Growth and Pubertal Timing in Boys With Adult-diagnosed Celiac Disease: A Population-based Longitudinal Cohort Study.

There are few longitudinal data on whether childhood growth and pubertal timing may be impaired by adult-diagnosed celiac disease (CD). Through school health care records and national registers, we retrieved serial growth measurements on 37,672 Swedish boys born in 1945 to 1961, out of whom 72 (0.2%) were clinically diagnosed with CD as adults. Boys with, versus without, adult-diagnosed CD exhibited no appreciable mean differences in body mass index (BMI, kg/m) and height (cm) at ages 8 or 20 to 21 years (childhood BMI, 15.9 [CD] vs 15.7 [comparators]; childhood height, 129.1 [CD] vs 128.6 [comparators]; adult BMI, 21.3 [CD] vs 21.4 [comparators]; adult height, 180.7 [CD] vs 180.4 [comparators]). Neither did we observe any between-group differences in growth development during puberty nor in the timing of pubertal growth spurt (all P values ≥0.30). Conclusively, in this population-based longitudinal study, boys with adult-diagnosed CD had similar growth and pubertal timing as their peers.


What Is New
Through school health care records and national registers, we analyzed serial growth measurements on more than 37,000 Swedish boys. Compared with population-based comparators, boys with adult-diagnosed celiac disease exhibited no appreciable differences in body mass index and height at ages 8 or 20 to 21 years. There were also no significant differences in growth development during puberty or timing of pubertal growth spurt. C eliac disease (CD) is an autoimmune disease where gluten intake causes a flattening of the gut mucosa (ie, villous atrophy). It is a common, yet underdiagnosed condition, whereas approximately 1% of children worldwide are affected by CD (1), most will, due to a scarcity of symptoms, be left undiagnosed until adulthood (2).
Commonly attributed to malnutrition from villous atrophy, poor growth, and pubertal delay may complicate untreated childhood CD (1). To what extent individuals with adult-diagnosed CD have impaired growth is less clear. Based on cross-sectional studies of patients at tertiary care centers (as detailed in the Supplemental Digital Content, http://links.lww.com/MPG/B795) (3), there is, however, a notion that adult-diagnosed CD could impair childhood growth and pubertal timing. Few population-based studies have longitudinally examined this relationship.
Therefore, using longitudinal data on growth, we aimed to determine whether adult-diagnosed CD was associated with growth impairment and pubertal delay as compared with general-population comparators.

METHODS
This study consists of 37,672 boys born in 1945 to 1961 who completed elementary/secondary school in Gothenburg, Sweden, with growth data in centrally archived school health care records. To enable register linkages, participation was restricted to individuals with correct personal identity numbers (4) and residing in Sweden as adults (Fig. 1, flowchart). This study was approved by the ethics committee at the University of Gothenburg, Sweden.

Longitudinal Data on Growth
School health care records included data on birth weights and prospectively recorded, direct measurements of weight and height until the end of secondary school. Through register linkages, we also collected self-reported adult height recorded in the nationwide passport register and directly measured height and weight at mandatory military conscription (age 18-20 years). Using all paired height and weight measurements at 6.5 to 9.5 and 17.5 to 22.0 years of age, we performed linear regression models to estimate body mass index (BMI, kg/m 2 ) at age 8 and 20 years, respectively. With a similar approach, we used all height measurements at ages 6.5 to 9.5 years to estimate the average height at age 8 years. Longitudinal growth typically ends at age 18 years, but for some, in particular men, there may be residual growth several years after that. Therefore, for adult height, we used self-reported heights in the passport register after 21 years of age (79.9% of the cohort). For those without such data (20.1% of the cohort), height at age 21 years was estimated through age adjustment of height measurements at ages 17.5 to 22.0 years from school health care and military conscription. Pubertal growth was in this study defined as the average change in BMI and height from 8 to 20 respectively 21 years of age. Through curve-fit of serial height measurements we estimated age at peak height velocity, that is, the adolescent period of fastest upward growth, as a marker of pubertal timing (5).
This study was restricted to men because Sweden did not during the study period have mandatory female military conscription, which prevented us from retrieving data on adult BMI and height for women.  (7). Reporting to this register is mandatory and it covers virtually all nonprimary health care in Sweden (7). For the purpose of this study, we predefined adult-diagnosed CD as a first diagnostic code of CD recorded at the age of 21 years or later. There is no mass screening for CD in Sweden, and hence in this sample CD diagnosis was likely mainly prompted by symptoms or signs of the disease.

Diagnosis of Celiac Disease
Although questioned by some (8,9), longitudinal screening studies have recently shown that serological evidence of CD usually develops in the first 5 to 10 years of life (10,11). Until the 1980s, before a wider use of serology tests, CD was, however, largely unrecognized as a disease in Sweden (12), and elsewhere (13). Hence, this birth cohort provides data on the natural history of childhood and adolescent growth in adult-diagnosed CD.

Other Data
Among available data (5), we preselected adjustment variables that may be associated with growth and CD diagnosis. Information on country of birth and adult education level was retrieved from the government agency Statistics Sweden as detailed in the Supplemental Digital Content (http://links.lww.com/MPG/B795). Country of birth was categorized as Sweden, when the participant and both of his parents were born in Sweden, or else categorized as not Sweden. From 7 predefined education categories, we grouped data into low (¼elementary school), middle (¼secondary education), and high education level (¼postsecondary education).

N=50,020
No data on BMI in childhood or adulthood 10,119 FIGURE 1. Flowchart of study formation. A The personal identity number is assigned to all Swedish residents and enables register linkages for this study. BMI ¼ body mass index.

