HEPATIC ABNORMALITIES IN YOUTH WITH TURNER SYNDROME

Abstract

Background & Aims:

Liver disease in children with Turner Syndrome (TS) is poorly understood relative to associated growth, cardiac, and reproductive complications. This study sought to better characterize hepatic abnormalities in a large national cohort of youth with TS.

Methods:

Using electronic health record data from PEDSnet institutions, 2,145 females with TS were matched to 8,580 females without TS on 8 demographic variables. Outcomes included liver enzymes (AST and ALT) stratified as normal, 1–2 times above the upper limit of normal (ULN), 2–3 times ULN, and >3 times ULN, as well as specific liver disease diagnoses.

Results:

Fifty-eight percent of youth with TS had elevated liver enzymes. Patients with TS had higher odds of enzymes 1–2 times ULN (OR: 1.7, 95% CI: 1.4–1.9), 2–3 times ULN (OR: 2.7, 95% CI: 1.7–3.3), and >3 times ULN (OR: 1.7, 95% CI: 1.3–2.2). They also had higher odds of any liver diagnosis (OR: 2.4, 95% CI: 1.7–3.3), fatty liver disease (OR: 1.9, 95% CI: 1.1–3.2), hepatitis (OR: 3.7, 95% CI: 1.9–7.1), cirrhosis/fibrosis (OR: 5.8, 95% CI: 1.3–25.0), and liver tumor/malignancy (OR: 4.8, 95% CI: 1.4–17.0). In a multinomial model, age, BMI, and presence of cardiovascular disease or diabetes significantly increased the odds of elevated liver enzymes in girls with TS.

Conclusions:

Youth with TS have higher odds for elevated liver enzymes and clinically significant liver disease compared with matched controls. These results emphasize the need for clinical screening and additional research into the etiology and treatment of liver disease in TS.

LAY SUMMARY

Turner Syndrome, a chromosomal condition in which females are missing the second sex chromosome, is often associated with short stature, infertility, and cardiac complications. Liver abnormalities are less well described in the literature. In this study, nearly 60% of youth with TS have elevated liver enzymes. Furthermore, patients with TS had a diagnosis of liver disease more often than patients without TS. Our results support the importance of early and consistent liver function screening and of additional research to define mechanisms that disrupt liver function in pediatric TS females.

INTRODUCTION

Turner Syndrome (TS) is a genetic disorder in which 1 in approximately 2,000–2,500 phenotypic females are missing the second sex chromosome. Medical complications associated with TS include short stature, infertility, and cardiac conditions.¹ The impact of the missing sex chromosome on the liver has been understudied, especially in the pediatric population, due to the rarity of TS and higher-priority cardiovascular disease that is responsible for half of a 3-fold increase in mortality reported in TS.² Still, there are multiple reports of unexplained liver biochemical abnormalities in the population, with a wide range of reported prevalence. Calanchini et al. found elevated liver enzymes in 49.6% of 125 adult women with TS. In this study, likelihood of elevation differed with age, height, triglyceride levels, and karyotype.³ In a comprehensive review by Roulot et al., the prevalence of liver test abnormalities in adolescent and adult patients was estimated to be 20–80% depending on age, with older patients being at highest risk.⁴ Wójcik assessed 100 adolescent patients with TS and found that 34% had liver enzyme abnormalities, concluding that elevated enzymes are common but not associated with obesity or hormone replacement therapy (HRT).⁵ Another study reported elevated alanine aminotransferase (ALT) in 11% of patients but suggested more comprehensive study of younger patients based on limited data revealing consistent elevation of at least one liver enzyme.⁶

The primary objectives of this study were (1) to compare prevalence of liver enzyme abnormalities and liver diagnoses in TS patients and controls in a large pediatric cohort and (2) to determine the risk factors for these outcomes within the pediatric TS population. The most recent clinical practice guidelines recommend annual monitoring of liver function tests throughout the lifespan beginning at age 10, as well as initiation of HRT appropriately for improvement of liver function.¹ The frequency with which these guidelines are being complied with or the resulting change in clinical outcomes remain unknown, and exploring these outcomes was a secondary objective of this study.

