2 METHODS
2.1 Participants
Asthmatic children, 5-18 years old, treated with ICS and NS at the Lung Institute, Tygerberg and Red Cross Children’s Hospitals in Cape Town, South Africa, whose hypothalamic-pituitary-adrenal axis (HPA) was previously assessed with a morning basal serum F and MTP testing,were re-recruited.1-2HPAS was diagnosed, if F< 83 nmol/l or the post-MTP (PMTP) ACTH < 106 pg/ml (23.5 pmol/l), 11-deoxycortisol (11DOC) < 208 nmol/l and 11DOC+C < 400 nmol/l.
2.2 Study design and methodology
A cross-sectional study was performed. Height, weight, sex and age were recorded as measured previously. The BMI z-score (Centre of Disease Control) was computed. Salivary samples were collected with an Oragene DNA collection kit (OG-500). Samples were stored at - 20º C. Once all the samples were collected, DNA was extracted. Genotyping for rs242941, rs1876828 and rs41423247 was performed by using TaqMan® polymerase chain reaction (PCR) assays.Ethical approval was granted by the ethics committees of both Stellenbosch University and the University of Cape Town. The study conformed to the standards of the Declaration of Helsinki. All participants and their parents gave their informed consent prior to inclusion of the study.
2.3 Statistical analysis and sample size considerations
2.3.1 Sample size
In the previous study130 patients had hypocortisolaemia or HPAS. The cross-sectional design allowed for an equal number of patients to be analysed across the full spectrum of post-MTP responses i.e. low, midrange and high (30 subjects each). The ACTH mean and standard deviation (SD) were calculated (ACTH mean=330pg/ml [73.3 pmol/l], SD=280pg/ml [62.2 pmol/l]) from the previous data. The proportional distribution of the genetic subgroups was derived from the paper of Tsartsali et al, table 4.7.5Power analysis with a sample size of 90 was done across the three ACTH groups. Significance level was taken as 0.05. A one-way analysis of variance test (ANOVA)was considered for the power analysis. The following mean values were specified under the alternative hypothesis: group 1 (low range) 230pg/ml (51.1 pmol/l), group 2 (midrange) 330pg/ml (73.3 pmol/l), group 3 (high range) 430pg/ml (95.5 pmol/l). This gives a 100 ACTH units difference between two adjacent groups. This is a relative difference of 36% of the SD. [(100/280)= 0.36] which is a conservative expected effect size. Even for unbalanced groups such as n1=24, n2=35 and n3=40, the power will be 81%. Thus the power is good if the sample size distribution across the genetic subgroups is moderately unbalanced. Power is improved by square root transformations of ACTH to stabilise the within group variances.
2.3.2 Statistical analysis
An online statistical tool (https://ihg.helmholtz-muenchen.de/cgi-bin/hw/hwa1.pl) and the R statistical packagewere used. Fisher’s Exact and Pearson’s Chi-squaretestswere conducted to test for statistical difference of minor allele counts. TheHardyWeinberg package version 1.6.3 installed in RStudio version 3.5.1 was used to assess whether the SNPs were inHardy-Weinberg equilibrium. To test for significance at SNP and genotype level, the Wald’s test in multinom() function from thennet package in R was utilized. PMTP ACTH data were square root transformed and a one-way ANOVA performed. The Wilcoxon test was used for pairwise multiple comparisons of differences in mean √PMTP ACTH per SNP. Linear analysis for BMI z-score, √PMTP ACTH and binomial analysis for HPAS were conducted with the glm() function in R. The best genetic model for rs41423247 was selected on the basis of the lowest significant Akaike information criterion. Significance level for all tests was taken as 0.05.
3 RESULTS
A total of 96 patients (94 from previous studies1,2and two newpatients) were recruited. One child had to be excluded, becauseof a discrepancy between the basal F level and the MTP test. The demographics, therapy and HPAS status of the 95 enrolled patients arelisted (table 1).Only one child was on oral prednisone at the time of testing. According to the basal F and the PMTP ACTH response, 35 children were classified in asuppressed (F < 83 nmol/l or ACTH < 106 pg/ml [23.5 pmol/l]), 29 in a middle (ACTH 106-319 pg/ml [23.5-70.8 pmol/l]) and 31 in a high (ACTH > 319 pg/ml [70.8 pmol/l]) ACTH response group.
All three SNPs were in Hardy-Weinberg equilibrium (rs242941A/C [suppressed ACTH response group p=0.830;middle/high ACTH response group p=0.580], rs1876828 A/T [suppressed ACTH response group p=0.791;middle/high ACTH response group p=0.481] and rs41423247G/C [suppressed ACTH response group p=0.445;middle/high ACTH response group; p=0.948]).
The rs41423247 (G/C) SNP was inversely associated with HPAS (OR = 0.27 [95% CI 0.06-0.90]). Both rs242941 (A/C) and rs1876828 (C/T) were not associated with HPAS, their OR being 0.99 (95% CI 0.43-2.27) and 0.54 (95% CI 0.05-3.48) respectively. Similarly, only the rs41423247 GC and GC+CC genotypes were inversely associated with the suppressed ACTH response group(OR=0.310 [95% CI: 0.137-0.920] and OR = 0.296 [95% CI: 0.1116 - 0.756]). The homozygous CC genotype just missed statistical significance (OR = 0.101 [95% CI: 0.005-1.908], p = 0.041). The mean √PMTP ACTH of the CC genotype was significantly higher (p = 0.002) than for the GC and GG genotypes (fig 1). The respective means and their 95% CIs were 22.59 (16.18-29.00), 16.00 (13.00-19.00) and 11.82 (10.11-13.63). These correspond to real ACTH mean levels of 553.0 pg/ml (122.9 pmol/l) for the CC, 331.0 pg/ml (73.6 pmol/l) for the GC and 187.5 pg/ml (41.7 pmol/l) for the GG genotype.
The frequency of the guanosine (G) allele (the major allele) of rs41423247 was 77%, while the frequency of the cytidine (C)allele (the minor allele) was 23%. The genotype frequencies for GG, GC and CC were 60, 34 and 6% respectively.
BMI was associated with the CC genotype (table 2) while the heterozygous GC was associated with HPAS, independently of BMI (table 3). Both the homozygous CC and the heterozygous GC genotypes were associated with √PMTP ACTH with CC having a significant larger impact than GC (table 4). The observed ACTH effect was independent of age, sex, height and weight (and hence of BMI). The higher the ACTH response, the greater and more significant the genotype (CC) effect (table 5).
After correcting for age, sex, height and weight, both the additive and the dominant genetic models were the best fit to describe the protective effect of the C alleleon HPAS. The effect size for the additive model was -1.30 (-2.32 – -0.42; p = 0.007) and for the dominant model, the effect size was -1.31 (95% CI -2.39 – -0.33; p = 0.012). Similarly, the effect of the C allele on √PMTP ACTH is best described by an additive (effect size = 4.42 [95% CI 1.77 – 7.07; p = 0.002]) and a dominant model (effect size = 4.79 [95% CI 1.54-8.04; p = 0.005).