers to positive history of parental allergy. number (%) in 3rd and 4th column. *Adjusted for child’s age, sex, and father’s education level. Parental history of allergic disease included a history of asthma, eczema, allergic rhinitis, allergic conjunctivitis and any other allergies.
We also investigated the combined effects of area of residence at birth and children’s atopic status (positive SPT to any of seven allergens). Taking non-atopic children born in rural areas as the reference group, we found that non-atopic children born in town had a modest increase in asthma risk [1.72 (1.16–2.55)], which increased further among city-born children [2.63 (1.41–4.91)], Table 6. However, atopic children had a modestly increased risk of asthma even if they were born in the rural area [1.77 (1.02–3.05)], which increased substantially among atopic children born in town [4.61 (3.07–6.90)] or in the city [4.67 (2.56–8.52)], Table 6.
Combined effects of residence at birth and atopy as asthma risk factors among schoolchildren in a case-control study between 2015–17 (N = 1,635).
Child’s residence at birth | Atopy (SPT to any of 7 allergens) | Asthma cases N = 546 | Non-asthmatic controls N = 1089 | Adj. OR (95% CI)* |
---|---|---|---|---|
Rural | - | 42 (7.7) | 191 (17.5) | 1 |
Town | - | 177 (32.4) | 504 (46.3) | 1.72 (1.16–2.55) |
City | - | 25 (4.6) | 43 (4.0) | 2.63 (1.41–4.91) |
Rural | + | 31 (5.7) | 81 (7.4) | 1.77 (1.02–3.05) |
Town | + | 228 (41.7) | 236 (21.7) | 4.61 (3.07–6.90) |
City | + | 43 (7.9) | 34 (3.1) | 4.67 (2.56–8.52) |
-
Adj. OR = adjusted odds ratio; CI = confidence interval; N = number; SPT = skin prick test; ‘-' refers to SPT negative, ‘+' refers to SPT positive to any of seven crude extracts of Blomia tropicalis, Dermatophagoides mix, cockroach, peanut, cat, weeds pollen mix, and mould mix. number (%) in 3rd and 4th column. *Adjusted for child’s age, sex and father’s education level.
We investigated the combined effects of parental education and urbanicity on asthma risk, by looking at the father’s education level in combination with the child’s area of residence at birth. We found that for each level of father’s education, there was an increasing risk gradient from rural-town-city; for each area of residence, increasing father’s education level was associated with increased asthma risk; the highest risk of asthma was among children born in the city whose fathers had a tertiary education [6.95 (3.52–13.71)] (Table 7). The same pattern in Table 7 was seen for mother’s education and for child’s area of residence in the first five years of life.
Combined effects of father’s education and residence at birth as asthma risk factors among schoolchildren in a case-control study from 2015 to 17 (N = 1,670).
Father’s education level | Child’s residence at birth | Asthma cases N = 555 | Non-asthma controls N = 1115 | Adj. OR (95% CI)* |
---|---|---|---|---|
Primary | Rural | 20 (3.6) | 115 (10.3) | 1 |
Primary | Town | 67 (12.1) | 206 (18.5) | 2.13 (1.22–3.74) |
Primary | City | 12 (2.2) | 17 (1.5) | 4.98 (2.02–12.31) |
Secondary | Rural | 26 (4.7) | 109 (9.8) | 1.39 (0.72–2.66) |
Secondary | Town | 156 (28.1) | 268 (24.0) | 3.97 (2.34–6.76) |
Secondary | City | 11 (1.9) | 24 (2.2) | 2.80 (1.16–6.76) |
Tertiary | Rural | 28 (5.0) | 54 (4.8) | 3.22 (1.64–6.33) |
Tertiary | Town | 189 (34.1) | 283 (25.4) | 4.96 (2.89–8.53) |
Tertiary | City | 46 (8.3) | 39 (3.5) | 6.95 (3.52–13.71) |
-
N = number; CI = confidence interval; Adj. OR = adjusted odds ratio; 3rd and 4th column contain n (%); *Adjusted for child’s age and sex. Similar pattern was observed for mother’s education level and child’s residence at birth.
For the combined effect of atopy (SPT positive) and parental history of allergic disease, we found that asthma risk was more than five times among children with both atopy and a parental history of allergic disease [5.74 (4.07–8.10)] (Supplementary file 1b). However, children who had both parents with a history of allergic disease had an effect size similar to children who had only one parent with a history of allergic disease [2.49 (1.74–3.58)] (Supplementary file 1c).
Discussion
We found a step-wise increase in asthma risk according to the child’s area of residence at the time of birth and in their first five years of life. Children born and raised in rural areas had the lowest risk, children born and raised in small towns had a 2-fold increase in risk, while children born and raised in the city had a 3-fold increase in asthma risk. This is the first study in Africa to show such a strong gradient in asthma risk by place of birth. Previous studies have been mostly cross-sectional and conducted in either rural or urban settings, and therefore focusing on current residence; these have shown that the prevalence of asthma is lower among rural residents compared to urban residents (Lawson et al., 2017; Morgan et al., 2018; Botha et al., 2019).
Our findings confirm that the area where a child is born (usually the same as the area where the mother was resident during pregnancy) is important for asthma risk. Indeed, we found that when children moved to other environments, they carried with them the asthma risk related to their area of residence in pregnancy: children born in urban areas who were subsequently raised in rural areas still had a 2-fold increase in asthma risk, similar to their counterparts born and raised in urban areas. These findings are comparable to observations from Europe showing that children born on a farm (usually in rural areas) have a lower risk of asthma in later life, even when they subsequently moved to urban areas (Leynaert et al., 2001), and that children who migrated to Europe after age five had a lower prevalence of asthma, similar to their country of origin, than children who were either born in Europe or migrated before the age of five (Migliore et al., 2007; Kuehni et al., 2007). However, there are also studies that show increased risk of wheezing and allergy among children following migration (Rodriguez et al., 2017; Stein et al., 2016). Unlike studies from Europe and North America which have reported a lower risk of asthma for children born or raised on farms (Genuneit, 2012; Timm et al., 2015), our study found no association between asthma risk and exposure to farm animals either during pregnancy or in early life. We hypothesise that this may be due to ubiquitous farm animal exposure due to subsistence farming, a widespread practice in Uganda, even in towns.
