br Methods br Results We enrolled children years old

Methods
Results We enrolled children≤10years old who presented with TB to King Edward VIII Hospital in Durban, South Africa, between May 2012 and March 2013. We enrolled 19 boys and 11 girls, whose clinical characteristics are shown in Fig. 1. The figure shows that most children were one year-old or less. All children had chest X rays performed, 19 (63%) had consolidation and 4 (13%) had cavitation, thus most children had pulmonary TB. Fig. 2 shows Z-scores and demonstrates that up to 30% of children were malnourished, though only one was severely malnourished (Z-score standard deviation<−3). Isoniazid acetylation follows a ping-pong Bi-Bi mechanism described by the standard Michaelis-Menten relationship for enzyme rate reaction rate (v) and substrate concentration [S] (Michaelis and Menten, 1913; Weber and Cohen, 1968): We generated Km and Vmax values for each child in ADAPT based on this relationship, with estimates shown in Fig. 3A–B. The mean population estimate for Km was 1.69mg/L, which translates to 0.012μM, while that of Vmax was 30.11mg/h, which translates to 0.06μM/s. Fig. 3A-B show that Km was tightly distributed around the mean, but there was greater variability in the Vmax. The isoniazid elimination rate (KINH) is shown in Fig. 3C. Fig. 3D shows the elimination ratio VAINH/fINH, calculated using the fraction of isoniazid metabolized (fINH), and the volume of distribution of metabolite VAINH. NAT2 sequencing results, including the T-5224 pairs and SNP calls for each allele, are shown in Extended Dataset 1. We identified 17 different alleles, including four reference high-activity NAT2*4 alleles. The most frequent genotypes were NAT2*5B/NAT2*5B encountered in 6/30 (20%) children, and NAT2*5C/NAT2*12C encountered in 6/30 (20%) children; the former genotype is commonly reported in literature to confer slow acetylator status, while the latter is predicted to confer the intermediate acetylator status (McDonagh et al., 2014). On the other hand, homozygosity for the rapid acetylator NAT2*4 reference allele (NAT2*4/NAT2*4 genotype) was encountered in 2/30 (7%) of enrolled children (Table 1). As a first step, we employed standard frequentist statistics to compare the distributions of Km, Vmax, and elimination rate between different inferred genotypes and predicted acetylation status. Table 2 shows no significant differences in mean values. Compared the distributions of the same enzyme function parameters, as well as measures of body size, age, and dose by inferred acetylator status, shown in Table 2. The only statistically significant finding was that the isoniazid dose administered was lower among rapid acetylators. Next, we used MARS algorithm in order to identify predictors of Vmax. Table 3 shows that dose (BF1), genotype (BF5), isoniazid peak concentration (BF7), and age (BF10) were identified as the best predictors of Vmax. The table illustrates the nature of MARS output, which is the form of basis functions (BF). BF1 is a simple hinge function “max (0, Dose–190)” which means that the value of the expression is zero for all doses (mg) up to 190 (i.e., for all doses satisfying “Dose minus 190” less than zero), as illustrated in Fig. 4A. The dose of 190mg is at the hinge. The variable importance score of 100% means that “Dose” is the most important variable for Vmax. Isoniazid peak concentration was also an important predictor (BF7), with a variable importance score of 90%, very close to that of “Dose”; thus both dose and peak concentration likely reflect the same thing, which is effect of isoniazid concentration on Vmax. BF7 means that isoniazid concentration had no effect on Vmax until the peak concentration achieved the threshold value of 8.83mg/L, at the hinge. Table 3 shows that the peak concentration (BF7) interacted with BF5 (genotype). However, the BF5 interaction with BF7 only applied to NAT2 genotypes selected as subset 1, which means that there is staggered or stratified effect so that relationship applied to some genotypes and not others. Specifically, BF5 comprised NAT2 genotypes that were grouped by MARS, which had a variable importance score of 90%, similar to that of isoniazid concentration. Of note, for MARS analyses we did not pre-categorize different genotypes by acetylator status (slow or rapid acetylator), since predictors do not take a pre-specified form but are derived from the data.Table 3 further shows that, in addition, the child\'s age also had an effect on Vmax as well, as reflected by BF10; the impact however was less than half that of either genotype or concentration. Fig. 4B shows the singular contribution of age; the effect of age starts at zero but reaches a maximum at 1935days (about 5.3years) when it reaches a plateau, and remains flat into the next decade. Overall, the relation between Vmax and these four predictors was given by the equation: