Racial/Ethnic Disparities in U.S. Pediatric GH Treatment

Abstract

BACKGROUND/AIMS:

To compare racial/ethnic proportions of GH-registry subjects to those expected, and secondarily, to assess racial/ethnic differences in subject characteristics at GH-treatment initiation.

METHODS:

Race/ethnicity-based expected frequencies of height <−2.25 SD were determined by applying relative risks for short stature, calculated from a regional population of 189,280 pediatric primary-care patients, to U.S. Census data, and compared by Chi-squared test to racial/ethnic proportions of U.S. subjects enrolled in the Pfizer International Growth Study (KIGS). Characteristics of white and black subjects at GH-treatment initiation were presented as medians and compared by Wilcoxon rank-sum test (significant P-value <0.01).

RESULTS:

White subjects exceeded expected frequency (63%) for all indications (83%) and each separately, ranging from 73% for congenital GH deficiency (GHD) to 85% for idiopathic short stature (ISS) (P<0.001). Compared to white subjects, black subjects treated for idiopathic GHD had greater height deficits relative both to the population (−2.97 vs −2.56 SD) and to their mid-parental heights (−2.47 vs −1.89 SD), lower stimulated GH peak levels (4.9 vs 6.0 ng/ml), and lower birth weights (−0.86 vs −0.48 SD). Black subjects with congenital GHD had lower stimulated GH peaks (2.1 vs 3.2 ng/ml) and started GH treatment at younger ages (2.9 vs 4.8 yrs), while those with acquired GHD had lower birth weights (−1.12 vs −0.08 SD). Male predominance did not differ by race for any or all indications.

CONCLUSION:

Over-representation of white children among those receiving GH treatment in the U.S. KIGS registry reflects racial/ethnic treatment biases, not just differences in growth rates.

Introduction

White children are over-represented relative to the general population from the earliest report of U.S. GH registries [1] and among those seeking subspecialist care for short stature [2]. The National Health and Nutrition Examination Survey (NHANES) found that among U.S.-born non-Hispanic children from birth cohorts of 1942–2002, black children were on average taller than white children in childhood and early adolescence [3,4]. This was related to a faster tempo of growth starting at age 3 years, such that the average height differences between black and white children were +1 cm for boys age 2–10 years, +1.5 cm for boys age 11–13 years, +1.2 cm for girls age 2–10 years, and +1.8 cm for girls age 11–13 years. However, white teens caught up to black teens in height and ended up slightly taller, with average height differences between black and white teens age 14–19 years, <−1 cm for boys and −1.5 cm for girls. Associated with this growth pattern is an earlier onset of puberty by black children, when measured by both median age of entry and mean age in each Tanner stage [5].

A recent regional study of 189,280 pediatric primary-care patients [6] confirmed lower odds of short stature among black than white children and adolescents. It also showed, similar to earlier studies [7–10], higher odds of short stature among Asian and Hispanic children compared to white. This regional study allowed quantification of the effect of race and ethnicity on the prevalence of short stature. Thus, we now sought to compare racial/ethnic proportions of GH-registry subjects to those expected based on U.S. Census and differences in growth rates, and secondarily, to assess racial/ethnic differences in other subject characteristics at initiation of GH treatment.

Subjects and Methods

Determination of expected frequencies of short stature by race/ethnicity

The heterogeneous pediatric primary-care population was drawn from the 28 practices in Pennsylvania, New Jersey and Delaware affiliated with a tertiary pediatric hospital. The Children’s Hospital of Philadelphia institutional review board approved this study, and determined that the criteria under 45 CFR 46.116(d) were met for waiver of informed consent. The patient population was described previously in detail [6]. Race/ethnicity-based odds ratios for height <−2.25 SD (i.e. short enough to meet FDA-approved criteria for GH treatment of idiopathic short stature [ISS]) were calculated from the 145,710 subjects with complete data by multivariable logistic regression modeling, controlling for gender, age, history of premature birth, BMI Z-score, insurance type, primary-care practice location (urban vs non-urban), and number of well visits during the study period [6]. The odds ratios were converted to race/ethnicity-based relative risks using the formula for converting adjusted odds ratios from logistic regression into crude estimates of adjusted risk ratios suggested by Zhang and Yu [11]. Race/ethnicity-based expected frequencies of height <−2.25 SD were then calculated by applying the risk ratios for short stature to racial/ethnic percentages from the 2011 U.S. Census [12].