Our observation that children with asthma were more likely to have parents with tertiary education and to use gas or electricity for indoor cooking (as opposed to charcoal stoves) has been made by an earlier study in Uganda (Nantanda et al., 2013). Similarly, our finding that children with asthma reported the highest frequency of ‘trucks passing on the street near their home’ has been reported elsewhere (Sharma and Banga, 2007; Shirinde et al., 2014; Venn et al., 2005). We suggest that these factors are proxy measures of a higher socio-economic status of asthma cases and of urbanicity, consistent with findings from other LMICs that have found a higher prevalence of asthma among children (and adults) in urban than rural areas (Morgan et al., 2018; Robinson et al., 2011; Gaviola et al., 2016). However, our findings contradict those from HICs in which asthma is associated with lower parental education (Lewis et al., 2017) and socio-economic status (Akinbami et al., 2011). This suggests that there may be similarities in lifestyle and environmental factors between the highly educated and high socio-economic status families in LMICs with the low educated and low socio-economic status families in HICs, which increase asthma risk, and therefore require further investigation.
Although children with asthma were more likely to be sensitised to allergens than controls, the pattern of allergic sensitisation was similar among cases and controls; majorly sensitised to house-dust mites and cockroach, and least sensitised to peanut, cat, pollen and mould. This SPT response pattern was similar to other studies from Africa (Nyembue et al., 2012; Mbatchou Ngahane et al., 2016), but different from Europe where the main allergens are dust mite, cat and pollen (Bousquet et al., 2007).
Although maternal smoking is a known risk factor for childhood asthma (Silvestri et al., 2015), our study found no association between maternal smoking and asthma. We attribute this to the low prevalence of smoking in this population, by the mothers during pregnancy (2.5%) and by any household members currently (11%). The lack of association between asthma and indoor cooking with biomass fuels in this study is consistent with other studies from Africa (Thacher et al., 2013; Oluwole et al., 2017a). Indeed, this is consistent with the general pattern of lower risk of asthma in rural areas (where biomass fuel use is highest) than urban areas.
We found an inverse association between asthma and current infection with any helminths, particularly T. trichuria. However, the association between asthma and current infections in Africa has been inconsistent across studies (Mpairwe and Amoah, 2019). What was novel in this study is that we also collected data on history of using de-worming medication in the last 12 months. We found that children with asthma were more likely to have used de-worming medication more than twice in the last 12 months compared to controls, and this was de-worming with albendazole (for geo-helminths) not with praziquantel (for schistosomiasis). This implies that the inverse association we noted between asthma and helminths may be partly explained by increased de-worming among children with asthma. The current de-worming schedule in this age-group in Uganda includes mass-deworming in schools, bi-annually for albendazole and once a year for praziquantel. We did not establish whether the additional de-worming was self-medicated or prescribed by medical workers.
We found that children with asthma were less likely to have a BCG scar, but there was no association with tuberculin skin test at enrolment. In Uganda, BCG vaccination is routinely given at birth. A lack of BCG scar does not mean the vaccine was not administered, but may indicate differences in immune responses among children who will eventually develop asthma. Alternatively, BCG vaccination may be protective against asthma, as suggested by previous studies (El-Zein et al., 2010).
The other important risk factors for asthma, that have been previously described, included parental history of allergic disease (Lim et al., 2010), a child’s atopy status and concomitant other allergic disease (Bao et al., 2017). The strength of this study was to demonstrate that having a combination of any two of these known risk factors for asthma had an additive effect on asthma risk, and that this risk also increased in relation to area of residence in early life, with an increasing rural-town-city gradient. This gradient and independent effects of parental education provide strong evidence for the role of environmental and lifestyle factors associated with urbanisation that are responsible for the increasing asthma risk in urban areas. More investigations are required to identify the specific factors, in order to design interventions to modify them so as to prevent the establishment of asthma risk in early life.
This study had limitations inherent to all case-control studies, such as potential recall bias, selection bias and confounding. We minimised recall bias by focusing on major early life events that a parent was likely to remember. It was re-assuring to note a strong correlation between the recalled events and objective measures such as SPT. We minimised the selection bias for controls by randomly selecting controls from the same class register where the cases were obtained, and by randomly selecting the 400 participants for the asIgE assay. We minimised confounding by adjusting for measured confounders in all our analyses, but cannot rule out the possible role of unmeasured confounders. Finally, a large number of potential participants were not included in this study, mostly because they were in the boarding school section and were unable to contact their parents/guardians in time to attend the parents’ meeting (to provide consent). We do not have information on whether these parents were more likely to reside in rural areas or the city, where the asthma risk is either lower or higher than the town where this study was conducted, respectively. Our results may or may not be generalisable to similar urban areas in Uganda and Sub-sahara Africa.
Our findings support the concept that environmental factors during early life may influence the risk of development of NCDs in later life (Fleming et al., 2018). However, our study does not involve information sufficiently detailed to identify whether exposures in utero or in early life are most relevant. Further research in this setting is required to investigate these hypotheses, and this would have important implications for the life course approach in the prevention of asthma globally.
Conclusion
The risk of asthma among schoolchildren in urban Uganda is strongly predicted by their area of residence in early life, particularly at birth, with the highest risk among children whose early life is spent in small towns and in the city. This risk increases further in the presence of other asthma risk factors such as parental history of allergic disease, children’s own atopy, higher measures of socio-economic status and urbanicity. Given the current rapid urbanisation in Africa and other LMICs, the prevalence of asthma is likely to increase further. This study provides the basis for future studies investigating environmental and lifestyle factors that increase asthma risk in the urban areas of LMICs.
Materials and methods
Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
---|---|---|---|---|
Commercial assay or kit | ImmunoCAPspecific IgE test | Thermo Fisher Scientific, Uppsala, Sweden | http://www.phadia.com/en/Products/ Allergy-testing-products/ImmunoCAP-Assays/sIgE/ |
Study design
Request a detailed protocolWe conducted an un-matched case-control study among schoolchildren, and report following STROBE guidelines (von Elm et al., 2014).
Study population, sampling and consent
Request a detailed protocolThe study base was determined a priori as schoolchildren, 5–17 years old, in primary and secondary schools in Entebbe Municipality and Katabi zone in Wakiso District, Central Uganda. This was a predominantly urban (town) setting. All schools in the study area were invited and 96% participated. Study enrolment was between May 2015 and July 2017. We estimated that a sample size of 2112 would have 80% power to detect odds ratio (OR) <0.5 or>1.5 for exposures with prevalence 10% among controls.