METHODS

STUDY DESIGN AND DATA SOURCE

This is a retrospective observational study using a national sample of electronic health record (EHR) data from 2009 and 2019 available through PEDSnet, the largest pediatric health learning system and the first and only database with the potential to capture rare pediatric diagnoses in large numbers.⁷ The PEDSnet collaborative was developed to facilitate pediatric research across multiple nonprofit health systems with funding from the Patient-Centered Outcomes Research Institute (PCORI). EHR data included all inpatient and outpatient encounters with corresponding provider specialties, dates of encounters, vital signs, diagnoses, and prescriptions, as well as results and dates for laboratory tests, radiology tests, and procedures. All dates were shifted within a three-month period to protect personal health information. The study was not classified as human subjects research as only de-identified data were obtained from the PEDSnet Data Coordinating Center.

DEFINING CASES AND CONTROLS

Pediatric patients with TS (“cases”) were identified as those who had a billing or problem list diagnosis coded as Turner syndrome, monosomy X, or gonadal dysgenesis (which was the ICD-9 code used for Turner syndrome prior to 2015) and at least one outpatient visit encounter in PEDSnet between 0–25 years of age (Table A-1). All patients with a male sex were excluded. A total of 2,145 unique patients with TS were included for analysis and subsequently matched from a pool of ~100,000 female patients < 25 years of age who had at least one outpatient visit encounter in PEDSnet. A greedy match algorithm with caliper width of 0.10 was used to obtain a 1 to 4 matching of cases to control.⁸ Variables included in the matching procedure included PEDSnet site, race, ethnicity, insurance type (public/private/none), birth year, duration in PEDSnet database (duration was set to 0 for individuals with a single encounter), and age at most recent visit in PEDSnet. Propensity score matching was chosen because it enables matching on a large number of variables with a relatively small loss in statistical power.⁹ The most commonly coded diagnoses among controls were well child visit, needs influenza immunization, (muscular) incoordination, constipation, and acute pharyngitis – consistent with what would be expected in a general pediatric population.

OUTCOMES OF INTEREST

The primary outcome of interest was liver enzyme elevation. Alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma-glutamyl transferase (GGT) were considered, but only ALT and AST were analyzed as too few patients had GGT measurements available. ALT was the preferred test since it is most specific for liver disease as a result of its presence mainly in the cytosol of the liver and low concentration elsewhere.¹⁰ Therefore, results are presented for ALT with some supplementary analyses for AST shown in Figure A-1. The upper limits of normal (ULNs) for ALT and AST were determined for each patient based on age at the time of the lab measurement and sex using the 97.5^th percentile of liver enzyme levels from a common source.¹¹ The ULNs from the reference were calculated using the ‘childsds’ package in R. For categorical analyses, liver enzyme concentrations were categorized as normal, 1–2 times the ULN, 2–3 times the ULN, and greater than 3 times the upper limit of normal for age and sex. For analyses that focused on whether a patient ever had liver enzyme abnormalities (prevalence of any liver dysfunction), we used only each patient’s highest recorded ALT/AST. Longitudinal analysis is presented in Figure 4, which maps the trajectory of patients grouped based on the severity of their first abnormal liver enzyme result.

Figure 4.

ALT progression from onset of abnormality to most recent visit

† ALT: Alanine transaminase, AST: Aspartate transaminase

Secondary outcomes of interest included any of the following liver diagnoses in the problem list or billing codes: fatty liver disease, hepatitis (any type), cirrhosis/fibrosis of the liver, liver tumor/malignancy, and congenital anomaly of the liver, which are previously reported liver abnormalities in the TS population.

To assess the benefit of liver screening in this population, we also analyzed how often changes of care resulted from testing. The changes of care considered were as follows: laboratory evaluation, liver ultrasound, ursodiol prescription, liver biopsy, and liver transplant or resection. We restricted the analysis to only include those with an elevated liver enzyme result to be able to reasonably infer that these changes of care were indeed in response to testing. Next, we divided the total number of patients with TS who had their enzymes checked by the number of patients with TS who experienced changes of care in response to abnormal enzymes to calculate a “Number Needed to Screen” (NNS), akin to the more commonly used “Number Needed to Treat (NNT)”.¹² This calculation reveals how many patients would have to be screened in order for one patient to have a change of care, reflecting the utility of screening.

