- Email: [email protected]
Rural and urban food allergy prevalence from the South African Food Allergy (SAFFA) study
Rural and urban food allergy prevalence from the South African Food Allergy (SAFFA) study Maresa Botha, MD,a Wisdom Basera, MPH,b Heidi E. Facey-Thomas, Dip. Nursing, RN,a Ben Gaunt, MD,c,d Claudia L. Gray, MD, PhD,a Jordache Ramjith, MSc,e,f Alexandra Watkins, MA, MSc,a and Michael E. Levin, MD, PhDa,g Cape Town and Zithulele, South Africa, and Nijmegen, The Netherlands Background: Food sensitization and challenge-proved food allergy (FA) have not been compared in urban and rural settings. Objective: We sought to determine and compare the prevalence of food sensitization and challenge-proved IgE-mediated FA in urban and rural South African toddlers aged 12 to 36 months. Methods: This cross-sectional study of unselected children included 1185 participants in urban Cape Town and 398 in the rural Eastern Cape. All participants completed a questionnaire and underwent skin prick tests (SPTs) to egg, peanut, cow’s milk, fish, soya, wheat, and hazelnut. Participants with SPT responses of 1 mm or greater to 1 or more foods and not tolerant on history underwent an open oral food challenge. Result: The prevalence of FA was 2.5% (95% CI, 1.6% to 3.3%) in urban children, most commonly to raw egg white (1.9%), followed by cooked egg (0.8%), peanut (0.8%), cow’s milk (0.1%), and fish (0.1%). Urban sensitization (SPT response _1 mm) to any food was 11.4% (95% CI, 9.6% to 13.3%) and > 9.0% (95% CI, 7.5% to 10.8%) at an SPT response of 3 mm or greater. Sensitization in rural cohorts was significantly lower than in the urban cohort (1-mm SPT response, 4.5% [95% CI, 2.5% to 6.6%]; 3-mm SPT response, 2.8% [95% CI, 1.4% to 4.9%]; P < .01). In the rural black African cohort 0.5% (95% CI, 0.1% to 1.8%) of children had food allergy, all to egg. This is significantly lower than the prevalence of the urban cohort overall (2.5%) and urban black African participants (2.9%; 95% CI, 1.5% to 4.3%; P 5 .006). Conclusion: FA prevalence in Cape Town is comparable with rates in industrialized middle-income countries and is significantly greater than in rural areas. Further analysis will describe and compare environmental exposures and other risk factors in this cohort. (J Allergy Clin Immunol 2018;nnn:nnn-nnn.) From athe Division of Paediatric Allergy, Department of Paediatrics, University of Cape Town, Cape Town; bSchool of Public Health and Family Medicine, Faculty of Health Sciences, University of Cape Town, Cape Town; cZithulele Hospital, Eastern Cape Department of Health, Zithulele; dthe Division of Primary Health Care, Health Sciences Faculty, University of Cape Town, Cape Town; ethe Division of Epidemiology & Biostatistics, School of Public Health & Family Medicine, University of Cape Town, Cape Town; fthe Department for Health Evidence, Biostatistics Research Group, Radboud Institute for Health Sciences, Radboud University Medical Center, Nijmegen, The Netherlands; and ginVIVO Planetary Health, Group of the Worldwide Universities Network (WUN). Supported by the Medical Research Council of South Africa, National Research Foundation of South Africa, Mylan, Thermo Fisher, Nestle, Cipla, Aspen, Pharma Dynamics, Astellas, Beiersdorf, Novartis, Nutricia, MSD, Takeda, and AstraZeneca. Disclosure of potential conflict of interest: The authors declare that they have no relevant conflicts of interest. Received for publication March 19, 2018; revised June 12, 2018; accepted for publication July 12, 2018. Corresponding author: Michael E. Levin, MD, PhD, Rm 516, ICH Building, Red Cross Hospital, Klipfontein Rd, Rondebosch 7700, South Africa. E-mail: michael.levin@ uct.ac.za. 0091-6749/$36.00 Ó 2018 American Academy of Allergy, Asthma & Immunology https://doi.org/10.1016/j.jaci.2018.07.023
Key words: Africa, allergy, egg, ethnicity, food allergy, food sensitization, peanut, prevalence, skin prick test, urban, rural
Food allergy (FA) is a growing public health concern in both middle- and high-income countries.1 There is as yet no cure, and moreover, the quality of life of children with FA and their families is significantly affected by the economic, nutritional, and social effect of having an FA.2 Worldwide, the prevalence of FA in children ranges from 2% to 10%, depending on the age of the children studied, range of foods tested, methodologies used, and geographic area.1,3,4 In South Africa, as in other transitional countries, the prevalence of noncommunicable diseases in children is a growing public health concern. Studies in South Africa have clearly shown an increase in asthma, allergic rhinitis, and atopic dermatitis over time.5 In addition, rural communities experience a lower prevalence of respiratory allergy, but the ‘‘protection’’ in rural areas appears to be decreasing over time, with a narrowing of the urban versus rural gradient.6 FA has been investigated in an urban tertiary care setting among children in Cape Town with moderate-to-severe atopic dermatitis, showing high rates of food sensitization (FS) and FA comparable with those found in similar highly selected populations from other countries.7 The same study also showed lower rates of FA, particularly peanut allergy, in black African (BA) children compared with children of mixed ancestry (MA). This raises the question of whether there are protective factors reducing the development of allergies in BA children that might have either a genetic or environmental basis. Data from studies elsewhere in the world have suggested differences in sensitization and allergy rates according to socioeconomic class and between children of different ethnicities and between children of newly urbanized immigrants and local populations.4,7 No previous large-scale epidemiologic data on FA and FS exist in South Africa, and no prior data on FA prevalence in an unselected population have been collected. The hypothesis for this study was that allergy and sensitization rates would be lower in urban BA participants than urban participants of white origin or MA and even lower in rural BA participants.
METHODS Study design and setting In this cross-sectional study 1207 children (age, 12-36 months) who attended registered early child development (ECD) centers in Cape Town were recruited between February 2013 and December 2016. ECD centers are government-regulated daycare centers for preschool children that are partially 1
2 BOTHA ET AL
Abbreviations used BA: Black African ECD: Early child development FA: Food allergy FS: Food sensitization MA: Mixed ancestry OFC: Oral food challenge SPT: Skin prick test
funded through subsidies and serve most children in South Africa. Four hundred fourteen participants in the same age group were recruited in the rural Mqanduli district of the Eastern Cape by raising awareness through local primary health care clinics and community stakeholders. This age range was chosen to cover the range at which we postulated that FA would be prevalent. The majority of BA subjects living in Cape Town are AmaXhosa. They are the second largest ethnolinguistic group in South Africa and reside mainly in the Eastern Cape province. Migration to Cape Town from the rural Eastern Cape is ongoing but has been particularly rapid in the last 30 years. The Mqanduli district was chosen because of the high rate of migration from this area to Cape Town and the availability of an excellent district hospital to use as a site for food challenges.
J ALLERGY CLIN IMMUNOL nnn 2018
the back when eczema precluded use of the forearm. An SPT response of 1 mm or greater than that elicited by the negative control was considered ‘‘reactive’’ and sufficiently high to merit a food challenge because previous studies have found that young children with sensitization manifesting in am SPT response as small as 1 mm can have challenge-proved FA.8 _1 mm) who were not Participants with any reactive SPTs (> (on questionnaire) tolerating a full age-appropriate portion of the specific food thus qualified for an open oral food challenge (OFC). Participants with a reactive SPT who were tolerating a regular age-appropriate portion with no history of a reaction were considered food sensitized but not allergic. OFCs were performed at Red Cross Children’s Hospital in Cape Town and Zithulele Hospital in the Eastern Cape. Challenges were performed as open incremental oral challenges by using a standardized protocol with predetermined objective criteria for a positive challenge response.8 Challenges used full-fat cow’s milk, soy milk formula, peanut butter, steamed codfish, crushed hazelnuts, Weet-bix (Pioneer Foods, Cape Town, South Africa), and, for egg sensitization, both raw egg white and whole scrambled egg on separate occasions. OFCs in which participants did not reach the top dose but had no reaction were classified as equivocal if the challenge could not successfully be repeated. This study was approved by the Human Research Ethics Committee of the University of Cape Town (HREC REF: 038/2012). Informed consent was obtained from parents or legal guardians of participants.