At each school, we pre-screened by registering all children with any breathing problems. We concurrently randomly selected from the class register two children without any breathing problems, using a random number generator programme in STATA (StataCorp, Texas, USA). The children delivered invitation letters to their parents; parents/guardians with telephones were invited to attend a meeting during which those interested provided written informed consent, and children aged ≥8 years provided written informed assent.
Definition of cases and controls
Request a detailed protocolFollowing consent and assent, we screened all participants with the International Study of Asthma and Allergies in Childhood (ISAAC) questionnaire (Asher et al., 1995). Children with a history of wheezing in the last 12 months underwent a detailed medical history and examination by study clinicians, to diagnose asthma. Asthma was defined as a history of recurrent symptoms of wheezing, cough (mostly dry, worse at night and morning) and/or difficulty in breathing experienced in the last 12 months, with or without forced expiratory volume in the first second (FEV1) <80% expected for age, sex and height for African children. Additional medical history included a prior physician diagnosis and good response to asthma medication. If the diagnosis was not straightforward, two clinicians reviewed the participant and if they disagreed, that participant was excluded in order to ensure proper case ascertainment. Children with a history of wheeze or any asthma symptoms but not in the last 12 months were excluded. Controls were defined as having no history of wheeze or any other asthma symptoms. There were no other exclusion criteria.
Ethical approval
Request a detailed protocolThe study was approved by the Uganda Virus Research Institute Research and Ethics Committee, and the Uganda National Council for Science and Technology.
Clinical assessments
Request a detailed protocolWe collected data about asthma risk factors and allergic conditions identified in literature using interviewer-led questionnaires to parents and adolescents, including the ISAAC questionnaire (Asher et al., 1995). When a parent was not available in person and the participant answering the questionnaire did not know that parent’s history of allergy, we telephoned the parent for the relevant information. We looked for the presence of a Bacillus Calmette–Guérin (BCG) scar since BCG vaccination has been inversely associated with asthma previously (El-Zein et al., 2010). We tested for fractional exhaled nitric oxide (FENO), using a hand-held device (NoBreath from Bedfonf Scientific, Maidstone, United Kingdom), and used the manufacturer’s cut-off for children of >35 parts per billion. FENO is considered a biomarker of allergic airway inflammation (Hoyte et al., 2018). We conducted lung function tests for asthma cases using a hand-held spirometer (Micro 1 Diagnostic Spirometer, CareFusion, Chatham Marine, United Kingdom).
Skin prick tests (SPT) and allergen-specific IgE (asIgE) are important measures of allergic sensitisation (Heinzerling et al., 2013). We conducted (SPT) following standard procedures (Heinzerling et al., 2013; Mpairwe et al., 2008), and crude extracts of seven allergens (Dermatophagoides mix of D. farinae and D. pteronyssminus, Blomia tropicalis, Blattella germanica, peanut, cat, pollen mix of weeds, mould mix of Aspergillus species; ALK Abello, Hoersholm, Denmark). A positive response was a wheal diameter >3 mm measured after 15 min, with a negative saline control and positive histamine. We collected blood samples which we processed to obtain plasma that we stored at −80°C. At the end of the study, we randomly selected 200 plasma aliquots from all asthma cases and 200 from all controls for measurement of asIgE to whole allergen extracts (D. pteronyssinus, B. germanica and peanut), using ImmunoCAP (Phadia, Uppsala, Sweden). The standard cut-off for allergic sensitisation of >0.35 allergen-specific kilo units per litre (kUA/L) was used (Key resources table).
Because previous studies have reported an association between asthma and helminths (Mpairwe and Amoah, 2019), we collected three fresh stool samples from each participant and tested for intestinal helminths using the Kato Katz method (Katz et al., 1972). The findings on the association between tuberculin skin test (TST) and allergy/asthma have been inconsistent (Obihara et al., 2006; Eifan et al., 2009). We investigated this association by performing the TST using standard procedures, as we have previously described (Nkurunungi et al., 2012).
Data management and statistical analysis
Request a detailed protocolData were double-entered using OpenClinica open source software version 3.1.4 (OpenClinica LLC and collaborators, Waltham, MA, USA). Data were analysed in STATA.
For continuous variables with clinically relevant cut-off points such as for SPT, asIgE, FENO and TST, we used the dichotomous variables in the analysis. Total IgE was analysed as a continuous variable. The variable for maternal or paternal history of ‘allergic disease’ combines the history of asthma, eczema, allergic rhinitis, allergic conjunctivitis and any other allergies (Greenland et al., 2016). We conducted a complete case analysis, and did not impute missing values.
The key variables adjusted for (see below) were age, sex, area of residence at time of birth and father’s education. Each of these variables (except sex) had small numbers of missing values (the largest number was 29 for father’s education). We therefore created a ‘complete case’ data set, comprising the 555 cases and 1115 controls which had no missing values for these variables. This ‘complete case’ data set was used for all analyses.
The odds ratio was the main outcome measure (as is appropriate for a case-control study), and we also estimated the 95% CIs. We identified age and sex as a priori confounders, and all analyses were adjusted for these variables. We identified area of residence at time of birth and father’s education as potential confounders; we did not also adjust for area where the child spent most of the first five years or mother’s education, since these were strongly associated with the above two factors, and therefore would have introduced collinearity. Previous studies in this setting have found that father’s and mother’s education are significantly associated with socioeconomic status (Aarø et al., 2009).
We did not identify any factors which were likely to be on the causal pathways (and therefore were not potential confounders and should not be adjusted for). We conducted initial logistic regression analyses for each exposure of interest, adjusted for age and sex, with random effects to allow for any clustering by school. We then ran the adjusted random effects logistic regression model (i.e. adjusted for the above factors in addition to age and sex) for each exposure of interest, and checked for collinearity by comparing the standard error of the exposure coefficient in the adjusted model and in the basic model (Greenland et al., 2016; Greenland and Pearce, 2015). We did not find any problems of collinearity, so we reported the findings for the adjusted model for each exposure.
Data availability
Data is available on request via https://datacompass.lshtm.ac.uk/1369/. To gain access to the data please complete the application process via the website. Requests will be reviewed and assessed by the corresponding author, in consultation with the LSHTM's Research Data Manager and relevant LSHTM staff members responsible for research governance and data protection. Applications will be evaluated on the basis of their compatibility with the study's research objectives and the ability to provide de-identified data sufficient to meet the intended purpose, without breaching participant confidentiality or the study's ethical and legal commitments. Successful applicants will be asked to sign a Data Transfer Agreement prior to being provided with the data.