KIGS data collection

The Pfizer International Growth Study (KIGS) database, established in 1987 and introduced in the US. in 1996, is the world’s largest post-marketing surveillance database of pediatric GH with data from 83,803 patients (277,264 patient-years) from 52 different countries [13]. The KIGS registry is an international database developed with the main objective of documenting long-term outcomes and safety of Genotropin (Pfizer Inc, USA) in children with short stature. Pediatric patients and their legal guardians were informed of and consented/assented to participate in the survey, which permits anonymous use of the data in compliance with privacy guidelines. Institutional review board approval was obtained per standards of the place and time. As of October, 2012, KIGS was closed for further data collection and therefore transitioned to a static database.

This study focused on the U.S. cohort of the KIGS database. Included in this analysis were data from all U.S. subjects treated for ISS, idiopathic GHD (IGHD), congenital GHD, acquired GHD, or born small-for-gestational age (SGA). Diagnoses were made by individual KIGS investigators according to the KIGS Etiology Classification List. For purposes of this analysis, diagnoses were reviewed and reclassified, if indicated (36 subjects with GHD were reclassified to ISS), to comply with the generally accepted threshold of peak GH levels on provocative testing of ≤10 ng/dl for all subjects with GHD and >10 ng/dl for all subjects with ISS. Key variables of interest were race/ethnicity (as reported by the investigator on CRD and based on discussions with patient/parent[s]), gender, birth weight Z-score, gender adjusted mid-parental height Z-score, maximal GH peak on provocative testing, IGF-I Z-score, diagnosis, and at the time of GH therapy initiation: age, height Z-score, height Z-score minus mid-parental height Z-score, weight and BMI Z-scores, and pubertal status.

Statistical analyses of the KIGS data

Standard descriptive statistics were tabulated for discrete (e.g., frequencies and percentages) and continuous {e.g. mean and standard deviation, median and interquartile range) subject characteristics. The racial/ethnic proportions of U.S. subjects enrolled in the KIGS database, for all combined and individual indications, were compared by Chi-squared test to race/ethnicity-based expected frequencies of height <−2.25 SD. Characteristics of the white and black subjects at initiation of GH treatment were presented as medians (10^th and 90^th centiles) and compared by Wilcoxon rank-sum test (significant P-value <0.01). Gender ratios of each and combined indications were compared by Chi-squared test. Statistical Analysis System version 9.2 (SAS Institute Inc, North Carolina, USA) was used for all statistical analyses.

Results

Of the 10,229 U.S. subjects in the KIGS database, 8,173 (80%) had race/ethnicity data recorded and were included in this study. White subjects exceeded the expected frequency of 63% (per U.S. Census data and the risk ratios for short stature, Table 1) for all indications (83%) and each separately, ranging from 73% for congenital GHD to 85% for ISS (every P<0.001; Table 1 and Figure 1). The congenital GHD group included 228 subjects with multiple pituitary hormone deficiencies (MPHD) and 242 with isolated GHD. Because short stature is more prevalent among Asian and Hispanic youth and adults in U.S. [6–10], their heights are more likely to be attributed by clinicians to familial short stature. Because treating familial short stature with GH is controversial, we will focus the rest of our comparisons to the black and white subjects in U.S. KIGS.

Table 1: Racial/ethnic proportions in U.S. KIGS compared to expected rates based on the 2011 U.S. Census and differences in growth patterns.

Race/ Ethnicity	RAW DATA		COMPA- RATOR	U.S. KIGS DATA
	Relative risk for Ht <−2.25 SD	US Census 2011	Expected frequency	All combined	ISS	IGHD	Congenital GHD	Acquired GHD	SGA
White	1	63%	63%	83%	85%	84%	73%	83%	80%
Black	0.66	13%	9%	4%	3%	3%	11%	4%	6%
Asian	1.72	5%	9%	2%	2%	2%	2%	2%	3%
Hispanic	1.86	17%	32%	7%	7%	7%	8%	8%	8%
Other	0.85	2%	2%	3%	3%	3%	4%	3%	3%
n				8173	1097	4810	470	563	1233
Chi-squared P value vs comparator				P<0.001	P<0.001	P<0.001	P<0.001	P<0.001	P<0.001

Figure 1. White children and adolescents are over-represented in U.S. KIGS relative to expected rate.