Sampling frame
Data entry and statistical analysis
In Cape Town we accessed the publicly available online database of registered ECD centers and approached the ECD centers in a randomized manner to participate in our study. Once an ECD center agreed to participate, all children who attended and were aged 12 to 36 completed months were invited to participate. ECD centers were enrolled until a target of more than 1200 participants was reached. There are almost no ECD centers in the Mqanduli district. Instead all eligible children from the areas surrounding the 10 district community health clinics were recruited. Participants who responded to the study invitation but decided not to participate or who could not complete the study were considered nonparticipants.
Data were entered into a Microsoft Access database and analyzed by using Stata software (version 11.1; StataCorp, College Station, Tex). The Mann-Whitney test was used to investigate differences between numeric variables, all of which were nonnormally distributed. The Fisher exact test was used to assess overall associations between pairs of categorical variables, and the z test using normal approximation to the binomial distribution was used to investigate differences in proportions. A P value of less than .05 was considered statistically significant. Multiple logistic regression was performed with an OFC as the outcome and each SPT for each allergen used as the predictor variables by using this data set to calculate adjusted prevalence rates in the urban cohort. The probability of a positive OFC response was then calculated and used for each participant who did not complete an indicated OFC to add them to the analysis of adjusted prevalence.
Sample size Our sample size calculations were based on estimated prevalences of 0%, 2%, and 8% of FA in rural BA, urban BA, and urban MA children, respectively. By using these estimates, sample sizes of 400 for each cohort would give us 80% power to detect a 5% difference in prevalence between the 2 urban cohorts and also detect a difference between the 2 BA cohorts of 2% or greater _2%). Pilot data showed that our sampling (ie, rural, almost 0%; urban, > method produced a representative cohort according to the Cape Town Census, meaning that about 40% to 50% of the urban cohort would be BA subjects.8 These proportions were used to expand the urban cohort proportionally to 1200, producing enough power to match a rural cohort of 400. Power calculations were done based on the z test for differences in proportions.
Recruitment and assessment Questionnaires were administered to all participating parents regarding other allergic diseases or a family history of allergic diseases. Children were examined for signs of atopic dermatitis (visible skin lesions) and allergic rhinitis (allergic shiners, nasal crease, Dennie lines, and enlarged turbinates). Skin prick test (SPTs) were performed with ALK-Abello SPT solutions (Thermo Fisher, Waltham, Mass) to 7 common food allergens (egg white extract, peanut, cow’s milk, codfish, soya, wheat [flour], and hazelnut). In addition, fresh egg white, peanut butter, and fresh cow’s milk, as well as positive (10 mg/mL histamine) and negative (saline) controls, were used. SPT responses were read at 15 minutes and recorded as average wheal diameter in millimeters. Where fresh extracts and commercial extracts were used (egg, milk, and peanut), the highest of the 2 SPT responses for the same allergen source were analyzed for the purposes of this study. SPTs were performed with ALK-Abello lancets on the volar aspect of the forearm or on
RESULTS One thousand one hundred eighty-five urban and 398 rural participants completed the study (see Table E1 in this article’s Online Repository at www.jacionline.org). There were no statistically significant differences in median age, sex, family history of allergy, or history of allergic diseases (eczema, asthma, and rhinitis) or food reactions between study participants and nonparticipants (see Table E2 in this article’s Online Repository at www.jacionline.org). The median age of the urban cohort was 26 months (interquartile range, 22-32 months), which was older than the rural cohort at 21 months (interquartile range, 17-28 months; P <.001; Table I). There were significantly more male than female participants in both cohorts (urban cohort, 52.5% male and 47.5% female [P 5 .01]; rural cohort, 56.5% male and 43.5% female [P 5 .01]), but these ratios were not significantly different between the urban and rural cohorts (P 5 .162). The urban cohort reflected the ethnic demographics of children less than 5 years old in the Cape Town metropole, as captured by the 2011 census, with 46.3% of participants being BA and 46.5% and 7.2% being of MA or white origin, respectively
BOTHA ET AL 3
J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn
TABLE I. Description of the study population by study site Urban (n 5 1185)
Rural (n 5 398)
P value*
26 (22-32)
21 (17-28)
<.001
622 (52.5%)
225 (56.5%)
M/F ratio: .16
563 (47.5%)
173 (43.7%)
Enrollment age (mo) Median (IQR) Sex, no. (%) Male Female
Self-reported family history of allergic disease, no. (%) First-degree relative with any allergy First-degree relative with asthma First-degree relative with allergic rhinitis First-degree relative with atopic dermatitis First-degree relative with FA Self-reported comorbid allergies, no. (%) Any Asthma Allergic rhinitis Atopic dermatitis Allergic disease confirmed on clinical examination, no. (%) Allergic rhinitis Atopic dermatitis
578 207 421 195 59
(48.8%) (17.5%) (35.5%) (16.5%) (5.0%)
30 12 6 2 10
(7.5%) (3.0%) (1.5%) (0.5%) (2.5%)
<.001 <.001 <.001 <.001 <.001
502 106 279 279
(42.4%) (9.0%) (23.5%) (23.5%)
22 4 13 7
(5.5%) (1.0%) (3.3%) (1.8%)
<.001 .002 <.001 <.001
0 (0.0%) 3 (0.8%)
<.001 <.001
101 (8.5%) 139 (11.7%)
IQR, Interquartile range. *P values are from Fisher exact tests for categorical values and 2-sample Wilcoxon rank-sum (Mann-Whitney) tests for continuous variables.