COVARIATES OF INTEREST

Several potential risk factors for liver enzyme elevation were considered for inclusion in the models: age at the time of liver enzyme measurement,¹³ highest body mass index (BMI) z-score,¹⁴ estrogen prescription,¹⁵ as well as comorbidity with celiac disease,¹⁶ diabetes,¹⁷ any thyroid condition,¹⁸ and any cardiac condition.¹⁹ The impact of estrogen was also separately analyzed by comparing enzymes of TS patients above the age of 13 years prescribed estrogen vs. those who were not. Because of reports of milder complications associated with 45,X/46,XX mosaicism, we also examined the impact of mosaicism (defined as one or more billing diagnoses for mosaic Turner syndrome).

STATISTICAL ANALYSIS

Descriptive statistics were used to compare TS cases and matched controls. Categorical variables were compared using the chi-squared goodness of fit test while continuous variables were compared using either the two-sample t-test (for normally distributed variables) or the Wilcoxon rank sum test (for non-normally distributed variables). Multinomial regression models were performed within the TS group for each of the covariates independently, followed by a combined model for all those that were individually significant p<0.05. This strict cutoff, compared to other studies that sometimes use more loose cutoffs around p-value 0.20, was chosen because we were interested in the most significant, not necessarily all, predictors for liver enzyme elevation in this group. Case-control analysis involving conditional logistic regression for the various liver enzyme elevation categories and liver diagnoses was conducted using a generalized estimating equations (GEE) approach (through the R package ‘geepack’) to account for the dependence introduced by the matching procedure. Univariate comparisons showed statistically significant differences in BMI z-score between the case and control groups, so we adjusted for the highest recorded BMI z-score for each patient in the logistic regression model (excluding outliers with BMI z-score <−5 or > 5). We elected to use the highest BMI (as opposed to lowest or average), to capture cases for which elevated liver enzymes are associated with the presence of obesity. These data were missing for 1,666 controls and 185 TS cases. A p-value < 0.05 was considered significant and odds ratios with 95% confidence intervals are reported. Analyses were conducted using R version 3.6.1.²⁰

In a sensitivity analysis to validate the accuracy of a diagnosis of Turner syndrome in the EHR, we repeated our primary analyses using only the subset of the TS population who had at least two diagnoses coded as Turner syndrome (not including gonadal dysgenesis) (n = 1,691), and there were no differences in results.

RESULTS

DEMOGRAPHIC SUMMARY

Matching successfully produced demographically homogenous case and control samples (Table 1). Patients were followed at PEDSnet institutions for an average of 7.6 years, which enabled longitudinal analyses of liver enzymes. The majority of patients were white, non-Hispanic, and privately insured, however there was diverse representation in this analytic cohort. The distribution of PEDSnet sites was similar across both groups to ensure that differences in outcomes were not attributable to regional differences in healthcare or systems. The median age for the first documented diagnosis of TS was 8.2 years (interquartile range (IQR): 2.2–13.1).

Table 1.

Demographic variables for Turner Syndrome (TS) cases and matched controls

	TS Cases (n = 2145)	Controls (n = 8580)	p-value
Age at First Visit (yrs)	4.4 (0.3–10.2)	4.3 (0.7–9.9)	0.04
Age at Last Visit (yrs)	14.3 (8.5–18.5)	14.6 (10.0–17.2)	0.08
Total # Outpatient Visits	17 (8–38)	7 (3–20)	< 0.01
# of Visits with TS Diagnosis	10 (4–22)	-
Follow up period (yrs)	7.6 (5.4)	7.6 (5.7)	0.99
Primary care encounter	850 (39.6%)	3480 (40.6%)	0.45
Race			> 0.99
White	1458 (68.0%)	5823 (67.9%)
Black	179 (8.3%)	696 (8.1%)
Asian	54 (2.5%)	215 (2.5%)
Other	284 (13.2%)	1158 (13.5%)
Unknown	170 (7.9%)	688 (8.0%)
Ethnicity			0.68
Hispanic	311 (14.5%)	1240 (14.5%)
Non-Hispanic	1704 (79.4%)	6775 (79.0%)
Unknown	130 (6.1%)	565 (6.6%)
Insurance			0.04
Public	820 (38.2%)	3318 (38.7%)
Private	1119 (52.2%)	4433 (51.7%)
Other	50 (2.3%)	543 (6.3%)
Unknown	27 (1.3%)	287 (3.3%)
PEDSnet Site			0.93
1	252 (11.7%)	1043 (12.2%)
2	423 (19.7%)	1612 (18.8%)
3	290 (13.5%)	1135 (13.2%)
4	552 (25.7%)	2230 (26.0%)
5	426 (19.9%)	1745 (20.3%)
6	202 (9.4%)	815 (9.5%)