-
London School of Hygiene & Tropical Medicine (LSHTM) Data CompassSONA project - Asthma risk factors data.https://doi.org/10.17037/DATA.00001369
References
-
Parental education as an Indicator of socioeconomic status: improving quality of data by requiring consistency across measurement occasionsScandinavian Journal of Public Health 37:16–27.https://doi.org/10.1177/1403494808086917
-
Risk factors for asthma in urban GhanaJournal of Allergy and Clinical Immunology 108:363–368.https://doi.org/10.1067/mai.2001.117464
-
Asthma prevalence, health care use, and mortality: united states, 2005-2009National Health Statistics Reports 32:1–14.
-
Prevalence of asthma, allergic rhinitis and eczema in 6-7-year-old schoolchildren from Luanda, AngolaAllergologia Et Immunopathologia 47:523–534.https://doi.org/10.1016/j.aller.2018.12.002
-
International study of asthma and allergies in childhood (ISAAC): rationale and methodsEuropean Respiratory Journal 8:483–491.https://doi.org/10.1183/09031936.95.08030483
-
Environmental risk factors for asthma in 13-14 year old african childrenPediatric Pulmonology 53:1475–1484.https://doi.org/10.1002/ppul.24162
-
Nutrition and allergic diseases in urban and rural communities from the south african food allergy cohortPediatric Allergy and Immunology : Official Publication of the European Society of Pediatric Allergy and Immunology 30:511–521.https://doi.org/10.1111/pai.13058
-
Asthma and the westernization 'package'International Journal of Epidemiology 31:1098–1102.https://doi.org/10.1093/ije/31.6.1098
-
No association between tuberculin skin test and atopy in a Bacillus Calmette-Guérin vaccinated birth cohortPediatric Allergy and Immunology 20:545–550.https://doi.org/10.1111/j.1399-3038.2008.00846.x
-
Does BCG vaccination protect against the development of childhood asthma? A systematic review and meta-analysis of epidemiological studiesInternational Journal of Epidemiology 39:469–486.https://doi.org/10.1093/ije/dyp307
-
Outcome modelling strategies in epidemiology: traditional methods and basic alternativesInternational Journal of Epidemiology 45:565–575.https://doi.org/10.1093/ije/dyw040
-
Statistical foundations for model-based adjustmentsAnnual Review of Public Health 36:89–108.https://doi.org/10.1146/annurev-publhealth-031914-122559
-
The skin prick test - European standardsClinical and Translational Allergy 3:3.https://doi.org/10.1186/2045-7022-3-3
-
Exhaled nitric oxide: an updateImmunology and Allergy Clinics of North America 38:573–585.https://doi.org/10.1016/j.iac.2018.06.001
-
A simple device for quantitative stool thick-smear technique in schistosomiasis mansoniRevista Do Instituto De Medicina Tropical De Sao Paulo 14:397–400.
-
Asthma in young south Asian women living in the United Kingdom: the importance of early lifeClinical & Experimental Allergy 37:47–53.https://doi.org/10.1111/j.1365-2222.2006.02627.x
-
Asthma and current intestinal parasite infection: systematic review and meta-analysisAmerican Journal of Respiratory and Critical Care Medicine 174:514–523.https://doi.org/10.1164/rccm.200603-331OC
-
Mother's education and offspring asthma risk in 10 European cohort studiesEuropean Journal of Epidemiology 32:797–805.https://doi.org/10.1007/s10654-017-0309-0
-
Does living on a farm during childhood protect against asthma, allergic rhinitis, and atopy in adulthood?American Journal of Respiratory and Critical Care Medicine 164:1829–1834.https://doi.org/10.1164/ajrccm.164.10.2103137
-
Asthma endotypes: a new approach to classification of disease entities within the asthma syndromeJournal of Allergy and Clinical Immunology 127:355–360.https://doi.org/10.1016/j.jaci.2010.11.037
-
Sensitization to common aeroallergens in a population of young adults in a sub-Saharan africa setting: a cross-sectional studyAllergy, Asthma & Clinical Immunology 12:1.https://doi.org/10.1186/s13223-015-0107-8
-
Asthma and allergic disorders in Uganda: a Population-Based study across urban and rural settingsThe Journal of Allergy and Clinical Immunology: In Practice 6:1580–1587.https://doi.org/10.1016/j.jaip.2017.11.032
-
Skin prick test reactivity to common allergens among women in Entebbe, UgandaTransactions of the Royal Society of Tropical Medicine and Hygiene 102:367–373.https://doi.org/10.1016/j.trstmh.2008.01.017
-
Parasites and allergy: Observations from AfricaParasite Immunology 41:e12589.https://doi.org/10.1111/pim.12589
-
Prevalence and determinants of allergic diseases in a congolese populationInternational Forum of Allergy & Rhinology 2:285–293.https://doi.org/10.1002/alr.21017
-
Mycobacterium tuberculosis infection may protect against allergy in a tuberculosis endemic areaClinical Experimental Allergy 36:70–76.https://doi.org/10.1111/j.1365-2222.2005.02408.x
-
Rural to urban migration is associated with increased prevalence of childhood wheeze in a Latin-American cityBMJ Open Respiratory Research 4:e000205.https://doi.org/10.1136/bmjresp-2017-000205
-
Prevalence and risk factors for wheezing in children from rural Areas of north IndiaAllergy and Asthma Proceedings 28:647–653.https://doi.org/10.2500/aap.2007.28.3059
-
Biomass fuel use and the risk of asthma in nigerian childrenRespiratory Medicine 107:1845–1851.https://doi.org/10.1016/j.rmed.2013.09.009
-
The Urban-Rural gradient in asthma: a Population-Based study in northern europeInternational Journal of Environmental Research and Public Health 13:93.https://doi.org/10.3390/ijerph13010093
-
The impact of asthma and COPD in sub-Saharan AfricaPrimary Care Respiratory Journal 20:240–248.https://doi.org/10.4104/pcrj.2011.00027
-
Proximity of the home to roads and the risk of wheeze in an ethiopian populationOccupational and Environmental Medicine 62:376–380.https://doi.org/10.1136/oem.2004.017228
-
The changing prevalence of asthma, allergic rhinitis and atopic eczema in African adolescents from 1995 to 2002Pediatric Allergy and Immunology 18:560–565.https://doi.org/10.1111/j.1399-3038.2007.00554.x
Decision letter
-
Eduardo FrancoSenior Editor; McGill University, Canada
-
Belinda NicolauReviewing Editor; McGill University, Canada
-
Akhil SomanReviewer
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
Acceptance summary:
There are only a few studies on asthma from African countries and this is particularly true of Uganda. The authors conducted a study to identify the main risk factors for asthma among schoolchildren living in urban Uganda. The authors show that the places of birth and residence in the early years of life contribute to the risk of asthma in later years. The findings raise important questions on early environmental and lifestyle determinants of asthma, specifically among children living in poverty and in low- and middle-income countries. This work is therefore of interest to the researchers working on asthma and early determinants of health.