TABLE II. Spectrum of sensitization and FA in the urban cohort _1 mm, SPT response > no. (% [95% CI])
Any food Egg (any) Peanut Cow’s milk Hazelnut Soya Wheat Fish
137 90 54 38 13 17 13 9
(11.6 [9.7- 13.4]) (7.6 [6.1-9.1]) (4.6 [3.4-5.8]) (3.2 [2.2-4.2]) (1.1 [0.5-1.7]) (1.4 [0.8-2.1]) (1.1 [0.5-1.7]) (0.8 [0.3-1.3])
_3 mm, SPT response > no. (% [95% CI])
107 78 40 16 5 5 1 2
(9.0 (6.6 (3.4 (1.4 (0.4 (0.4 (0.1 (0.2
[7.4-10.7]) [5.0-7.8]) [2.4-4.6]) [0.7-2.0]) [0.1-0.8]) [0.1-0.8]) [0.1-0.3]) [0.1-0.4])
(Census 2011: 46.4% BA, 45.4% MA, and 8.2% white). The rural cohort was entirely BA. Urban participants were significantly more likely than rural participants to have a family history of allergic diseases (48.8% [578/1185] vs 7.5% [30/398], P < .001) and reported a much higher prevalence of comorbid eczema and allergic rhinitis than their rural counterparts (23.5% [279] vs 3.3% [130] and 1.8% [7], P < .001; Table I). One hundred thirty-nine (11.7%) urban participants had objective signs of atopic dermatitis (AD), and 101 (8.5%) had signs of allergic rhinitis (AR). In the rural cohort AD and AR were significantly less frequent (0.8% [3] and 0.0%, both P <.001) than in the urban sample (Table I). Seventy-five (6.3%) urban participants self-reported a previous reaction to one of the 7 foods (cow’s milk, 2.5%; egg, 1.9%; fish, 1.8%; peanuts, 1.0%; soya, 0.3%; and wheat, 0.1%) compared with 13 (3.3%) of 398 (P 5 .02) of rural participants (cow’s milk, 1.5%; fish, 1.0%; soya, 0.5%; and wheat, 0.3%). One hundred thirty-seven (11.6%) of the urban cohort were _1-mm response). The highest sensitized to food at any level (>
OFC-confirmed FA, no. (% [95% CI])
27 (2.3 [1.5-3.3]) Raw egg white: 21 (1.8 [1.1-2.7]) Cooked whole egg: 9 (0.8 [0.3-1.4]) 8 (0.7 [0.3-1.3]) 1 (0.1 [0.0-0.5]) 0 0 0 1 (0.1 [0.0-0.5])
prevalence of sensitization was to egg, followed by peanut, cow’s milk, hazelnut, soya, wheat, and fish (Table II). Fifty-one (4.3%) of the urban participants were polysensitized, with 27 (2.3%) sensitized to 2 foods and 24 (2.0%) sensitized to 3 foods or more. Per-protocol analysis of the prevalence of FA was 2.3% (95% CI, 1.5% to 3.7%) in urban children. The most common allergy was to raw egg (1.8%), followed by cooked egg (0.8%), peanut (0.7%) cow’s milk (0.1%), and fish (0.1%; Table II). Male participants had a higher rate of FA than female participants (3.2% vs 1.2%, P 5 .02). Food challenges were indicated in 34 participants based on an SPT response of 1 mm or greater without a history of tolerating a full age-appropriate portion of the food. No participants were classified as having FA on the basis of a recent severe reaction. In 7 (20.6%) of these participants, 1 or more challenges were not completed because participants could not be contacted or did not attend. Individual uncompleted food challenges were to milk (2/3 indicated), cooked egg (6/15 indicated), peanut (2/10 indicated), and raw egg (4/25 indicated). Per-protocol analysis represents only the proved FA cases and therefore likely underestimates the true prevalence of
4 BOTHA ET AL
J ALLERGY CLIN IMMUNOL nnn 2018
TABLE III. Adjusted FA prevalence in the urban cohort Per-protocol analysis (minimum FA prevalence)
Any food, no. (% [95% CI]) Raw egg white, no. (% [95% CI]) Cooked whole egg, no. (% [95% CI]) Peanut, no. (% [95% CI]) Cow’s milk, no. (% [95% CI]) Fish, no. (% [95% CI])
27 21 9 8 1 1
(2.3 (1.8 (0.8 (0.7 (0.1 (0.1
Completed plus missed OFCs
[1.5-3.7]) [1.1-2.7]) [0.3-1.4]) [0.3-1.3]) [0.0-0.5]) [0.0-0.5])
27 21 9 8 1 1
1 1 1 1 1 1
Maximum FA prevalence
7 4 6 2 2 0
34 25 15 10 3 1
(2.9 (2.1 (1.3 (0.8 (0.3 (0.1
[2.0-4.0]) [1.4-3.1]) [0.7-2.1]) [0.4-1.6]) [0.1-0.7]) [0.0-0.5])
Adjusted FA prevalence
2.5 1.9 0.8 0.8
(1.6-3.3) (1.1-2.7) (0.3-2.3) (0.3-1.3) NA NA
NA, Not applicable.
TABLE IV. Urban sensitization and FA by ethnicity _1 mm > BA, no. (%)
Any food Egg Peanut Cow’s milk Soya Wheat Fish Hazelnut
71 51 28 16 7 5 2 5
(12.9) (9.3) (5.1) (2.9) (1.3) (0.9) (0.4) (0.9)
MA, no. (%)
61 34 25 22 10 8 7 8
(10.7) (6.2) (4.6) (4.0) (1.8) (1.5) (1.3) (1.5)
_3 mm > White, no. (%)
P value
5 (5.9) 5 (5.9) 1 (1.2) 0 0 0 0 0
.15 .15 .3 .12 .47 .60 .21 .60
BA, no. (%)
56 44 21 6 3 1
(10.2) (8.0) (3.8) (1.1) (0.6) (0.5) 0 2 (0.4)
FA. The maximum prevalence (Table III) assumes all missed challenge responses were positive and likely overestimates the true prevalence of FA. Therefore an adjusted prevalence rate was calculated to more accurately model the true rate in this cohort.
Allergy and ethnicity in the urban cohort The rate of sensitization at SPT responses of 1 mm or greater and 3 mm or greater to any food, peanut, and egg was greater in urban BA participants than in white participants and participants of mixed ethnicity (Table IV). This difference was only statistically significant for an SPT response of 1 mm or greater regarding sensitization to egg. There was a nonsignificant trend showing the BA > MA > white for cumulative SPT response sizes for all food allergens combined (P 5 .109, Kruskal-Wallis test; Fig 1). Allergy in the rural cohort Twenty-one (5.3%) participants in the rural BA cohort were _1 mm). The highest prevalence of sensitized to food at any level (> sensitization was to egg, followed by cow’s milk, peanut, soya, _3 mm), and fish. At greater levels of sensitization (SPT response > rural participants were only sensitized to egg and milk (Table V). The prevalence of FS at 1 and 3 mm in the rural BA cohort (5.3% and 2.8%) is significantly lower than the prevalence in the urban cohort overall (11.6% [137] and 9.0% [107], both P <.001) and in urban BA participants (12.9% [71] and 10.2% [61], both P <.001; Fig 2). Three (0.8%) of the rural participants were polysensitized (compared with 4.3% of urban participants), with 2 (0.5%) sensitized to 2 foods and 1 (0.3%) sensitized to 3 or more foods (P < .001). Per-protocol analysis of the rural cohort showed 0.5% (95% CI, 0.1% to 1.8%) of children had FA (all to raw egg, with 0.3% allergic to cooked egg). The prevalence of FA in the rural
Positive OFC result
MA, no. (%)
White, no. (%)
P value
BA, no. (%)
MA, no. (%)
White, no. (%)
P value
46 27 17 10 2
5 (5.9) 5 (5.9) 1 (1.2) 0 0 0 0 0
.39 .12 .51 .41 1.0 1.0 .357 .78
20 20 8 2
23 17 8 2
3 (3.5) 3 (5.3) 0 0 0 0 0 0
.94 .82 .85 1.0 — — — 1.0
(8.4) (4.9) (3.1) (1.8) (0.4) 0 2 (0.4) 3 (0.6)
(3.6) (3.6) (1.5) (0.4) 0 0 0 1 (0.2)
(4.2) (3.1) (1.5) (0.4) 0 0 0 1 (0.2)
BA cohort is significantly less (P 5 .014) than the prevalence in the urban cohort overall (2.5%) and in urban BA participants (2.9%; 95% CI, 1.5% to 4.3%; P 5 .007; Fig 2).
DISCUSSION The study found that the prevalence of FA in urban South African children is comparable with other middle-income countries in which industrialization and rapid urbanization are taking place.1 Contrary to what we expected, there was no protection against FS or FA in urban BA participants versus urban children of white ancestry and MA. The pattern of sensitization was similar in all cohorts, most commonly to egg then peanut. The higher rates of raw egg allergy versus cooked egg allergy indicate that egg allergy can already be in the process of being outgrown in some children of this age group. Our urban cohort was recruited from ECD centers, representing a true unselected population. High participation and completion rates (98.2% urban and 96.1% rural) and no significant difference in atopic background between participants versus nonparticipants means selection bias is unlikely (see Table E2). The rural cohort was selected by means of community sampling by using community health clinics as hubs, which might be expected to increase the measured FS and FA rates; however, this is unlikely to be marked because the concern regarding allergies in the community is very low. The completion rate to complete food challenges of 79.4% in those in whom a challenge was indicated prompted a correction for missing challenges by allocating a probability of a positive challenge response using the correlation between sensitization and FA. A strength of this approach is the use of our own data in this cohort because the correlation between FS and FA is known to be population specific. Correction of FA prevalence for incomplete challenges differed only slightly from the intent-to-treat data, indicating that the missed challenges did not markedly affect our prevalence estimates.