^†Numerical data presented as mean (SD) if normally distributed, median (IQR) if non-normally distributed. Categorical data presented as n (%). Variables included in this table were matched upon using propensity-score methods: age at most recent visit, duration in PEDSnet, race, ethnicity, insurance status, and site. Other matching variables not shown in table are sex (all female) and birth year.

LIVER ENZYME ABNORMALITIES AND LIVER DISEASE IN TURNER SYNDROME

59% of the 2,145 girls with TS had liver enzyme values in the dataset, compared to 27% of controls. When considering only those who are at least 10 years of age by their last visit (as the clinical care guidelines recommend annual assessment of liver enzymes beginning at 10 years), 66% of girls with TS had enzyme values compared to 31% of controls. The median number of ALT measurements per year in patients with TS after age 10 was 0.50 (IQR: 0.25–1.00), which is less frequent than recommended in the current clinical practice guidelines and statistically no more common than in controls (0.33, IQR: 0.17–0.67).

Overall, 58% of patients had an ALT above the ULN and 37% had an AST above the ULN (Table 2A). Substantial portions of the TS cohort had clinically significant elevation (> 2x ULN) of ALT (18%) and AST (9%). The onset of these abnormalities was early in life, with ALT abnormalities occurring at a median age of 12.4 years (IQR: 7.3–15.4). Similarly, AST abnormalities occurred at a median age of 12.1 years (IQR: 7.2–14.9). As seen in Figure 1, at least 38% of patients across all age groups had abnormal ALT. With relevance to the TS clinical practice guidelines recommending screening for liver enzyme abnormalities begin at age 10, 24% of girls with ALT >2x ULN were under 10 years of age, and 35% of those with AST >2x ULN were under 10 years of age.

Table 2.

Summary of liver enzymes and diagnoses among Turner Syndrome (TS) cases compared to matched controls

	TS Cases (n = 2145)	Median Age at Onset (yrs)	Controls (n = 8580)	Median Age at Onset (yrs)
ALT Elevation	667/1159 (57.6%)	12.4 (7.3–15.4)	914/2134 (42.8%)	13.0 (7.7–15.9)
1–2x ULN	462 (39.9%)	12.2 (6.8–15.3)	673 (31.5%)	13.0 (7.7–15.9)
2–3x ULN	89 (7.7%)	14.4 (10.5–17.1)	79 (3.7%)	13.2 (7.4–16.8)
>3x ULN	116 (10.0%)	13.6 (8.7–17.0)	162 (7.6%)	13.3 (7.7–16.1)
AST Elevation	433/1158 (37.4%)	12.1 (7.2–14.9)	499/2116 (23.6%)	12.5 (8.0–15.9)
1–2x ULN	327 (28.2%)	12.0 (7.2–14.9)	341 (16.1%)	12.3 (7.9–15.8)
2–3x ULN	53 (4.6%)	13.0 (6.7–16.4)	67 (3.2%)	13.8 (9.1–16.1)
>3x ULN	53 (4.6%)	13.2 (8.0–15.7)	91 (4.3%)	13.5 (9.0–16.2)
Liver Diagnoses	68/2145 (3.2%)	13.1 (2.2–15.8)	88/8580 (1.0%)	11.9 (8.2–16.0)
Fatty Liver	28 (1.3%)	14.5 (12.9–16.6)	39 (0.5%)	12.4 (9.6–16.2)
Hepatitis	21 (1.0%)	13.0 (4.0–15.2)	17 (0.2%)	8.8 (4.3–14.6)
Liver tumor/malignancy	8 (0.3%)	7.5 (0.9–15.1)	4 (0.05%)	16.6 (14.9–17.6)
Congenital anomaly	3 (0.1%)	1.8 (1.1–10.0)	3 (0.03%)	15.7 (7.9–17.3)
Cirrhosis/Fibrosis	5 (0.2%)	15.3 (12.7–19.0)	3 (0.03%)	13.6 (7.0–17.7)

^†ALT: Alanine transaminase, AST: Aspartate transaminase

Figure 1.