Decision letter after peer review:
Thank you for submitting your article "Risk factors for asthma among schoolchildren who participated in a case-control study in urban Uganda" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Eduardo Franco as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Akhil Soman (Reviewer #3).
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
Summary:
There are only a few studies on asthma from African countries and this is particularly true of Uganda. Therefore, this paper will add to the sparse literature. Using a case control design, the authors have shown that the places of birth and residence in the early years of life contribute to the risk of asthma in later years. The paper is generally well written. However, there are a few methodological and analytic issues that limit the contribution of the paper to the literature. The following review lists several items in need of revision.
Essential revisions:
Introduction:
1) With respect to scarcity of evidence on asthma risk factors from Africa, the authors attempt to address an important gap in the literature. However, the current Introduction does not provide a strong rationale for the analysis undertaken. It would be worthwhile to more comprehensively summarise the relevant literature by elaborating on the African context of this study through greater description of existing knowledge and explicit comparisons with studies from other African countries. The paper would also benefit from comparisons with studies from other low and middle income countries.
Materials and methods:
1) It would help to logically sequence the Materials and methods section. The type of case-control study, the secondary study base, time period of study and case and control size need to be highlighted at the start of the Materials and methods section. A helpful participant flow diagram is presented in Figure 1. However, in the current text, sample size is introduced in results and sampling of controls comes in data management and statistical analysis.
2) Two sources of interviews were used: in person and telephone interviews; does this mean that only children whose parents had a telephone were included in the study? Also, using two methods of interview can lead to differential responses. What was the distribution of this differential interview pattern? Can the authors comment on the potential of information bias arising from this pattern of interview? There are also some concerns related to pre-screening procedures: how confident are the authors that the children reported breathing problems accurately? The age range (5-17) is wide; this increases the possibility of obtaining erroneous information.
3) The several examinations mentioned under clinical assessment may be important for the topic addressed. Yet the rationale or relevance of conducting these examinations (e.g. BCG scar, exhaled nitric oxide, skin prick tests, tests for intestinal helminths, TST) is not clearly stated. In addition, a definition and diagnostic criteria for asthma should be provided. Another point related to data collection was not mentioned: was there information on smoking habits of the children/adolescents? This is especially important for secondary school students.
4) Only 28.7% of the original sample agreed to participate in the study; could you please elaborate on the potential impact of this low response rate on the results of study? Also, what was the difference in the response rate between cases and controls?
Analytic strategy:
The statistical analysis section should more coherently present confounders and mediators identified among the measured variables, statistical models used, and measures of strength and precision used to describe results. Below are a few specific comments that should either be clarified or considered through the conduct of further analysis:
1) One of the major concerns is that the sample in the study is drawn from 50+ schools and thus prone to clustering effects. There is no indication in the data analysis of whether this issue was considered. We strongly recommend that the authors address and justify what they have done. If they cannot, please redo the analyses.
2) The authors state that they have assessed combined effects of risk factors on additive and multiplicative scales. They have also interpreted the results given in tables under the combined effect Results section as additive or multiplicative. However, it can be argued that the authors do not present sufficient results to make such interpretations. What were the measures and associated confidence intervals based on and which combined effect on additive or multiplicative scales were assessed? P-values are not enough to estimate strength and precision of interaction results. If the authors have not used any such measures, it is recommended to just present stratum specific effects of these exposures and refrain from making any claims about interaction. Please highlight these in the limitations.
3) While we appreciate that the causal pathway factors were not adjusted in the models, there is no explanation of what those factors were. Was there any technique used (e.g., causal graphs) to identify them?
4) The paper would benefit from an explanation for the choice of socioeconomic indicator used in the final models (Why father's education? Were the continuous variables used as linear functions in the models? Was the potential for non-linearity tested? Several exposures were tested against a single outcome. Did the authors correct for multiple comparisons? If not, why?
5) The 95% CIs were estimated in the analysis. However, this is not stated in the analysis section. Also, please clarify the relevance of p-values (again not stated in the Materials and methods section) in the results tables. There is a strong argument in the literature that estimates and CIs, not p-values, convey the strength and precision of results (please refer to Nature comment: "Scientists rise up against statistical significance", 20 March 2019)
6) Table 1 shows missingness (ranging from N=0 to N=420) related to several variables. How did the authors deal with missing values? Were imputation and sensitivity analysis considered? Any comments related to potential bias?
[Editors' note: further revisions were requested prior to acceptance, as described below.]
Thank you for resubmitting your work entitled "Risk factors for asthma among schoolchildren who participated in a case-control study in urban Uganda" for further consideration by eLife. Your revised article has been evaluated by Eduardo Franco (Senior Editor) and Belinda Nicolau (Reviewing Editor).
The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:
1) In subsection “Current features of asthma cases versus controls”, the authors say that there was weak evidence for an inverse association between asthma and infection with any helminth's species. Based on the odds ratio and confidence interval presented, we suggest the authors to further elaborate this conclusion.
2) Please check the order of sections in the manuscript, table numbers, titles etc.
3) Please check for duplicates in reference, e.g., 53 and 56.
https://doi.org/10.7554/eLife.49496.sa1Author response
Essential revisions:
Introduction:
1) With respect to scarcity of evidence on asthma risk factors from Africa, the authors attempt to address an important gap in the literature. However, the current Introduction does not provide a strong rationale for the analysis undertaken. It would be worthwhile to more comprehensively summarise the relevant literature by elaborating on the African context of this study through greater description of existing knowledge and explicit comparisons with studies from other African countries. The paper would also benefit from comparisons with studies from other low and middle income countries.
We thank the reviewer for pointing out ways we could improve the Introduction section. We have added more relevant literature from Africa and other LMICs, mindful not to make the introduction too long.
“There is little previously published data on asthma risk factors from Africa. Unlike in HICs, higher parental education and socioeconomic status has been associated with asthma among children in Africa (Addo-Yobo et al., 2007; Nantanda et al., 2013;, Wolff et al., 2012).”
Materials and methods:
1) It would help to logically sequence the Materials and methods section. The type of case-control study, the secondary study base, time period of study and case and control size need to be highlighted at the start of the Materials and methods section. A helpful participant flow diagram is presented in Figure 1. However, in the current text, sample size is introduced in results and sampling of controls comes in data management and statistical analysis.
In accordance to the reviewer’s suggestions, we have made changes to the start of the Materials and methods section, which now reads as shown below.
Study design
We conducted an un-matched case-control study among schoolchildren, and report following STROBE guidelines (von Elm et al., 2014).
Study population, sampling and consent
The study base was determined a priori as schoolchildren, 5-17 years old, in primary and secondary schools in Entebbe Municipality and Katabi zone in Wakiso District, Central Uganda. This was a predominantly urban (town) setting. All schools in the study area were invited and 96% participated. Study enrolment was between May 2015 and July 2017. We estimated that a sample size of 2,112 would have 80% power to detect odds ratio (OR) <0.5 or >1.5 for exposures with prevalence 10% among controls.
At each school, we pre-screened by registering all children with any breathing problems. We concurrently randomly selected from the class register two children without any breathing problems, using a random number generator programme in STATA (StataCorp, Texas, USA).”
“Following consent and assent, we screened all participants with the International Study of Asthma and Allergies in Childhood (ISAAC) questionnaire (Asher et al., 1995).”
2) Two sources of interviews were used: in person and telephone interviews; does this mean that only children whose parents had a telephone were included in the study? Also, using two methods of interview can lead to differential responses. What was the distribution of this differential interview pattern? Can the authors comment on the potential of information bias arising from this pattern of interview?
We agree that the issue of telephone interviews requires more clarification. We only telephoned a small proportion of children’s mothers (6.6%) and fathers (3.4%) to specifically answer the question regarding their history of allergy. This small proportion of telephone interviews is unlikely to have introduced any information bias to the study. We have clarified this point more in the Materials and methods section.
“When a parent was not available in person and the participant answering the questionnaire did not know that parent’s history of allergy, we telephoned the parent for the relevant information.”
Table 1 legend: “For parental history of allergy, 6.6% mothers and 3.4% fathers responded by telephone.”
There are also some concerns related to pre-screening procedures: how confident are the authors that the children reported breathing problems accurately?
It is true that the children may not have reported breathing problems accurately at the pre-screening stage. However, the purpose of the pre-screening stage was to identify all children with any breathing problems in our primary study base. Then, during the screening and enrolment stages of the study, proper case ascertainment was made by the clinical study team. We are confident of the accuracy of the final diagnosis of the asthma cases.
“If the diagnosis was not straightforward, two clinicians reviewed that participant and if they disagreed, the participant was excluded in order to ensure proper case ascertainment.”
The age range (5-17) is wide; this increases the possibility of obtaining erroneous information.
We agree with the reviewer that age 5-17 is a wide range. In order to minimise the possibility of obtaining erroneous information, we used interviewer-led questionnaires for both adolescents and parents (who answered on behalf of the younger children).
3) The several examinations mentioned under clinical assessment may be important for the topic addressed. Yet the rationale or relevance of conducting these examinations (e.g. BCG scar, exhaled nitric oxide, skin prick tests, tests for intestinal helminths, TST) is not clearly stated.
We agree with the reviewer and have now rewritten the clinical assessments section to include the rationale, with relevant references, for the various tests that we conducted.
“We looked for the presence of a Bacillus Calmette–Guérin (BCG) scar since BCG vaccination has been inversely associated with asthma previously (El-Zein et al., 2010).”
“ENO is considered a biomarker of allergic airway inflammation (Hoyte, Gross and Katial, 2018).”
“Skin prick tests (SPT) and allergen-specific IgE (asIgE) are important measures of allergic sensitisation (Heinzerling et al., 2013).”
“Because previous studies have reported an association between asthma and helminths (Mpairwe and Amoah, 2019), we collected three fresh stool samples from each participant and tested for intestinal helminths using the Kato Katz method (Katz, Chaves and Pellegrino, 1972). The findings on the association between tuberculin skin test (TST) and allergy/asthma have been inconsistent (Obihara et al., 2006; Eifan et al., 2009). We investigated this association by performing the TST using standard procedures, as we have previously described (Nkurunungi et al., 2012).”
In addition, a definition and diagnostic criteria for asthma should be provided.
Thank you for pointing out this important omission. We have now clarified the definition and diagnostic criteria for asthma.
“Following consent and assent, we screened all participants with the International Study of Asthma and Allergies in Childhood (ISAAC) questionnaire (Asher et al., 1995). Children with a history of wheezing in the last 12 months underwent a detailed medical history and examination by study clinicians, to diagnose asthma. Asthma was defined as a history of recurrent symptoms of wheezing, cough (mostly dry, worse at night and morning) and/or difficulty in breathing experienced in the last 12 months, with or without forced expiratory volume in the first second (FEV1) <80% expected for age, sex and height for African children. Additional medical history included a prior physician diagnosis and good response to asthma medication. If the diagnosis was not straightforward, two clinicians reviewed that participant and if they disagreed, the participant was excluded in order to ensure proper case ascertainment.”
Another point related to data collection was not mentioned: was there information on smoking habits of the children/adolescents? This is especially important for secondary school students.
It is true that smoking is important in asthma. We collected data on current smoking by any household member (including the adolescents themselves) and the overall prevalence was only 11% and did not differ between asthma cases and controls (Table 2, rows 10-11).
Generally, details of the items from the questionnaires are presented in the results tables. We did this to keep the Materials and methods sections short and simple. But we are willing to include details of questionnaire items in the Materials and method’s section as well, if that is preferred.
“This included the adolescents’ own smoking history.” added to Table 2 legend.
4) Only 28.7% of the original sample agreed to participate in the study; could you please elaborate on the potential impact of this low response rate on the results of study?
True, the response rate was low. This was mainly due to the difficulty in contacting parents of children in the boarding school section, if they did not have a telephone. We discussed this as an important study limitation, and stated that we do not have information on whether they were more likely to come from rural areas or the city, which areas are associated with a lower or higher asthma risk, respectively. We have included this last point in the discussion.
“Finally, a large number of potential participants were not included in this study, mostly because they were in the boarding school section and were unable to contact their parents/guardians in time to attend the parents’ meeting (to provide consent). We do not have information on whether these parents were more likely to reside in rural areas or the city, where the asthma risk is lower or higher than the town where this study was conducted, respectively.”