BOTHA ET AL 5
J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn
FIG 1. Cumulative SPT response size in urban participants according to ethnicity.
TABLE V. The spectrum of sensitization and FA in the rural cohort _1 mm, SPT response > no. (% [95% CI])
Any food Egg (any) Peanut Cow’s milk Hazelnut Soya Wheat Fish
21 13 1 9
(5.3 (3.3 (0.3 (2.3
[2.1-7.5]) [1.5-5.0]) [0.0-0.8]) [0.8-3.7]) 0 1 (0.3 [0.3-1.0]) 0 1 (0.3 [0.0-1.4])
_3 mm, SPT response > no. (% [95% CI])
OFC-confirmed FA, no. (% [95% CI])
11 (2.8 [1.2-4.4]) 12 (3.0 [1.3-4.7]) 0 1 (0.3 [0.2-0.7]) 0 0 0 0
2 (0.5 [0.1-1.8]) Raw egg white: 2 (0.5 [0.1-1.8]) Cooked whole egg: 1 (0.3 [0.0-1.4]) 0 0 0 0 0 0
The higher prevalence of comorbid allergy, as well as a positive family history in the urban population, can reflect either a parallel phenomenon (with the same factors over time having affected both family history and comorbid allergy) or (especially in the case of atopic dermatitis) a precursor to the increased prevalence of FA found in urban children. Raw egg challenges were used to determine FA prevalence rates, as well as (raw) egg allergy prevalence. However, determination of tolerance in the questionnaire and an indication for OFC was based on tolerance to a full age-appropriate portion of cooked egg. Thus it is likely that some cases of raw egg allergy would have been missed. The rate of allergy to raw egg white was approximately 60% greater than the rate of cooked egg allergy. Comparison of studies of FA prevalence should ideally document
the form in which the challenge was performed and include both raw and cooked egg if accurate comparison with prior studies is envisaged. The study suggests that environment might be more important than ethnicity in the development of FS and FA in these cohorts. The expectation that we would find a lower prevalence of FS in urban BA subjects because of differences in affluence was not met, indicating that it is likely not affluence itself but some other factor that can account for rural versus urban differences. Factors that can differ in the rural and urban participants that might influence allergy include differences in microbiome, pregnancy and birth metrics, diet, breast-feeding, unpasteurized milk ingestion, exposure to aeroallergens and pets, tobacco and fossil fuel exposure, parasite infestation, medication and
6 BOTHA ET AL
J ALLERGY CLIN IMMUNOL nnn 2018
FIG 2. FS and FA in urban and rural participants (and BA urban participants).
probiotic use, sunlight exposure and vitamin D status, and potentially others. Future statistical analyses aim at identifying the contributions of these factors to the observed differences between urban and rural participants. In view of the increasing prevalence of aeroallergen sensitization, asthma, and allergic rhinitis and the narrowing of the urban versus rural gradient over time, the possibility exists for an increase in FA prevalence as rural populations become more urbanized, with loss of rural protective factors and exposure to urban proinflammatory factors. Furthermore, although the rates of sensitization and allergy in the rural cohort are very low compared with those in the urban cohort, there is a similar pattern of FS, indicating that loss of tolerance to food allergens might already be occurring in deeply rural BA communities.
In addition, the higher rate of FS in BA participants in urban communities indicates that as such populations become urbanized, an FA epidemic can arise, which is of significant concern in a middle-income country that carries the double burden of infectious diseases and a rapid increase in noncommunicable diseases. Further analysis of this cohort to be published will look at environmental risk factors, including the number of children in the house, family position, tobacco smoke exposure during pregnancy, tobacco smoke and fossil fuel emissions in the environment, breast-feeding, infant diet, exposure to nonpasteurized milk, presence of domestic and farm animals, exposure to antibiotics and paracetamol, immunizations received, and past infections.
BOTHA ET AL 7
J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn
Limitations In the urban cohort we assessed children attending ECD centers, therefore excluding children who are at home. The effect of this selection bias, if any, would be postulated to yield a lower prevalence of allergy in children attending ECD centers than those staying at home, with an underestimation of urban prevalence. Although sampling of ECD centers was random, consent for participation was required from both ECD principals and parents. We do not have data on characteristics of centers that did and did not participate. Although our participation and completion rates overall were considerably greater than those reported in other studies, we had significant loss to follow-up for participants who qualified for a food challenge. The use of adjusted prevalence rates mitigates against this limitation. CONCLUSION This is the first population-based study determining the prevalence of food allergies in South African children. It demonstrates that urban versus rural differences are likely a larger factor in the development of FS and FA than ethnicity in this cohort. This epidemiologic study lays a foundation for further study of allergic diseases in transitional communities with the view to looking at specific environmental and nutritional factors that accelerate the increase in noncommunicable diseases associated with urbanization and industrialization. It also highlights the importance of further mechanistic work to evaluate biologically plausible mechanisms9 and find solutions to mitigate their effects. We thank the management and clinical team of Zithulele Hospital and the community of Mqanduli district for their invaluable support of the study.
Key messages d
This is the first population-based comparative study of FA prevalence in urban and rural children and puts FA prevalence in urban South Africa on par with that in other rapidly urbanizing middle-income countries.
d
Significantly lower FA prevalence in rural communities is likely a result of environmental rather than genetic differences.
REFERENCES 1. Levin ME, Gray CL, Marrugo J. Food allergy: international and developing world perspectives. Curr Pediatr Rep 2016;4:129-37. 2. Walkner M, Warren C, Gupta RS. Quality of life in food allergy patients and their families. Pediatr Clin North Am 2015;62:1453-61. 3. Liu AH, Jaramillo R, Sicherer SH, Wood RA, Bock SA, Burks AW, et al. National prevalence and risk factors for food allergy and relationship to asthma: results from the National Health and Nutrition Examination Survey 2005-2006. J Allergy Clin Immunol 2010;126:798-806.e14. 4. Allen KJ, Koplin JJ. Why does Australia appear to have the highest rates of food allergy? Pediatr Clin North Am 2015;62:1441-51. 5. Zar HJ, Ehrlich RI, Workman L, Weinberg EG. The changing prevalence of asthma, allergic rhinitis and atopic eczema in African adolescents from 1995 to 2002. Pediatr Allergy Immunol 2007;18:560-5. 6. Levin ME, Muloiwa R, Motala C. Associations between asthma and bronchial hyper-responsiveness with allergy and atopy phenotypes in urban black South African teenagers. South African Med J 2011;101:472-6. 7. Keet CA, Savage JH, Seopaul S, Peng RD, Wood RA, Matsui EC. Temporal trends and racial/ethnic disparity in self-reported pediatric food allergy in the United States. Ann Allergy Asthma Immunol 2014;112:222-9. 8. Basera W, Botha M, Gray CL, Lunjani N, Watkins AS, Venter C, et al. The South African Food Sensitisation and Food Allergy population-based study of IgE-mediated food allergy: validity, safety, and acceptability. Ann Allergy Asthma Immunol 2015;115:113-9. 9. Matsui EC, Keet CA. Weighing the evidence: bias and confounding in epidemiologic studies in allergy/immunology. J Allergy Clin Immunol 2017;139:448-50.