Registry proportions compared to expected rate by Chi-squared test. ISS, Idiopathic short stature; IGHD, idiopathic GH deficiency; SGA, born small-for-gestational age.

Comparing the white and black subjects in U.S. KIGS (Table 2), black subjects in the idiopathic GHD group had greater height deficits relative to the population and their mid-parental heights, lower GH peaks on provocative testing, and lower birth weights. Black subjects with congenital GHD had lower GH peaks on testing and started GH treatment at younger ages, while those with acquired GHD had lower birth weights. Baseline characteristics did not differ significantly among those treated for ISS or SGA.

Table 2. Baseline characteristics of black vs white subjects in U.S. KIGS by indication.

Traits compared included sex, birth weight Z-score, gender adjusted mid-parental height Z-score, maximal GH peak on provocative testing, IGF-I Z-score, diagnosis, and at the time of GH therapy initiation: age, height Z-score, height Z-score minus mid-parental height Z-score, weight and BMI Z-scores, and pubertal status. Only those with significant differences (P<0.01) are shown.

Indication	Black subjects			White subjects			P-value
	n	Median	10th to 90th percentiles	n	Median	10th to 90th percentiles
Idiopathic GHD
Height SDS Prader	168	−3.0	−4.9 to −1.8	4063	−2.6	−3.8 to −1.4	0.0001
Height - Mid Parental Ht SDS	142	−2.5	−4.9 to −0.5	3806	−1.9	−3.6 to −0.3	0.0001
GH peak on stim test (ng/ml)	168	4.9	1.1 to 9.1	4063	6.0	1.9 to 9.2	0.0014
Birth weight (SDS)	135	−0.9	−2.6 to 1.1	3393	−0.5	−2.0 to 0.7	0.0021
Age at GH start (yr)	168	10.9	3.7 to 15.8	4063	11.5	4.4 to 14.9	0.7069
Congenital GHD
Height SDS Prader	54	−3.0	−5.8 to −0.4	345	−2.6	−4.9 to 0.0	0.2347
Height - Mid Parental Ht SDS	6	−2.9	−5.9 to −0.4	285	−2.5	−5.1 to 0.2	0.2493
GH peak on stim test (μg/L)	24	2.1	0.5 to 3.9	198	3.2	0.6 to 9.7	0.0050
Birth weight (SDS)	44	−0.5	−2.4 to 0.9	288	−0.2	−1.6 to 1.3	0.0893
Age at GH start (yr)	54	2.9	0.2 to 13.1	345	4.8	0.8 to 13.3	0.0037
Acquired GHD
Height SDS Prader	24	−2.4	−3.7 to −1.2	469	−2.0	−3.6 to −0.3	0.2683
Height - Mid Parental Ht SDS	19	−2.3	−3.8 to 1.1	406	−1.8	−3.8 to −0.2	0.4173
GH peak on stim test (ng/ml)	16	2.0	0.5 to 6.1	334	2.4	0.4 to 9.0	0.6534
Birth weight (SDS)	16	−1.1	−2.7 to 1.0	341	−0.1	−1.3 to 1.6	0.0058
Age at GH start	24	11.8	8.3 to 17.3	469	11.5	7.0 to 15.8	0.3601
SGA
Height SDS Prader	80	−3.0	−4.5 to −2.3	981	−3.0	−4.4 to −1.8	0.3609
Height - Mid Parental Ht SDS	72	−2.1	−4.6 to −0.2	868	−2.2	−4.2 to −0.6	0.4013
GH peak on stim test (ng/ml)	25	17.2	7.8 to 40.0	253	15.90	7.1 to 29.3	0.3697
Birth weight (SDS)	71	−2.4	−4.1 to 0.4	907	−2.3	−3.8 to −0.5	0.7012
Age at GH start (yr)	80	8.3	3.0 to 13.1	981	8.1	3.1 to 13.3	0.9496
ISS
Height SDS Prader	30	−2.6	−3.4 to −2.1	929	−2.7	−3.7 to −1.6	0.6604
Height – Mid Parental Ht SDS	26	−1.8	−3.0 to −0.5	876	−1.8	−3.3 to −0.3	0.7626
GH peak on stim test (μg/L)	30	14.2	10.4 to 23.2	929	14.0	10.7 to 24.0	0.7960
Birth weight (SDS)	22	−1.1	−3.0 to 0.3	782	−0.6	−2.2 to 0.9	0.1102
Age at GH start (yr)	30	11.7	8.0 to 14.6	929	11.8	6.5 to 14.8	0.8581

Male predominance did not differ by race/ethnicity (Table 3). Congenital GHD, the indication least dependent on a phenotype of short stature for diagnosis, had the least male predominance for both races. The combined effects of gender- and race/ethnicity-based disparities were pronounced for the ISS indication (Figure 2). In absolute numbers, with male predominance superimposed on white predominance, of the 1,097 subjects with ISS in U.S. KIGS, only 6 were black females.

Table 3. Male predominance did not differ by race for any or all of the indications.

Indication	White males		Black males		Chi-Square P
	n	%	n	%
ISS	704	76	24	80	0.5948
IGHD	3034	75	132	79	0.2540
Congenital GHD	196	57	28	52	0.4946
Acquired GHD	285	61	16	67	0.5632
SGA	657	67	62	78	0.0527
All combined	4876	72	262	74	0.4732

Figure 2. White boys are over-represented and black girls are particularly under-represented among U.S. KIGS subjects treated for idiopathic short stature (ISS).

Ratios of frequencies, by race and gender, of U.S. KIGS subjects treated for ISS to those expected based on population frequencies of height <−2.25 SD [6].

Discussion

In summary, white subjects in the U.S. cohort of the KIGS database exceeded the expected frequency for all indications and each separately, least for congenital GHD and most for ISS. The former indication is least related to short stature, while the latter involves treatment primarily for the psychosocial aspects of short stature. Comparing the black to white subjects in the U.S. cohort of KIGS, the lower GH peaks on provocative testing for both the idiopathic and congenital GHD groups, the greater height deficits relative to the population and their mid-parental heights for the idiopathic GHD group, and the younger age at initiation of GH treatment for the congenital GHD group, all suggest that the black subjects were more severely affected at presentation. Taken altogether, these findings suggest racial/ethnic biases in U.S. pediatric GH treatment. Statistically significant racial differences were not found in severity among KIGS subjects treated for ISS, which is the most subjective indication, but this may be due to the small numbers of black children treated for this indication. Male predominance did not differ by race.

In the absence of data on race/ethnicity-specific prevalence of the various growth disorders in the U.S. population, one must turn to the relative proportions of the various racial/ethnic groups in the U.S. population as comparators to the proportions seen in the KIGS database. However, the NHANES studies found race/ethnicity-based differences in growth rates among U.S. children and adolescents [3–5]. Because all groups are evaluated clinically based on the same CDC growth charts, the faster growth rates reported in NHANES for black children may mask under-lying growth problems in this group more so than for white children. Due to this phenomenon, to avoid over-attributing racial differences in GH treatment to disparities, we corrected population race/ethnicity proportions for relative risk of height <−2.25 SD before comparing to race/ethnicity proportions in KIGS. While such a correction may be more directly applicable to indications like ISS and IGHD than congenital GHD or SGA, not implementing this correction (i.e. comparing KIGS to the relative proportions of the various racial/ethnic groups in the U.S. population) would not change any of our findings apart from making the apparent disparities that much bigger.

Where could such racial/ethnic biases come from? Race is often a surrogate for socioeconomic status and access to healthcare. This is illustrated by a comment in a parent focus group exploring parental concerns that impact their medical decision-making regarding a child’s short stature; “There’s so many other issues to be concerned about in the African-American community that I don’t think that…height and trying to address that through medicine would come at the top of the tier here…on the other side, a family that may be more affluent, uh, particular a Caucasian family, maybe that would be something that they would consider.” [14]