Age breakdown of liver enzyme elevation: A. Categorization of alanine transaminase (ALT) abnormalities for different age groups in TS. B. Prevalence of ALT abnormality for different age groups in TS. C. Categorization of ALT abnormalities for different age groups in controls. D. Prevalence of ALT abnormality for different age groups in controls.

† TS: Turner syndrome

While most patients with TS had liver enzyme elevation in their EHR, only 3% of the TS population had liver-related diagnoses (as seen in Table 2A). However, this was significantly higher than controls. Specific diagnoses consisted of fatty liver (1.3% of cases), hepatitis (1% of cases), and fewer than 10 cases had diagnoses of liver tumor/malignancy, congenital anomaly, and cirrhosis/fibrosis. When these diagnoses did occur, the median age at first documentation of the diagnosis was 13.1 (IQR 2.2–15.8). Of the 59 TS patients with both liver diagnoses and liver enzyme results available in the dataset, 93% had elevated enzymes and 75% had enzymes > 2x ULN.

ANALYSIS OF CASES AND CONTROLS

Cases were significantly more likely to be 1–2x ULN, 2–3x ULN, and >3x ULN, with odds ratios (OR) of 1.65, 2.68, and 1.70 respectively as shown in Figure 2A. Cases also had 2.37 times higher odds to have a specific liver diagnosis, including fatty liver (OR: 1.88), hepatitis (OR: 3.69), cirrhosis/fibrosis (OR: 5.79), and liver tumor/malignancy (OR: 4.78), as shown in Figure 2B.

Figure 2.

Odds ratios (OR) and 95% confidence intervals of liver enzyme elevation and liver diagnoses in Turner syndrome cases compared to matched controls. OR of 1.0 indicates no difference in odds between groups.

† ALT: Alanine transaminase

RISK FACTORS FOR LIVER ENZYME ABNORMALITIES IN TURNER SYNDROME

In univariate regression, age, BMI, presence of a thyroid or heart condition or diabetes, and estrogen prescription significantly impacted risk of having abnormal liver enzymes. When included in a combined multinomial logistic regression model, age had a significant but weak impact for all elevation categories (Figure 3). BMI impacted odds of being in the clinically significant elevation categories (2–3x ULN and >3x ULN) with OR of 1.25 and 1.20 respectively. Neither celiac disease nor presence of a thyroid condition significantly affected odds of liver enzyme elevation in any of the elevation categories. Heart conditions affected odds of clinically significant liver enzyme elevation (2–3x ULN, >3x ULN) with OR of 1.82 and 2.13 respectively. Diabetes had a significant impact on odds of severe elevation (> 3x ULN) with an odds ratio of 2.63.

Figure 3.

Multinomial logistic regression investigating variables that were independently associated with Turner syndrome liver enzyme abnormalities. Point estimate is the odds ratio and error bars represent the 95% confidence interval. Asterisks indicate variables that significantly impact odds for liver enzyme elevation.

† ALT: Alanine transaminase

A prescription for estrogen did not impact risk of elevation in the combined model. Furthermore, estrogen prescription did not have a significant impact on ALT, and no differences in liver enzymes in oral versus transdermal estrogens were found (Table A-2). Patients who were older at estrogen initiation had higher odds of severe elevation (OR: 1.13, 95% CI: 1.01–1.26).

LONGITUDINAL ANALYSIS OF LIVER ENZYMES IN TS

After the first abnormal liver enzymes are observed, the most likely outcome is to remain at the same degree of elevation (~50% of patients) or improve (~45% of patients) during the follow-up period in PEDSnet (Figure 4). Few patients worsened from their first abnormal ALT or AST: 6% for ALT and only 3% for AST. 54% of those with ALT > 3x ULN improved.

