HORMONE RESEARCH IN PÆDIATRICS
Horm Res Paediatr DOI: 10.1159/000488028
Multiples of Median-Transformed, Normalized Reference Ranges of Steroid Profiling Data Independent of Age, Sex, and Units
Dominika Zalasa,c Thomas Reinehrb Marek Niedzielac Christoph Borzikowskyd Maciej Flader“ Gunter Simic-Schleichere Halit Ilker Akkurtf Sabine Heger9 Nadine Horniga Paul-Martin Holterhusa Alexandra E. Kulleª
ªDivision of Pediatric Endocrinology and Diabetes, Department of Pediatrics, University Medical Center SH, Campus Kiel, Kiel, Germany; bVestische Hospital for Children and Adolescents, University of Witten/Herdecke, Datteln, Germany; “Department of Pediatric Endocrinology and Rheumatology, Poznan University of Medical Sciences, Poznan, Poland; dInstitute of Medical Informatics and Statistics, University Medical Center SH, Campus Kiel, Kiel, Germany; eChildren’s Hospital of Bremen-Nord, Bremen, Germany; fPediatric Endocrinology, Children’s Hospital Altona, Hamburg, Germany; 9Children’s Hospital “Auf der Bult”, Hannover, Germany
Keywords
Congenital adrenal hyperplasia · Multiples of median · Reference ranges · Liquid chromatography tandem mass spectrometry
Abstract
Background/Aims: The high complexity of pediatric refer- ence ranges across age, sex, and units impairs clinical appli- cation and comparability of steroid hormone data, e.g., in congenital adrenal hyperplasia (CAH). We developed a mul- tiples-of-median (MoM) normalization tool to overcome this major drawback in pediatric endocrinology. Methods: Liq- uid chromatography tandem mass spectrometry data com- prising 10 steroid hormones representing 905 controls (555 males, 350 females, 0 to >16 years) from 2 previous datasets were MoM transformed across age and sex. Twenty-three genetically proven CAH patients were included (21-hydroxy- lase deficiency [210HD], n = 19; 11ß-hydroxylase deficiency [11OHD], n = 4). MoM cutoffs for single steroids predicting
21OHD and 11OHD were computed and validated through new, independent patients (21OHD, n = 8; adrenal cortical carcinoma, n = 6; obesity, n = 40). Results: 21OHD and 11OHD patients showed disease-typical, easily recognizable MoM patterns independent of age, sex, and concentration units. Two single-steroid cutoffs indicated 21OHD: 3.87 MoM for 17-hydroxyprogesterone (100% sensitivity and 98.83% specificity) and 12.28 MoM for 21-deoxycortisol (94.74% sen- sitivity and 100% specificity). A cutoff of 13.18 MoM for 11-deoxycortisol indicated 11OHD (100% sensitivity and 100% specificity). Conclusions: Age- and sex-independent MoMs are straightforward for a clinically relevant display of multi-steroid patterns. In addition, defined single-steroid MoMs can serve alone as predictors of 21OHD and 11OHD. Finally, MoM transformation offers substantial enhance- ment of routine and scientific steroid hormone data ex- change due to improved comparability.
@ 2018 S. Karger AG, Basel
KARGER
Introduction
Endocrine diseases, e.g., Cushing syndrome, polycys- tic ovary syndrome, and different forms of congenital ad- renal hyperplasia (CAH), are often associated with typical profiles of steroid hormone concentrations [1, 2]. Accu- rate determination of plasma steroids and their precur- sors is, therefore, the prerequisite for correctly diagnosing steroid-dependent diseases in pediatric and adult endo- crinology [1, 2]. Liquid chromatography tandem mass spectrometry (LC-MS/MS) has become the method of choice for measuring steroids nowadays [3-9]. The com- plexity of steroid profiles resulting from this technique - although providing comprehensive information - might, however, complicate a straightforward disease-related in- terpretation.