However, digging deeper, there may be other factors at play as well. In a survey of 1820 parents of pediatric primary-care patients, when asked “how short is too short for an adult” male or female, the median height provided by white parents was 2 inches taller for both men and women than that provided by parents of all other racial and ethnic backgrounds [15]. Because short stature is more prevalent in U.S. among Asian and Hispanic Americans, their reported shorter acceptable height thresholds may reflect personal experience or the tendency to use oneself as a benchmark. However, this cannot explain the difference in height thresholds between black and white parents, whose adult heights tend to be more similar. Further, respondent traits found to be significant main effects in a parsimonious model of the reported acceptable height threshold were parents who were female, taller, white, wealthier, from non-urban primary-care practices, and personally concerned about their own child’s height [15]. The fact that race remained a significant explanatory variable in a model that included measures of socioeconomic status and access to healthcare (highest level of education completed, current employment status, total household annual income, insurance type, and primary-care practice location) suggests that race is not merely a confounder; there are racial group differences in height-related perceptions that warrant further investigation.

Another possible factor is an interaction between race and religion. In both the qualitative parent groups and the quantitative survey portions of a mixed-methods study of height-related medical decision-making by parents of pediatric primary-care patients, black parents were more likely than white parents to endorse a religious objection to pursuing medical height enhancement. They were more likely to express the belief that one’s height is “the way God made you” and thus, should be accepted [14].

Gender-based disparities in U.S. pediatric GH treatment have been reported previously [6,16], as have gender-based disparities among patients seeking subspecialist care for short stature [17], diagnostic screening tests of the GH axis by primary-care clinicians of their patients with growth faltering [18], and the likelihood of prescribing GH by pediatric endocrinologists [19]. Thus, it was interesting to see how the gender-based disparities interacted with racial differences in our current study. The previously reported 2:1 to 3:1 male:female ratios in U.S. pediatric GH treatment [6] were seen in the U.S. cohort of the KIGS-database, similarly for both black and white subjects for each and combined indications. Thus, the social pressures for tallness on males in our society seem to cut across racial lines. Further, for black and white subjects alike, congenital GHD was the indication with the smallest gender gap. Congenital GHD is the indication least dependent on a phenotype of short stature for diagnosis; often those patients present with neonatal hypoglycemia, microphallus in boys, and/or midline or central nervous system defects.

The voluntary ascertainment of a registry study design introduced several possible limitations. Black patients may be less likely to obtain subspecialist care at a center participating in the KIGS study. Given the American history of racially based research maleficence [20], black patient-families may be less likely to consent to participate in medical research like the KIGS registry, or they may be less likely to report their racial/ethnic identity when asked. Other potential factors may include black patient-families not trusting doctors as much as they trust their religious leaders; too few same-race doctors, especially endocrinologists; and as seen in school-based or other non-referred populations, not all families of short children are interested in seeking height-related medical care. All of these possibilities could lead to under-representation of black patients in the current study. Interestingly, the percentage of KIGS subjects with missing race/ethnicity data was greater in the U.S. cohort (20%) than the rest of the world (5%). However, racial disparities also were seen in U.S. studies of subspecialist care that were not dependent on voluntary ascertainment. In a study of patients with growth faltering from 4 urban pediatric primary-care practices, black children were less likely to see a subspecialist than non-Hispanic white children, and the racial disparity was greater for endocrinology than gastroenterology appointments [18]. Likewise, white children were over-represented (91%) relative to the local general population in a study of parents of children seeking subspecialist care for short stature in a mid-sized, ethnically diverse U.S. city [2]. Thus, these non-voluntary ascertainment-based studies support the notion of racial disparities in subspecialist referrals for short stature, which would then lead to racial disparities in GH treatment (GH is primarily prescribed by endocrinologists), as found in our current study. However, the 20% of U.S. KIGS registry subjects with missing race/ethnicity data could not be analyzed and this also could have contributed to the study’s findings.

In conclusion, over-representation of white children among those receiving GH treatment and enrolled in the U.S. KIGS registry suggests racial/ethnic treatment biases and not just differences in growth rates. Growth is a vital sign of child health and should receive equal import for all children and adolescents [21]. Thus, more worrisome than GH treatment itself is the possibility that social biases can lead to missed diagnosis of underlying disease in short girls and children of racial/ethnic minorities, while promoting over-zealous treatment of short, primarily white boys with an expensive medication.