7.e1 BOTHA ET AL
J ALLERGY CLIN IMMUNOL nnn 2018
TABLE E1. CONSORT table for study cohort
Responders Nonparticipants Participants Ineligible (>36 months) Eligible participants Questionnaire incomplete SPT incomplete Total Sensitized OFC indicated OFC incomplete (none done at all) OFC partially complete
Urban
Rural
1230 25 1205 5 1200 4 11 1185 136 (11.5%) 64 (5.5%) 7 7
410 0 410 0 410 0 12 398 18 (4.5%) 7 (1.8%) 1 0
BOTHA ET AL 7.e2
J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn
TABLE E2. Description of participants versus nonparticipants Urban
Median age (mo [IQR]) Sex distribution, no. (%) Male Female First-degree relative with any allergy First-degree relative with asthma First-degree relative with hay fever First-degree relative with eczema First-degree relative with FA Comorbid allergies Asthma (self-reported) Allergic rhinitis (self-reported) Allergic rhinitis (clinically confirmed) Eczema (self-reported) Eczema (clinically confirmed) Any comorbid allergy (clinically confirmed) Any comorbid allergy (self-reported or clinically confirmed) IQR, Interquartile range.
Rural
Participants (n 5 1185)
Nonparticipants (n 5 27)
P value
Participants (n 5 398)
Nonparticipants (n 5 12)
P value
26 (22–32)
26 (21–32)
.65
21 (17–28)
16.5 (13.5–24.5)
.06
622 563 578 207 421 195 59
(52.5) (47.5) (48.8) (17.5) (35.5) (16.5) (5.0)
10 17 12 7 7 2 2
.11
7 (58.3) 5 (41.7) 0 0 0 0 0
.90 1.00 1.00 1.00 1.00 1.00
106 279 101 279 139 221 535
(9.0) (23.5) (8.5) (23.5) (11.7) (18.6) (45.2)
4 9 2 6
0 0 0 0 0 0 0
1.00 1.00 — 1.00 1.00 1.00 1.00
(37.0) (63.0) (44.4) (25.9) (25.9) (7.4) (7.4)
.66 .30 .30 .29 .64
(14.8) (34.6) (7.4) (23.1) 0 2 (7.4) 13 (48.2)
.29 .19 .84 .96 1.00 .14 .76
225 173 30 12 6 2 10
(56.5) (43.5) (7.5) (3.0) (1.5) (0.5) (2.5)
4 (1.0) 13 (3.3) 0 7 (1.8) 3 (0.8) 3 (0.8) 23 (5.8)
Key words: Africa, allergy, egg, ethnicity, food allergy, food sensitization, peanut, prevalence, skin prick test, urban, rural
Food allergy (FA) is a growing public health concern in both middle- and high-income countries.1 There is as yet no cure, and moreover, the quality of life of children with FA and their families is significantly affected by the economic, nutritional, and social effect of having an FA.2 Worldwide, the prevalence of FA in children ranges from 2% to 10%, depending on the age of the children studied, range of foods tested, methodologies used, and geographic area.1,3,4 In South Africa, as in other transitional countries, the prevalence of noncommunicable diseases in children is a growing public health concern. Studies in South Africa have clearly shown an increase in asthma, allergic rhinitis, and atopic dermatitis over time.5 In addition, rural communities experience a lower prevalence of respiratory allergy, but the ‘‘protection’’ in rural areas appears to be decreasing over time, with a narrowing of the urban versus rural gradient.6 FA has been investigated in an urban tertiary care setting among children in Cape Town with moderate-to-severe atopic dermatitis, showing high rates of food sensitization (FS) and FA comparable with those found in similar highly selected populations from other countries.7 The same study also showed lower rates of FA, particularly peanut allergy, in black African (BA) children compared with children of mixed ancestry (MA). This raises the question of whether there are protective factors reducing the development of allergies in BA children that might have either a genetic or environmental basis. Data from studies elsewhere in the world have suggested differences in sensitization and allergy rates according to socioeconomic class and between children of different ethnicities and between children of newly urbanized immigrants and local populations.4,7 No previous large-scale epidemiologic data on FA and FS exist in South Africa, and no prior data on FA prevalence in an unselected population have been collected. The hypothesis for this study was that allergy and sensitization rates would be lower in urban BA participants than urban participants of white origin or MA and even lower in rural BA participants.
METHODS Study design and setting In this cross-sectional study 1207 children (age, 12-36 months) who attended registered early child development (ECD) centers in Cape Town were recruited between February 2013 and December 2016. ECD centers are government-regulated daycare centers for preschool children that are partially 1
2 BOTHA ET AL
Abbreviations used BA: Black African ECD: Early child development FA: Food allergy FS: Food sensitization MA: Mixed ancestry OFC: Oral food challenge SPT: Skin prick test
funded through subsidies and serve most children in South Africa. Four hundred fourteen participants in the same age group were recruited in the rural Mqanduli district of the Eastern Cape by raising awareness through local primary health care clinics and community stakeholders. This age range was chosen to cover the range at which we postulated that FA would be prevalent. The majority of BA subjects living in Cape Town are AmaXhosa. They are the second largest ethnolinguistic group in South Africa and reside mainly in the Eastern Cape province. Migration to Cape Town from the rural Eastern Cape is ongoing but has been particularly rapid in the last 30 years. The Mqanduli district was chosen because of the high rate of migration from this area to Cape Town and the availability of an excellent district hospital to use as a site for food challenges.
J ALLERGY CLIN IMMUNOL nnn 2018
the back when eczema precluded use of the forearm. An SPT response of 1 mm or greater than that elicited by the negative control was considered ‘‘reactive’’ and sufficiently high to merit a food challenge because previous studies have found that young children with sensitization manifesting in am SPT response as small as 1 mm can have challenge-proved FA.8 _1 mm) who were not Participants with any reactive SPTs (> (on questionnaire) tolerating a full age-appropriate portion of the specific food thus qualified for an open oral food challenge (OFC). Participants with a reactive SPT who were tolerating a regular age-appropriate portion with no history of a reaction were considered food sensitized but not allergic. OFCs were performed at Red Cross Children’s Hospital in Cape Town and Zithulele Hospital in the Eastern Cape. Challenges were performed as open incremental oral challenges by using a standardized protocol with predetermined objective criteria for a positive challenge response.8 Challenges used full-fat cow’s milk, soy milk formula, peanut butter, steamed codfish, crushed hazelnuts, Weet-bix (Pioneer Foods, Cape Town, South Africa), and, for egg sensitization, both raw egg white and whole scrambled egg on separate occasions. OFCs in which participants did not reach the top dose but had no reaction were classified as equivocal if the challenge could not successfully be repeated. This study was approved by the Human Research Ethics Committee of the University of Cape Town (HREC REF: 038/2012). Informed consent was obtained from parents or legal guardians of participants.
Sampling frame
Data entry and statistical analysis
In Cape Town we accessed the publicly available online database of registered ECD centers and approached the ECD centers in a randomized manner to participate in our study. Once an ECD center agreed to participate, all children who attended and were aged 12 to 36 completed months were invited to participate. ECD centers were enrolled until a target of more than 1200 participants was reached. There are almost no ECD centers in the Mqanduli district. Instead all eligible children from the areas surrounding the 10 district community health clinics were recruited. Participants who responded to the study invitation but decided not to participate or who could not complete the study were considered nonparticipants.
Data were entered into a Microsoft Access database and analyzed by using Stata software (version 11.1; StataCorp, College Station, Tex). The Mann-Whitney test was used to investigate differences between numeric variables, all of which were nonnormally distributed. The Fisher exact test was used to assess overall associations between pairs of categorical variables, and the z test using normal approximation to the binomial distribution was used to investigate differences in proportions. A P value of less than .05 was considered statistically significant. Multiple logistic regression was performed with an OFC as the outcome and each SPT for each allergen used as the predictor variables by using this data set to calculate adjusted prevalence rates in the urban cohort. The probability of a positive OFC response was then calculated and used for each participant who did not complete an indicated OFC to add them to the analysis of adjusted prevalence.