In general, laboratory results are highly influenced by the specific assay methodology used [10-12] and by pre- analysis treatment of the sample [13]. In pediatric endo- crinology, well-defined reference ranges covering the en- tire pediatric age range, gender, and pubertal develop- ment are also of utmost importance [14]. To date, a lack of such reference data in pediatric endocrinology often affects clinical and scientific interpretation, the ability to exchange and compare steroid hormone data between different labs, centers, and countries, as well as a com- parison of reference hormone data in medical textbooks.
Standardization of reference ranges transforms the data from absolute concentrations to relative values, facilitating a direct comparison of data from different laboratories [15, 16]. Several approaches have been attempted to standard- ize data: the estimation of percentiles used by Elveback and Taylor [17] or the traditional approach based on the sam- ple quantiles, or distribution-free quantile estimators in- troduced by Harrell and Davis [18]. The idea of expressing laboratory measurements as multiples of medians (MoMs) was first suggested in 1977 by Wald et al. [19], who used them for comparing maternal serum a-fetoprotein levels in neural-tube defects in early pregnancy. The application of MoMs was afterwards reported in various fields of med- icine [20, 21]. To date, most maternal lab values during pregnancy are expressed in MoMs [22, 23].
Our first aim was to simplify the display of complex steroid profiling data and make them independent of sex, age, and units. We, therefore, used the MoM standardiza- tion procedure, which we applied for the first time in LC- MS/MS analysis.
The underlying reference data on hormone concentra- tions and median values needed for this project were based on 2 previously published studies in children and
adolescents [24, 25]. We established a matrix of 140 refer- ence medians and a MoM reference range for 10 LC-MS/ MS-determined steroid hormones, namely progesterone, deoxycorticosterone, corticosterone, 17-hydroxyproges- terone, 11-deoxycortisol, 21-deoxycortisol, cortisol, cor- tisone, androstenedione, and testosterone.
In order to validate this concept, we subsequently ana- lyzed steroid profiles in treatment-naïve children with 21-hydroxylase deficiency (21OHD) and 11ß-hydroxylase deficiency (11OHD), respectively, the most common forms of CAH [26, 27]. In addition, we aimed to further improve the efficacy of the endocrine diagnostic process for 21OHD and 11OHD based on a diagnostic single-ste- roid MoM model defining specific cutoffs, which we val- idated in children with adrenal cortical carcinoma (ACC) and obesity and in independent new CAH patients.
Patients and Methods
The study was approved by the ethics committee of the Chris- tian-Albrechts-University of Kiel, Germany (file number D531/16). Mainly, anonymized routine ultra-performance (UP) LC-MS/MS steroid data from the Pediatric Endocrine Laboratory at the De- partment of Pediatrics, University Hospital of Schleswig-Holstein, Kiel, Germany, were used. Additional samples from the Depart- ment of Pediatric Endocrinology and Rheumatology, Poznan Uni- versity of Medical Sciences, Poznan, Poland, were analyzed. Data were then anonymized and subsequently included in the study ac- cording to the requirements of the ethics committee of the Poznan University of Medical Sciences.
MoM Calculation
We have previously published gender- and age-specific refer- ence values for an LC-MS/MS-based method for determination of progesterone, deoxycorticosterone, corticosterone, 17-hydroxy- progesterone, 11-deoxycortisol, 21-deoxycortisol, cortisol, corti- sone, androstenedione, and testosterone [24, 25]. These publica- tions include the median values used for calculating the MoMs. The reference subjects are described in these previous publications [24, 25]. In brief, the subjects did not suffer from an endocrine or systemic disease and they were not on steroid medication.
The calculation of each subject’s MOM followed the equation by Wald et al. [19] in which the measured single plasma hormone concentration (subject’s result) is divided by the population me- dian derived from the corresponding age- and gender-specific group. For example, a female patient aged 4.3 years showed a plasma 17-hydroxyprogesterone concentration of 10.4 ng/mL. The applicable sex- and age-specific median in the reference cohort is 0.19 ng/mL. The equation for this patient is as follows: MoM17-hydroxyprogesterone = 10.4/0.19 = 54.7. In essence, the medians for the 2 different sexes, 7 different age groups, and each of the 10 different hormones resulted in a total of 140 medians.