Of those with abnormal liver enzyme values, 25% had a change of care (hepatitis-specific laboratory workup, liver ultrasound, liver biopsy, or ursodiol prescription) reflected in the EHR. This corresponds to a number needed to test of 7 to result in a change of care (Figure 5). Amongst those with clinically significant elevation (> 2x ULN), 49% had a change of care.

Figure 5.

Number of patients who experienced changes of care out of those with abnormal liver enzymes and number needed to screen in order to result in a change of care.

† TS: Turner syndrome

DISCUSSION

This study used a national EHR dataset to study pediatric liver enzyme abnormalities and compliance with consensus TS care guidelines in one of the largest pediatric TS cohorts reported to date. We found that elevation of liver enzymes is significantly more common in children with TS than in matched controls, with over half of youth with TS who had liver enzyme values available in the EHR having values in the elevated range. Girls with TS were also more likely to have moderate to severe liver enzymes elevation (>2x ULN) and specific liver disease diagnoses, supporting that there is clinical significance to the differences we observed in liver enzymes. Additionally, we observed that based on EHR, patients with TS often do not have their liver enzymes tested, rarely with the frequency recommended by current guidelines, and half do not receive further evaluation in response to enzyme elevation. This big data secondary data analysis supports that further investigation into TS liver dysfunction is warranted.

Previous literature on liver enzyme abnormalities in TS is limited, with prevalence estimates ranging from 20 to 80%, although younger patients are likely at the lower end of the spectrum.⁴ Our finding of 58% of pediatric patients with TS who had liver enzymes values available in the EHR having had at least one ALT above the ULN is on the higher side of this estimate. This higher prevalence is likely because our analysis included several years of data whereas most studies to date evaluate a single time point. We cannot rule out the possibility that some of our captured liver enzyme elevations were secondary to transient acute illness rather than chronic TS-related pathophysiology, however, we found more than half of TS patients with liver enzyme abnormalities persisted or worsened over time. Furthermore, nearly 20% had liver enzymes greater than 2x the ULN, which is more suggestive of clinically significant liver pathology. Our results support that both mild and severe liver enzyme elevation is more common in girls with TS than well-matched controls and is likely to represent a chronic process.

Although the prevalence of liver enzyme elevation was high, only 3% of patients with TS had a recorded liver diagnosis in their EHR. This may support a TS-specific pseudo-elevation of enzymes that is non-progressive, a hypothesis with some limited evidence in the literature.²¹ Nonetheless, it is important to be aware of aminotransferase elevations since we did find higher risk for liver diagnosis in TS patients compared to controls, specifically due to fatty liver disease, hepatitis, cirrhosis/fibrosis, and liver tumor/malignancy, and there was a strong relationship between these diagnoses and elevated liver enzymes. Given clinical outcomes may take years to develop, it is possible that the high prevalence of liver enzyme elevation we observed may progress to liver disease in adulthood after these patients are outside of the PEDSnet system. This is consistent with the finding of increased liver and gastrointestinal disease in a study of 1156 women with TS identified through the Danish Cytogenetic Central Registry.²²

A few studies have been conducted in attempt to capture the etiology of these abnormalities. Roulot concluded the main causes of liver involvement in TS were likely vascular disorders of a congenital origin and nonalcoholic fatty liver disease.¹⁹ Others have hypothesized liver abnormalities may be due to underlying autoimmune hepatitis connected to higher rates of autoimmune disorders in girls and women with TS.²³ Many studies find that lack of estradiol is involved in liver dysfunction and HRT leads to a reduction of liver enzymes, although perhaps not quantitative liver functions (such as hepatic capacities for conversion of galactose, indocyanine green, and antipyrine).^24–26 Nonetheless, the mechanism behind liver involvement in TS is not entirely clear. Our study identified age, BMI, and comorbidities such as diabetes and cardiovascular disease were associated with higher odds for enzyme elevation. Although age was a significant covariate, the median age for onset of abnormality was 12 years old with 40% of patients under age 10 having liver enzyme elevation, emphasizing that it is not just age-related etiologies and suggesting screening for liver dysfunction should begin in childhood.