Sample size Our sample size calculations were based on estimated prevalences of 0%, 2%, and 8% of FA in rural BA, urban BA, and urban MA children, respectively. By using these estimates, sample sizes of 400 for each cohort would give us 80% power to detect a 5% difference in prevalence between the 2 urban cohorts and also detect a difference between the 2 BA cohorts of 2% or greater _2%). Pilot data showed that our sampling (ie, rural, almost 0%; urban, > method produced a representative cohort according to the Cape Town Census, meaning that about 40% to 50% of the urban cohort would be BA subjects.8 These proportions were used to expand the urban cohort proportionally to 1200, producing enough power to match a rural cohort of 400. Power calculations were done based on the z test for differences in proportions.
Recruitment and assessment Questionnaires were administered to all participating parents regarding other allergic diseases or a family history of allergic diseases. Children were examined for signs of atopic dermatitis (visible skin lesions) and allergic rhinitis (allergic shiners, nasal crease, Dennie lines, and enlarged turbinates). Skin prick test (SPTs) were performed with ALK-Abello SPT solutions (Thermo Fisher, Waltham, Mass) to 7 common food allergens (egg white extract, peanut, cow’s milk, codfish, soya, wheat [flour], and hazelnut). In addition, fresh egg white, peanut butter, and fresh cow’s milk, as well as positive (10 mg/mL histamine) and negative (saline) controls, were used. SPT responses were read at 15 minutes and recorded as average wheal diameter in millimeters. Where fresh extracts and commercial extracts were used (egg, milk, and peanut), the highest of the 2 SPT responses for the same allergen source were analyzed for the purposes of this study. SPTs were performed with ALK-Abello lancets on the volar aspect of the forearm or on
RESULTS One thousand one hundred eighty-five urban and 398 rural participants completed the study (see Table E1 in this article’s Online Repository at www.jacionline.org). There were no statistically significant differences in median age, sex, family history of allergy, or history of allergic diseases (eczema, asthma, and rhinitis) or food reactions between study participants and nonparticipants (see Table E2 in this article’s Online Repository at www.jacionline.org). The median age of the urban cohort was 26 months (interquartile range, 22-32 months), which was older than the rural cohort at 21 months (interquartile range, 17-28 months; P <.001; Table I). There were significantly more male than female participants in both cohorts (urban cohort, 52.5% male and 47.5% female [P 5 .01]; rural cohort, 56.5% male and 43.5% female [P 5 .01]), but these ratios were not significantly different between the urban and rural cohorts (P 5 .162). The urban cohort reflected the ethnic demographics of children less than 5 years old in the Cape Town metropole, as captured by the 2011 census, with 46.3% of participants being BA and 46.5% and 7.2% being of MA or white origin, respectively
BOTHA ET AL 3
J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn
TABLE I. Description of the study population by study site Urban (n 5 1185)
Rural (n 5 398)
P value*
26 (22-32)
21 (17-28)
<.001
622 (52.5%)
225 (56.5%)
M/F ratio: .16
563 (47.5%)
173 (43.7%)
Enrollment age (mo) Median (IQR) Sex, no. (%) Male Female
Self-reported family history of allergic disease, no. (%) First-degree relative with any allergy First-degree relative with asthma First-degree relative with allergic rhinitis First-degree relative with atopic dermatitis First-degree relative with FA Self-reported comorbid allergies, no. (%) Any Asthma Allergic rhinitis Atopic dermatitis Allergic disease confirmed on clinical examination, no. (%) Allergic rhinitis Atopic dermatitis
578 207 421 195 59
(48.8%) (17.5%) (35.5%) (16.5%) (5.0%)
30 12 6 2 10
(7.5%) (3.0%) (1.5%) (0.5%) (2.5%)
<.001 <.001 <.001 <.001 <.001
502 106 279 279
(42.4%) (9.0%) (23.5%) (23.5%)
22 4 13 7
(5.5%) (1.0%) (3.3%) (1.8%)
<.001 .002 <.001 <.001
0 (0.0%) 3 (0.8%)
<.001 <.001
101 (8.5%) 139 (11.7%)
IQR, Interquartile range. *P values are from Fisher exact tests for categorical values and 2-sample Wilcoxon rank-sum (Mann-Whitney) tests for continuous variables.
TABLE II. Spectrum of sensitization and FA in the urban cohort _1 mm, SPT response > no. (% [95% CI])
Any food Egg (any) Peanut Cow’s milk Hazelnut Soya Wheat Fish
137 90 54 38 13 17 13 9
(11.6 [9.7- 13.4]) (7.6 [6.1-9.1]) (4.6 [3.4-5.8]) (3.2 [2.2-4.2]) (1.1 [0.5-1.7]) (1.4 [0.8-2.1]) (1.1 [0.5-1.7]) (0.8 [0.3-1.3])
_3 mm, SPT response > no. (% [95% CI])
107 78 40 16 5 5 1 2
(9.0 (6.6 (3.4 (1.4 (0.4 (0.4 (0.1 (0.2
[7.4-10.7]) [5.0-7.8]) [2.4-4.6]) [0.7-2.0]) [0.1-0.8]) [0.1-0.8]) [0.1-0.3]) [0.1-0.4])
(Census 2011: 46.4% BA, 45.4% MA, and 8.2% white). The rural cohort was entirely BA. Urban participants were significantly more likely than rural participants to have a family history of allergic diseases (48.8% [578/1185] vs 7.5% [30/398], P < .001) and reported a much higher prevalence of comorbid eczema and allergic rhinitis than their rural counterparts (23.5% [279] vs 3.3% [130] and 1.8% [7], P < .001; Table I). One hundred thirty-nine (11.7%) urban participants had objective signs of atopic dermatitis (AD), and 101 (8.5%) had signs of allergic rhinitis (AR). In the rural cohort AD and AR were significantly less frequent (0.8% [3] and 0.0%, both P <.001) than in the urban sample (Table I). Seventy-five (6.3%) urban participants self-reported a previous reaction to one of the 7 foods (cow’s milk, 2.5%; egg, 1.9%; fish, 1.8%; peanuts, 1.0%; soya, 0.3%; and wheat, 0.1%) compared with 13 (3.3%) of 398 (P 5 .02) of rural participants (cow’s milk, 1.5%; fish, 1.0%; soya, 0.5%; and wheat, 0.3%). One hundred thirty-seven (11.6%) of the urban cohort were _1-mm response). The highest sensitized to food at any level (>
OFC-confirmed FA, no. (% [95% CI])
27 (2.3 [1.5-3.3]) Raw egg white: 21 (1.8 [1.1-2.7]) Cooked whole egg: 9 (0.8 [0.3-1.4]) 8 (0.7 [0.3-1.3]) 1 (0.1 [0.0-0.5]) 0 0 0 1 (0.1 [0.0-0.5])
prevalence of sensitization was to egg, followed by peanut, cow’s milk, hazelnut, soya, wheat, and fish (Table II). Fifty-one (4.3%) of the urban participants were polysensitized, with 27 (2.3%) sensitized to 2 foods and 24 (2.0%) sensitized to 3 foods or more. Per-protocol analysis of the prevalence of FA was 2.3% (95% CI, 1.5% to 3.7%) in urban children. The most common allergy was to raw egg (1.8%), followed by cooked egg (0.8%), peanut (0.7%) cow’s milk (0.1%), and fish (0.1%; Table II). Male participants had a higher rate of FA than female participants (3.2% vs 1.2%, P 5 .02). Food challenges were indicated in 34 participants based on an SPT response of 1 mm or greater without a history of tolerating a full age-appropriate portion of the food. No participants were classified as having FA on the basis of a recent severe reaction. In 7 (20.6%) of these participants, 1 or more challenges were not completed because participants could not be contacted or did not attend. Individual uncompleted food challenges were to milk (2/3 indicated), cooked egg (6/15 indicated), peanut (2/10 indicated), and raw egg (4/25 indicated). Per-protocol analysis represents only the proved FA cases and therefore likely underestimates the true prevalence of
4 BOTHA ET AL
J ALLERGY CLIN IMMUNOL nnn 2018
TABLE III. Adjusted FA prevalence in the urban cohort Per-protocol analysis (minimum FA prevalence)
Any food, no. (% [95% CI]) Raw egg white, no. (% [95% CI]) Cooked whole egg, no. (% [95% CI]) Peanut, no. (% [95% CI]) Cow’s milk, no. (% [95% CI]) Fish, no. (% [95% CI])
27 21 9 8 1 1
(2.3 (1.8 (0.8 (0.7 (0.1 (0.1
Completed plus missed OFCs
[1.5-3.7]) [1.1-2.7]) [0.3-1.4]) [0.3-1.3]) [0.0-0.5]) [0.0-0.5])
27 21 9 8 1 1
1 1 1 1 1 1
Maximum FA prevalence
7 4 6 2 2 0
34 25 15 10 3 1
(2.9 (2.1 (1.3 (0.8 (0.3 (0.1
[2.0-4.0]) [1.4-3.1]) [0.7-2.1]) [0.4-1.6]) [0.1-0.7]) [0.0-0.5])
Adjusted FA prevalence
2.5 1.9 0.8 0.8
(1.6-3.3) (1.1-2.7) (0.3-2.3) (0.3-1.3) NA NA
NA, Not applicable.