Statistical Analyses
We used analysis of covariance (ANCOVA) to compare average BMI and height in childhood (age 8 years), young adulthood (age 20-21 years), and during puberty (average difference in BMI and height from 8 to 20 respectively 21 years of age) in men with versus without CD diagnosis; ANCOVA was also used to test for differences in the age at peak height velocity. ANCOVAs were adjusted for birth year (ie, secular trends), country of birth, and adult education level (5). Chi-square test, and when appropriate Fisher exact test, was used to test for differences in background characteristics according to adult-diagnosed CD. Further details on the used data sources are provided in the Supplemental Digital Content (http://links.lww.com/MPG/B795).

Power Estimation
Assuming a common standard deviation across groups and a 2sided significance level of 0.05, we estimated that our sample would provide 80% statistical power to detect the following minimum mean differences between men with versus without CD: 1.3 cm difference in adult height, 0.5 kg/m 2 in adult BMI and 0.37 year (135 days) age difference in timing of peak height velocity.

RESULTS
The median calendar year of birth was 1953 (range, 1945-1961). By December 31, 2016, 72 men (0.2%) had been diagnosed with CD at a median age of 52.1 years (range, 32.5-70.4 years). Characteristics were largely similar between men with versus without adult-diagnosed CD ( Table 1). The participants had, on average, 1.9 paired height and weight measurements between 6.5 and 9.5 years of age and 1.3 paired height and weight measurements between 17.5 and 22.0 years of age.
The average age at peak height velocity, that is, pubertal timing, was 13.9 years in boys with adult-diagnosed CD, compared with 14.1 years in general-population comparators, yielding a P value of 0.30 for the difference of means when adjusting for the potential confounding effect of birth year, country of birth, and

DISCUSSION
This is likely the first population-based study using serial growth measurements as a mean to study if childhood growth and/or pubertal timing differ between individuals with versus without adult-diagnosed CD. We found similar growth and pubertal timing in boys with versus without adult-diagnosed CD. There may be several explanations to this null finding.
First, adult-diagnosed CD may not significantly impair childhood growth. Among 9 cross-sectional studies in this field, only 5 studies have shown significantly lower height and/or BMI in men with adult-diagnosed CD compared with comparators (prior studies are detailed in the Supplemental Digital Content, http:// links.lww.com/MPG/B795) (3,(14)(15)(16)(17). Previous research has also mostly been limited to tertiary care settings, typically including those suffering from more severe disease than the average patient with CD, which may lead to exaggerated risk estimates. Furthermore, despite recommendations to test children with pubertal delay for CD (1), we are, besides case series (18), unaware of earlier research on pubertal timing in CD. Our results are also in line with human perfusion studies suggesting that (19), although CD impairs the absorptive function of the proximal small intestine, the distal part of it can adapt to such damage with increased absorptive capacity, which may conceivably prevent malabsorption and poor growth in the context of adequate food access.
Second, despite our large sample size, we cannot rule out that our null findings may be related to a type 2 error (ie, to erroneously accept a false null hypothesis). The risk of committing a type 2 error is associated with factors impairing the precision of our analysis, including imprecise growth estimates. Although our approach will most likely, on average, yield correct growth estimates, we cannot rule out misclassification of self-reported adult height related to erroneous recall or that direct measurements of growth may have been incorrectly recorded. Still, using prospectively collected data such misclassification should not differ according to the future diagnosis of CD.
Third, although we found no significant differences in BMI and height at 8 or 20 to 21 years of age, adult-diagnosed CD may still have transient effects on childhood growth or impair growth metrics not captured by this study. So, although caution should be exercised when interpreting our results, they suggest that the absence of significant growth impairment in childhood should not preclude CD testing in men suspected of having the disease.
Strengths of this study include its population-based design and prospectively collected longitudinal data on growth, which reduce the risk of bias introduced by the selection of participants or from erroneous recall, hampering previous studies in this field (3). Another strength is our use of serial height measurements to estimate the age at peak height velocity which is an indirect, but objective assessment of pubertal timing (5). To our knowledge, there is no better way to estimate pubertal timing when data on Tanner stage markers of pubertal timing are missing. In fact, pubertal timing by age at peak height velocity in boys seems to be highly correlated across Tanner stage markers of pubertal timing (20).
A limitation of this study is our lack of information on the timing of seroconversion of CD. Therefore, we cannot rule out that individual participants may have had a late-onset rather than a latediagnosed (ie, delayed) adult CD. If CD may truly develop for the first time in adulthood, there is in that scenario little reason to believe that childhood growth would be affected. Reports on when in life CD typically develops have, however, been highly inconsistent; although large birth cohort studies have shown that CD usually develops in the first 5 to 10 years of life (10,11), their follow-up period did not include adulthood, and cross-sectional data from periodic CD screening of adults have indicated that serological evidence of CD may develop late in life (8,9). The latter-mentioned studies, however, lacked the means to confidentially rule out that the seroconversion of CDspecific autoantibodies actually equaled incident disease rather than transient/fluctuating autoantibody positivity. Another limitation of this study is that it was restricted to men with clinically diagnosed CD. Hence, we do not know to what extent these results can be generalized to women or to individuals with screen-detected disease, who may have milder disease presentations than those with clinically diagnosed CD. A related limitation is our lack of clinical data, including how CD was diagnosed and information on the type and duration of manifestations preceding the diagnosis of CD. Data on clinical manifestations are important because the impact on growth from CD may vary according to the clinical phenotype of the disease, for example, be related to whether the patient also suffers from other malabsorptive symptoms or not.
Conclusively, in this population-based study, boys with adultdiagnosed CD had similar growth and pubertal timing as their peers. Our results should be corroborated in large-scale, longitudinal screening studies for CD with the means to precisely disentangle the potential effects of late-diagnosed from late-onset adult CD.