Also, what was the difference in the response rate between cases and controls?
This is an important point, but we do not have information on response rates of cases separate from controls because the majority of potential study participants were lost at the pre-screening stage, before we could ascertain whether they were true cases or controls.
Analytic strategy:
The statistical analysis section should more coherently present confounders and mediators identified among the measured variables, statistical models used, and measures of strength and precision used to describe results. Below are a few specific comments that should either be clarified or considered through the conduct of further analysis:
1) One of the major concerns is that the sample in the study is drawn from 50+ schools and thus prone to clustering effects. There is no indication in the data analysis of whether this issue was considered. We strongly recommend that the authors address and justify what they have done. If they cannot, please redo the analyses.
Thank you for this observation. We initially had not considered clustering in the analysis, since several similar previous asthma studies, particularly the ISAAC studies, have not found any significant clustering effects (Asher MI, Anderson HR, Stewart AW, Crane J, Ait-Khaled N, Anabwani G, Anderson HR, Beasley R, Björkstén B, Burr ML, Clayton TO, Crane J, Ellwood P, Keil U, Lai CKW, Mallol J, Martinez FD, Mitchell EA, Montefort S, Pearce N, Robertson CF, Shah JR, Sibbald B, Strachan DP, Weiland SK, Williams HC (ISAAC Steering Committee). Worldwide variations in the prevalence of asthma symptoms: International Study of Asthma and Allergies in Childhood (ISAAC). Eur Respir J 1998; 12: 315-335)
In accordance with the reviewers’ recommendation, we have repeated the analysis, using a random effects logistic regression analysis to allow for clustering by school. As with the ISAAC studies, allowing for clustering had minimal effects on the results, but we present these results for the benefit of some of your readers who prefer this type of analysis. We have changed the data management and statistical analysis section to reflect this change. We have also changed the adjusted odds ratios in all results tables and in the text of the Results section.
“We conducted initial logistic regression analyses for each exposure of interest, adjusted for age and sex, with random effects to allow for any clustering by school. We then ran the adjusted random effects logistic regression model (i.e. adjusted for the above factors in addition to age and sex) for each exposure of interest, and checked for collinearity by comparing the standard error of the exposure coefficient in the adjusted model and in the basic model(Greenland, Daniel and Pearce, 2016; Greenland and Pearce, 2015). We did not find any problems of collinearity, so we reported the findings for the adjusted model for each exposure.”
2) The authors state that they have assessed combined effects of risk factors on additive and multiplicative scales. They have also interpreted the results given in tables under the combined effect Results section as additive or multiplicative. However, it can be argued that the authors do not present sufficient results to make such interpretations. What were the measures and associated confidence intervals based on and which combined effect on additive or multiplicative scales were assessed? P-values are not enough to estimate strength and precision of interaction results. If the authors have not used any such measures, it is recommended to just present stratum specific effects of these exposures and refrain from making any claims about interaction. Please highlight these in the limitations.
We thank the reviewers for this assessment and we have followed their advice to present only the stratum specific effects of the exposures, and have refrained from making any claims about interaction. Because we have no longer included interactions, we have not needed to include any discussion of interactions in the limitations section.
We have removed p-values from Tables of results 4-7, and from Supplementary files 2 and 3.
3) While we appreciate that the causal pathway factors were not adjusted in the models, there is no explanation of what those factors were. Was there any technique used (e.g., causal graphs) to identify them?
Thank you for raising this important issue. We did not identify any mediators on the causal pathway. We have changed the text in the statistical analysis section to read as seen below.
“We did not identify any factors which were likely to be on the causal pathways (and therefore were not potential confounders and should not be adjusted for).”
4) The paper would benefit from an explanation for the choice of socioeconomic indicator used in the final models (Why father's education? Were the continuous variables used as linear functions in the models? Was the potential for non-linearity tested?
We collected data on both father’s and mother’s highest education level attained, as a proxy indicator for socioeconomic status. This was based on our experience conducting research in this setting, and from findings from a large study conducted in Dar es Salaam, Tanzania and Cape Town, South Africa, which found that father’s and mother’s education were significantly associated with socioeconomic status sum score (Aarø LE, Flisher AJ, Kaaya S, Onya H, Namisi FS, Wubs A. Parental education as an indicator of socioeconomic status: improving quality of data by requiring consistency across measurement occasions. Scand J Public Health. 2009 Jun;37 Suppl 2:16-27. doi: 10.1177/1403494808086917).
But since father’s and mother’s education were strongly associated, we chose to include only father’s education to avoid collinearity. Fathers are the breadwinners in most households in Uganda. We have included this point in the Materials and methods section.
“Previous studies in this setting have found that father’s and mother’s education are significantly associated with socioeconomic status.”
Several exposures were tested against a single outcome. Did the authors correct for multiple comparisons? If not, why?
We did not correct for multiple comparisons, because all of the tested exposures were selected because there was good prior evidence that they were likely to be risk factors for asthma. Thus, we assessed the findings for each exposure individually, in light of prior knowledge and evidence.
5) The 95% CIs were estimated in the analysis. However, this is not stated in the analysis section.
We thank the reviewer for reminding us to state explicitly in the statistical analysis section that we estimated 95% CIs. We have done this and now the statement reads as below.
“The odds ratio was the main outcome measure (as is appropriate for a case-control study), and we also estimated the 95% CIs. We identified age and sex as a priori confounders, and all analyses were adjusted for these variables. We identified area of residence at time of birth and father’s education as potential confounders; we did not also adjust for area where the child spent most of the first five years or mother’s education, since these were strongly associated with the above two factors, and therefore would have introduced collinearity.”
Also, please clarify the relevance of p-values (again not stated in the Materials and methods section) in the results tables. There is a strong argument in the literature that estimates and CIs, not p-values, convey the strength and precision of results (please refer to Nature comment: "Scientists rise up against statistical significance", 20 March 2019)
In line with the reviewer’s recommendation, we have now excluded p-values from all results tables. However, we note that your decision letter states that “please report exact p-values wherever possible alongside the summary statistics and 95% confidence intervals”, which is what we had done originally. We are happy to follow your advice whether to include p-values or not.
All results tables now do not include p-values.
6) Table 1 shows missingness (ranging from N=0 to N=420) related to several variables. How did the authors deal with missing values? Were imputation and sensitivity analysis considered? Any comments related to potential bias?