TABLE IV. Urban sensitization and FA by ethnicity _1 mm > BA, no. (%)
Any food Egg Peanut Cow’s milk Soya Wheat Fish Hazelnut
71 51 28 16 7 5 2 5
(12.9) (9.3) (5.1) (2.9) (1.3) (0.9) (0.4) (0.9)
MA, no. (%)
61 34 25 22 10 8 7 8
(10.7) (6.2) (4.6) (4.0) (1.8) (1.5) (1.3) (1.5)
_3 mm > White, no. (%)
P value
5 (5.9) 5 (5.9) 1 (1.2) 0 0 0 0 0
.15 .15 .3 .12 .47 .60 .21 .60
BA, no. (%)
56 44 21 6 3 1
(10.2) (8.0) (3.8) (1.1) (0.6) (0.5) 0 2 (0.4)
FA. The maximum prevalence (Table III) assumes all missed challenge responses were positive and likely overestimates the true prevalence of FA. Therefore an adjusted prevalence rate was calculated to more accurately model the true rate in this cohort.
Allergy and ethnicity in the urban cohort The rate of sensitization at SPT responses of 1 mm or greater and 3 mm or greater to any food, peanut, and egg was greater in urban BA participants than in white participants and participants of mixed ethnicity (Table IV). This difference was only statistically significant for an SPT response of 1 mm or greater regarding sensitization to egg. There was a nonsignificant trend showing the BA > MA > white for cumulative SPT response sizes for all food allergens combined (P 5 .109, Kruskal-Wallis test; Fig 1). Allergy in the rural cohort Twenty-one (5.3%) participants in the rural BA cohort were _1 mm). The highest prevalence of sensitized to food at any level (> sensitization was to egg, followed by cow’s milk, peanut, soya, _3 mm), and fish. At greater levels of sensitization (SPT response > rural participants were only sensitized to egg and milk (Table V). The prevalence of FS at 1 and 3 mm in the rural BA cohort (5.3% and 2.8%) is significantly lower than the prevalence in the urban cohort overall (11.6% [137] and 9.0% [107], both P <.001) and in urban BA participants (12.9% [71] and 10.2% [61], both P <.001; Fig 2). Three (0.8%) of the rural participants were polysensitized (compared with 4.3% of urban participants), with 2 (0.5%) sensitized to 2 foods and 1 (0.3%) sensitized to 3 or more foods (P < .001). Per-protocol analysis of the rural cohort showed 0.5% (95% CI, 0.1% to 1.8%) of children had FA (all to raw egg, with 0.3% allergic to cooked egg). The prevalence of FA in the rural
Positive OFC result
MA, no. (%)
White, no. (%)
P value
BA, no. (%)
MA, no. (%)
White, no. (%)
P value
46 27 17 10 2
5 (5.9) 5 (5.9) 1 (1.2) 0 0 0 0 0
.39 .12 .51 .41 1.0 1.0 .357 .78
20 20 8 2
23 17 8 2
3 (3.5) 3 (5.3) 0 0 0 0 0 0
.94 .82 .85 1.0 — — — 1.0
(8.4) (4.9) (3.1) (1.8) (0.4) 0 2 (0.4) 3 (0.6)
(3.6) (3.6) (1.5) (0.4) 0 0 0 1 (0.2)
(4.2) (3.1) (1.5) (0.4) 0 0 0 1 (0.2)
BA cohort is significantly less (P 5 .014) than the prevalence in the urban cohort overall (2.5%) and in urban BA participants (2.9%; 95% CI, 1.5% to 4.3%; P 5 .007; Fig 2).
DISCUSSION The study found that the prevalence of FA in urban South African children is comparable with other middle-income countries in which industrialization and rapid urbanization are taking place.1 Contrary to what we expected, there was no protection against FS or FA in urban BA participants versus urban children of white ancestry and MA. The pattern of sensitization was similar in all cohorts, most commonly to egg then peanut. The higher rates of raw egg allergy versus cooked egg allergy indicate that egg allergy can already be in the process of being outgrown in some children of this age group. Our urban cohort was recruited from ECD centers, representing a true unselected population. High participation and completion rates (98.2% urban and 96.1% rural) and no significant difference in atopic background between participants versus nonparticipants means selection bias is unlikely (see Table E2). The rural cohort was selected by means of community sampling by using community health clinics as hubs, which might be expected to increase the measured FS and FA rates; however, this is unlikely to be marked because the concern regarding allergies in the community is very low. The completion rate to complete food challenges of 79.4% in those in whom a challenge was indicated prompted a correction for missing challenges by allocating a probability of a positive challenge response using the correlation between sensitization and FA. A strength of this approach is the use of our own data in this cohort because the correlation between FS and FA is known to be population specific. Correction of FA prevalence for incomplete challenges differed only slightly from the intent-to-treat data, indicating that the missed challenges did not markedly affect our prevalence estimates.
BOTHA ET AL 5
J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn
FIG 1. Cumulative SPT response size in urban participants according to ethnicity.
TABLE V. The spectrum of sensitization and FA in the rural cohort _1 mm, SPT response > no. (% [95% CI])
Any food Egg (any) Peanut Cow’s milk Hazelnut Soya Wheat Fish
21 13 1 9
(5.3 (3.3 (0.3 (2.3
[2.1-7.5]) [1.5-5.0]) [0.0-0.8]) [0.8-3.7]) 0 1 (0.3 [0.3-1.0]) 0 1 (0.3 [0.0-1.4])
_3 mm, SPT response > no. (% [95% CI])
OFC-confirmed FA, no. (% [95% CI])
11 (2.8 [1.2-4.4]) 12 (3.0 [1.3-4.7]) 0 1 (0.3 [0.2-0.7]) 0 0 0 0
2 (0.5 [0.1-1.8]) Raw egg white: 2 (0.5 [0.1-1.8]) Cooked whole egg: 1 (0.3 [0.0-1.4]) 0 0 0 0 0 0
The higher prevalence of comorbid allergy, as well as a positive family history in the urban population, can reflect either a parallel phenomenon (with the same factors over time having affected both family history and comorbid allergy) or (especially in the case of atopic dermatitis) a precursor to the increased prevalence of FA found in urban children. Raw egg challenges were used to determine FA prevalence rates, as well as (raw) egg allergy prevalence. However, determination of tolerance in the questionnaire and an indication for OFC was based on tolerance to a full age-appropriate portion of cooked egg. Thus it is likely that some cases of raw egg allergy would have been missed. The rate of allergy to raw egg white was approximately 60% greater than the rate of cooked egg allergy. Comparison of studies of FA prevalence should ideally document
the form in which the challenge was performed and include both raw and cooked egg if accurate comparison with prior studies is envisaged. The study suggests that environment might be more important than ethnicity in the development of FS and FA in these cohorts. The expectation that we would find a lower prevalence of FS in urban BA subjects because of differences in affluence was not met, indicating that it is likely not affluence itself but some other factor that can account for rural versus urban differences. Factors that can differ in the rural and urban participants that might influence allergy include differences in microbiome, pregnancy and birth metrics, diet, breast-feeding, unpasteurized milk ingestion, exposure to aeroallergens and pets, tobacco and fossil fuel exposure, parasite infestation, medication and
6 BOTHA ET AL
J ALLERGY CLIN IMMUNOL nnn 2018
FIG 2. FS and FA in urban and rural participants (and BA urban participants).