On the other hand, we found mosaicism, estrogen, and celiac disease did not significantly change the odds of liver enzyme elevations among patients with TS. Patients with non-mosaic 45,X present with a more severe phenotype compared to those with a 46,XX cell line, therefore we hypothesized they would also have more severe liver disease. The lack of relationship between mosaicism and liver enzymes could be because karyotype cannot reliably predict phenotype or due to limitations in the dataset as a code for Turner mosaicism could be inaccurate or include cell lines other than 46,XX. As actual genetic results were not available in our dataset, future work to investigate the reliability of EHR diagnostic coding for correctly identifying TS and mosaicism status is underway.

The null results of HRT impacting liver enzymes is perhaps not surprising as the relationship between estrogen deficiency, estrogen replacement (particularly oral formulations), and liver function are complex. We cannot determine from our database who has hypogonadism (and therefore warrants hormone replacement therapy) and who has adequate ovarian function. Future work should investigate the effect of HRT timing, dosing, and formulation on development of liver disease.

Finally, the result regarding celiac disease is especially relevant given that females aged >10 years with TS have a fivefold greater risk of celiac.²⁷ While celiac disease has been found to be the underlying cause of unexplained elevation of liver enzymes in some studies, we would expect our cross-sectional approach to unveil any abnormality in liver enzymes prior to gluten-free diet adjustment.²⁸ A caveat to our results suggesting a lack of relationship between celiac and liver enzyme abnormalities in TS is that undiagnosed celiac disease would not be captured in our analysis and could result in liver enzyme abnormalities.

Our findings may inform important practice changes regarding care for TS patients. First, abnormalities appear to occur earlier than the current guidelines recommend initiating screening. While we recognize that the patients in our study who had liver enzymes assessed at a young age may be have an increased risk of liver disease (acute or chronic) that clinically warranted testing, we would expect the same among controls as well. Despite this, the odds of elevated liver enzymes were higher among TS compared to controls at all ages. These results support considering a revision to the clinical practice guidelines that currently recommend screening liver function beginning age 10, particularly when other risk factors such as high BMI or presence of comorbidities are present. However, longitudinal assessment to study risk stratification across different age groups is needed. Furthermore, recommendations for evaluation and management of liver enzyme elevation are not provided in the current clinical practice guidelines, which is reflected in our data with fewer than half of those with liver enzymes > 2x ULN had further evaluation. This is a documented issue in TS care, with a study in Germany finding that, while medical management focused on endocrine issues was strong, liver involvement was often neglected in women with TS.²⁹ Given the strong relationship between liver disease diagnoses and elevated liver enzymes in this cohort, as well as the observation that the majority of liver enzyme elevations did not resolve with longitudinal follow up suggests that further hepatology evaluation is warranted when elevated liver enzymes are present.

A variety of aspects of our study design uniquely equip us to understand the effect of the missing sex chromosome on the liver. The sample size in this study (n=2,145) makes it one of the largest focused on pediatric patients with TS. We also have a large control group well-matched on multiple demographic variables, which increases the confidence in our case-control analysis and sensitivity for detection of differences in liver-related abnormalities. The detailed information we have on patients in PEDSnet (including visits, conditions, prescriptions, measurements, and test results) as well as the longitudinal nature of the data allows us to assess many different aspects of liver health in TS that have not previously been investigated.

However, our lack of information about any visits that occur outside of the PEDSnet sites presents a significant limitation for this study to evaluate care and compliance with guidelines. Another relevant limitation is that our analyses relied on de-identified compilations of medical records, which are susceptible to errors in coding that might occur. The diagnosis of TS has not been specifically validated in this database. There is risk of both omission (missed cases because the diagnosis was not entered) and commission (erroneously entered). Furthermore, when looking at the liver enzyme outcomes, analyses relying on highest ALT/AST values could reflect acute and temporary illness rather than chronic liver dysfunction. However, this method is still preferable to using first, average, or most recent values in this heterogeneous dataset as these approaches would likely underestimate liver disease that develops over time or improves with interventions.