We thank the reviewer for raising this important issue and agree that this needs further clarification. We have repeated the analysis using the ‘complete case’ data set, by dropping 30 participants from the analysis who had missing values in the four key variables. This has not changed the study results. We have reflected this change in the number of participants in the flow diagram, results tables and in the re-written statistical section, as shown below.
“The key variables adjusted for (see below) were age, sex, area of residence at time of birth and father’s education. Each of these variables (except sex) had small numbers of missing values (the largest number was 29 for father’s education). We therefore created a ‘complete case’ data set, comprising the 555 cases and 1,115 controls which had no missing values for these variables. This ‘complete case’ data set was used for all analyses.”
[Editors' note: further revisions were requested prior to acceptance, as described below.]
The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:
1) In subsection “Current features of asthma cases versus controls”, the authors say that there was weak evidence for an inverse association between asthma and infection with any helminth's species. Based on the odds ratio and confidence interval presented, we suggest the authors to further elaborate this conclusion.
Thank you for this observation. We agree this point needed further clarification. We have made the following changes.
“There was weak evidence for an inverse association between asthma and infection with any helminths species [0.75 (0.53-1.07)] (Table 2); overall this was not statistically significant as shown by the fact that the confidence interval included the null value of 1, but it was statistically significant for T. trichiura [0.33 (0.13-0.89)] (Supplementary file 1a).”
2) Please check the order of sections in the manuscript, table numbers, titles etc.
Thank you for this comment. We structured the sections of our article as follows: Abstract, Introduction, Results, Discussion, Materials and methods, References, Tables of results and Figure legends. We are willing to change this order to any other that you advise us to. We noticed that one Table was in the wrong place. We have now moved that Table from the Materials and methods section to the ‘Tables section’ and have now labelled it as “Table 8”, and have given it a title. In additional, we have changed the names for the supplementary tables to supplementary files.
3) Please check for duplicates in reference, e.g., 53 and 56.
Thank you for that observation. We have deleted reference 56 and checked that there are no more duplicate references.
https://doi.org/10.7554/eLife.49496.sa2Article and author information
Author details
Funding
Wellcome (Training fellowship 102512)
- Harriet Mpairwe
Wellcome (Senior fellowship 095778)
- Alison M Elliott
European Research Council (Project grant 668954)
- Neil Pearce
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
We thank the study participants, their parents and guardians for their enthusiastic participation. We thank the teachers and school administrations for providing us with a conducive environment in which to conduct this study. We acknowledge the support we received from the Entebbe Municipality and Wakiso District Education Officials. Many thanks to Kisubi Hospital staff, particularly Dr. Robert Asaba and Dr. Rogers Sendijja, for their support during the results dissemination meetings we held in their premises. Many thanks to Ronald van Ree from Leiden University Medical Centre (The Netherlands) for working with GN on the asIgE ImmunoCAP assays.
Ethics
Human subjects: Parents or guardians of the children provided written informed consent, and children eight years or older provided written informed assent. This consent was to participate in the study, and to publish anonymous results. The study was approved by the Uganda Virus Research Institute Research and Ethics Committee, and the Uganda National Council for Science and Technology [reference number HS 1707]. The two bodies follow Good Clinical Practice guidelines.
Senior Editor
- Eduardo Franco, McGill University, Canada
Reviewing Editor
- Belinda Nicolau, McGill University, Canada
Reviewer
- Akhil Soman
Version history
- Received: June 19, 2019
- Accepted: November 13, 2019
- Accepted Manuscript published: November 15, 2019 (version 1)
- Version of Record published: December 16, 2019 (version 2)
Copyright
© 2019, Mpairwe et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 1,356
- Page views
-
- 191
- Downloads
-
- 15
- Citations
Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Epidemiology and Global Health
Background:
Machine learning (ML) techniques improve disease prediction by identifying the most relevant features in multidimensional data. We compared the accuracy of ML algorithms for predicting incident diabetic kidney disease (DKD).
Methods:
We utilized longitudinal data from 1365 Chinese, Malay, and Indian participants aged 40–80 y with diabetes but free of DKD who participated in the baseline and 6-year follow-up visit of the Singapore Epidemiology of Eye Diseases Study (2004–2017). Incident DKD (11.9%) was defined as an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m2 with at least 25% decrease in eGFR at follow-up from baseline. A total of 339 features, including participant characteristics, retinal imaging, and genetic and blood metabolites, were used as predictors. Performances of several ML models were compared to each other and to logistic regression (LR) model based on established features of DKD (age, sex, ethnicity, duration of diabetes, systolic blood pressure, HbA1c, and body mass index) using area under the receiver operating characteristic curve (AUC).
Results:
ML model Elastic Net (EN) had the best AUC (95% CI) of 0.851 (0.847–0.856), which was 7.0% relatively higher than by LR 0.795 (0.790–0.801). Sensitivity and specificity of EN were 88.2 and 65.9% vs. 73.0 and 72.8% by LR. The top 15 predictors included age, ethnicity, antidiabetic medication, hypertension, diabetic retinopathy, systolic blood pressure, HbA1c, eGFR, and metabolites related to lipids, lipoproteins, fatty acids, and ketone bodies.
Conclusions:
Our results showed that ML, together with feature selection, improves prediction accuracy of DKD risk in an asymptomatic stable population and identifies novel risk factors, including metabolites.
Funding:
This study was supported by the National Medical Research Council, NMRC/OFLCG/001/2017 and NMRC/HCSAINV/MOH-001019-00. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
-
- Epidemiology and Global Health
The NIH-funded RECOVER study is collecting clinical data on patients who experience a SARS-CoV-2 infection. As patient representatives of the RECOVER Initiative’s Mechanistic Pathways task force, we offer our perspectives on patient motivations for partnering with researchers to obtain results from mechanistic studies. We emphasize the challenges of balancing urgency with scientific rigor. We recognize the importance of such partnerships in addressing post-acute sequelae of SARS-CoV-2 infection (PASC), which includes ‘long COVID,’ through contrasting objective and subjective narratives. Long COVID’s prevalence served as a call to action for patients like us to become actively involved in efforts to understand our condition. Patient-centered and patient-partnered research informs the balance between urgency and robust mechanistic research. Results from collaborating on protocol design, diverse patient inclusion, and awareness of community concerns establish a new precedent in biomedical research study design. With a public health matter as pressing as the long-term complications that can emerge after SARS-CoV-2 infection, considerate and equitable stakeholder involvement is essential to guiding seminal research. Discussions in the RECOVER Mechanistic Pathways task force gave rise to this commentary as well as other review articles on the current scientific understanding of PASC mechanisms.