probiotic use, sunlight exposure and vitamin D status, and potentially others. Future statistical analyses aim at identifying the contributions of these factors to the observed differences between urban and rural participants. In view of the increasing prevalence of aeroallergen sensitization, asthma, and allergic rhinitis and the narrowing of the urban versus rural gradient over time, the possibility exists for an increase in FA prevalence as rural populations become more urbanized, with loss of rural protective factors and exposure to urban proinflammatory factors. Furthermore, although the rates of sensitization and allergy in the rural cohort are very low compared with those in the urban cohort, there is a similar pattern of FS, indicating that loss of tolerance to food allergens might already be occurring in deeply rural BA communities.
In addition, the higher rate of FS in BA participants in urban communities indicates that as such populations become urbanized, an FA epidemic can arise, which is of significant concern in a middle-income country that carries the double burden of infectious diseases and a rapid increase in noncommunicable diseases. Further analysis of this cohort to be published will look at environmental risk factors, including the number of children in the house, family position, tobacco smoke exposure during pregnancy, tobacco smoke and fossil fuel emissions in the environment, breast-feeding, infant diet, exposure to nonpasteurized milk, presence of domestic and farm animals, exposure to antibiotics and paracetamol, immunizations received, and past infections.
BOTHA ET AL 7
J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn
Limitations In the urban cohort we assessed children attending ECD centers, therefore excluding children who are at home. The effect of this selection bias, if any, would be postulated to yield a lower prevalence of allergy in children attending ECD centers than those staying at home, with an underestimation of urban prevalence. Although sampling of ECD centers was random, consent for participation was required from both ECD principals and parents. We do not have data on characteristics of centers that did and did not participate. Although our participation and completion rates overall were considerably greater than those reported in other studies, we had significant loss to follow-up for participants who qualified for a food challenge. The use of adjusted prevalence rates mitigates against this limitation. CONCLUSION This is the first population-based study determining the prevalence of food allergies in South African children. It demonstrates that urban versus rural differences are likely a larger factor in the development of FS and FA than ethnicity in this cohort. This epidemiologic study lays a foundation for further study of allergic diseases in transitional communities with the view to looking at specific environmental and nutritional factors that accelerate the increase in noncommunicable diseases associated with urbanization and industrialization. It also highlights the importance of further mechanistic work to evaluate biologically plausible mechanisms9 and find solutions to mitigate their effects. We thank the management and clinical team of Zithulele Hospital and the community of Mqanduli district for their invaluable support of the study.
Key messages d
This is the first population-based comparative study of FA prevalence in urban and rural children and puts FA prevalence in urban South Africa on par with that in other rapidly urbanizing middle-income countries.
d
Significantly lower FA prevalence in rural communities is likely a result of environmental rather than genetic differences.
REFERENCES 1. Levin ME, Gray CL, Marrugo J. Food allergy: international and developing world perspectives. Curr Pediatr Rep 2016;4:129-37. 2. Walkner M, Warren C, Gupta RS. Quality of life in food allergy patients and their families. Pediatr Clin North Am 2015;62:1453-61. 3. Liu AH, Jaramillo R, Sicherer SH, Wood RA, Bock SA, Burks AW, et al. National prevalence and risk factors for food allergy and relationship to asthma: results from the National Health and Nutrition Examination Survey 2005-2006. J Allergy Clin Immunol 2010;126:798-806.e14. 4. Allen KJ, Koplin JJ. Why does Australia appear to have the highest rates of food allergy? Pediatr Clin North Am 2015;62:1441-51. 5. Zar HJ, Ehrlich RI, Workman L, Weinberg EG. The changing prevalence of asthma, allergic rhinitis and atopic eczema in African adolescents from 1995 to 2002. Pediatr Allergy Immunol 2007;18:560-5. 6. Levin ME, Muloiwa R, Motala C. Associations between asthma and bronchial hyper-responsiveness with allergy and atopy phenotypes in urban black South African teenagers. South African Med J 2011;101:472-6. 7. Keet CA, Savage JH, Seopaul S, Peng RD, Wood RA, Matsui EC. Temporal trends and racial/ethnic disparity in self-reported pediatric food allergy in the United States. Ann Allergy Asthma Immunol 2014;112:222-9. 8. Basera W, Botha M, Gray CL, Lunjani N, Watkins AS, Venter C, et al. The South African Food Sensitisation and Food Allergy population-based study of IgE-mediated food allergy: validity, safety, and acceptability. Ann Allergy Asthma Immunol 2015;115:113-9. 9. Matsui EC, Keet CA. Weighing the evidence: bias and confounding in epidemiologic studies in allergy/immunology. J Allergy Clin Immunol 2017;139:448-50.
7.e1 BOTHA ET AL
J ALLERGY CLIN IMMUNOL nnn 2018
TABLE E1. CONSORT table for study cohort
Responders Nonparticipants Participants Ineligible (>36 months) Eligible participants Questionnaire incomplete SPT incomplete Total Sensitized OFC indicated OFC incomplete (none done at all) OFC partially complete
Urban
Rural
1230 25 1205 5 1200 4 11 1185 136 (11.5%) 64 (5.5%) 7 7
410 0 410 0 410 0 12 398 18 (4.5%) 7 (1.8%) 1 0
BOTHA ET AL 7.e2
J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn
TABLE E2. Description of participants versus nonparticipants Urban
Median age (mo [IQR]) Sex distribution, no. (%) Male Female First-degree relative with any allergy First-degree relative with asthma First-degree relative with hay fever First-degree relative with eczema First-degree relative with FA Comorbid allergies Asthma (self-reported) Allergic rhinitis (self-reported) Allergic rhinitis (clinically confirmed) Eczema (self-reported) Eczema (clinically confirmed) Any comorbid allergy (clinically confirmed) Any comorbid allergy (self-reported or clinically confirmed) IQR, Interquartile range.
Rural
Participants (n 5 1185)
Nonparticipants (n 5 27)
P value
Participants (n 5 398)
Nonparticipants (n 5 12)
P value
26 (22–32)
26 (21–32)
.65
21 (17–28)
16.5 (13.5–24.5)
.06
622 563 578 207 421 195 59
(52.5) (47.5) (48.8) (17.5) (35.5) (16.5) (5.0)
10 17 12 7 7 2 2
.11
7 (58.3) 5 (41.7) 0 0 0 0 0
.90 1.00 1.00 1.00 1.00 1.00
106 279 101 279 139 221 535
(9.0) (23.5) (8.5) (23.5) (11.7) (18.6) (45.2)
4 9 2 6
0 0 0 0 0 0 0
1.00 1.00 — 1.00 1.00 1.00 1.00
(37.0) (63.0) (44.4) (25.9) (25.9) (7.4) (7.4)
.66 .30 .30 .29 .64
(14.8) (34.6) (7.4) (23.1) 0 2 (7.4) 13 (48.2)
.29 .19 .84 .96 1.00 .14 .76
225 173 30 12 6 2 10
(56.5) (43.5) (7.5) (3.0) (1.5) (0.5) (2.5)
4 (1.0) 13 (3.3) 0 7 (1.8) 3 (0.8) 3 (0.8) 23 (5.8)