Another statistical limitation is related to the differences in number of times the cases and controls are seen. Since the cases are seen more often, there is a higher likelihood of observing abnormalities among this group. Finally, it’s important to recognize that the pediatric centers included here are tertiary care referral centers. As a result, the TS patients in this study may not be entirely representative of the pediatric TS population as a whole. Despite these limitations, this work adds valuable knowledge on liver disease in girls with TS to the literature.

Many questions remain regarding TS and the liver, including pathophysiology and natural history. Whether elevated liver enzymes in girls with TS represent pathology or perhaps different physiology that need to be interpreted with TS-specific normal ranges requires further investigation. Basic research investigating mechanisms of how monosomy X, aging, BMI, various comorbidities and medications affect liver function is lacking. Furthermore, while our longitudinal analysis suggested liver enzyme elevation did not progress through childhood, tracking patients into adulthood when the majority of liver disease progression occurs will be critical to identify at risk subgroups and preventive strategies.

In conclusion, liver enzyme elevation in pediatric patients with TS is common and the odds of both elevated liver enzymes and specific liver diseases are higher in TS compared to matched controls. Enzyme elevation occurs at an early age but does not significant worsen over time in the pediatric population. Both biochemical abnormalities and liver disease were positively associated with age, BMI, and specific medical comorbidities. Given these findings, it is important that liver function is assessed in pediatric patients with TS and additional diagnostic evaluation pursued if abnormalities are identified.

List of Abbreviations:

TS

Turner syndrome

ALT

alanine aminotransferase

AST

aspartate aminotransferase

GGT

gamma-glutamyl transferase

ULN

upper limit of normal

OR

odds ratio

CI

confidence interval

IQR

interquartile range

GEE

generalized estimating equations

APPENDICES

Table A-1. SNOMED codes categorized as TS diagnosis

SNOMED Code	Description
38804009	Turner syndrome
710008008	Monosomy X
710019006	Mosaicism 45, X or other cell line with abnormal sex chromosome
205686009	Karyotype 46, X iso
205687000	Karyotype 46, X with abnormal sex chromosome except iso
205688005	Turner’s phenotype, mosaicism 45,X; 46,XX or 45,X;46,XY
205681004	Gonadal dysgenesis
205683001	Ovarian dysgenesis
710010005	Mosaic Turner syndrome
205708001	Mosaic XO/XX
302960008	Mosaicism 45, X; 46, XX
254281006	Turner’s phenotype - ring chromosome karyotype*
254280007	Turner’s phenotype, partial X deletion karyotype†

^†No cases

Table A-2.

A-2A.

Analysis of liver enzymes among girls with TS > age 13 with and without estrogen initiation

	Estrogen Initiated (n = 559)	Estrogen Not Initiated (n = 206)
Highest ALT			OR (95%CI)
Normal	211 (37.7%)	85 (41.3%)
1–2x ULN	232 (41.5%)	82 (39.8%)	1.14 (0.80–1.63)
2–3x ULN	50 (8.9%)	18 (8.7%)	1.12 (0.62–2.03)
> 3x ULN	66 (11.8%)	21 (10.2%)	1.27 (0.73–2.20)

A-2B.

Regression on ALT categories for girls with estrogen vs. age of initiation

Highest ALT	Odds ratio	95% CI	P-value
1–2x ULN	1.02	0.94–1.11	0.58
2–3x ULN	1.12	0.99–1.27	0.07
> 3x ULN	1.13	1.01–1.26	0.04*

A-2C.

Oral vs. transdermal estradiol

Highest ALT	Odds ratio (oral vs. transdermal)	95% CI	P-value
1–2x ULN	0.83	0.54–1.27	0.38
2–3x ULN	0.54	0.26–1.11	0.09
> 3x ULN	1.34	0.69–2.63	0.39

† ALT: Alanine transaminase, AST: Aspartate transaminase, CI: Confidence interval

Figure A-1.

A-1A.

Age breakdown of liver enzyme elevation: A. Categorization of AST abnormalities for different age groups in TS. B. Prevalence of AST abnormality for different age groups in TS. C. Categorization of AST abnormalities for different age groups in controls. D. Prevalence of AST abnormality for different age groups in controls.

A-1B.

AST progression from onset of abnormality to most recent visit

† AST: Aspartate transaminase