In conclusion, a MoM of a particular hormone in a given refer- ence subject or in a given patient enables the display of hormone data independent of sex and age and without a plasma concentra-
tion unit, but the data still remain mathematically related to the original sex- and age group-specific median. In order to make the MoMs more conspicuous graphically, we decided to display MoM calculations in the log10 scale.
Statistical Analysis
Statistical evaluation of study data was performed using the statistical software GraphPad Prism version 5 (GraphPad Software Inc., San Diego, CA, USA) and R (R Core Team [2015], Vienna, Austria). The nonparametric Wilcoxon rank sum test was used to evaluate statistical differences between the reference group and the CAH group. A p value of <0.05 was considered to be statistically significant. Receiver operating characteristic (ROC) curves were used to define the cutoff values for the MoM values for the differ- ent significant steroid hormones of the patients and controls.
CAH Patients
Diagnostic routine plasma samples derived from 23 patients, who were subsequently genetically confirmed to be suffering from CAH, were analyzed. They had all been admitted to the Pediatric Endocrinology Division of the Christian-Albrechts-University, Kiel, Germany, between May 2005 and May 2014. Nineteen of these had 21OHD (21OHD cohort 1; 8 were male, 11 were female, and they were aged 2 days to 10.1 years), representing classical and nonclassical 21OHD (online suppl. Table 3; for all online suppl. material, see www.karger.com/doi/10.1159/000488028). Four of these patients had 11OHD (1 was male, 3 were female, and they were aged 3 days to 5.7 years; online suppl. Table 3).
First, the MoMs for each of the 10 steroid hormones were de- termined in each of the 23 CAH patients. Secondly, these MoMs were statistically compared between the 21OHD patients and the reference cohort, and between the 11OHD patients and the refer- ence cohort, to uncover potential significantly different MoMs be- tween the CAH patients and the reference cohort. This informa- tion was used for the third step: diagnostic cutoff values were cal- culated for the hormone MoMs with a significant difference between 21OHD and the reference cohort and for those with a significant difference between 11OHD and the reference cohort. In total, 2 cutoff values for MoMs in 21OHD and 1 for a MoM in 11OHD were used for further validation.
CAH Validation Group
A new and independent group of CAH patients served for val- idation of the MoM-derived cutoffs for diagnosis of 21OHD (21OHD cohort 2). These patients were admitted to the Pediatric Endocrinology Department of the Christian-Albrechts-Universi- ty, Kiel, Germany, for the first time between June 2014 and Octo- ber 2015 and to the Department of Endocrinology and Rheumatol- ogy, Poznan University of Medical Sciences, Poznan, Poland, be- tween September 2014 and September 2015. All patients had the presumed diagnosis of 21OHD, and serum samples were mea- sured to confirm or exclude this diagnosis. This group consisted of 9 patients with 210HD (n = 8: 2 males, 6 females; aged 2 days to 15.1 years) representing classical and nonclassical 21OHD (online suppl. Table 3). MoM values for the group were calculated accord- ing to the procedure described.
Non-CAH Cross-Validation Groups
In order to further validate the cutoffs, 2 groups of patients were selected who had a high clinical potential for hyperandrogen-
ism comparable with nonclassical CAH: patients with proven ACC and patients with documented obesity. Neither group showed evi- dence of further endocrine disorders. The inclusion criteria for ACC patients were clinical and histopathological evidence. Plasma was obtained prior to initiation of therapy (n = 6: 2 males, 4 fe- males; aged 3 months to 9.5 years). The ACC patients had origi- nally been analyzed for routine diagnostic purposes. The hormone concentration data of the obesity group (n = 40: 20 males, 20 fe- males; aged 6.4 to 10.6 years) has been published previously by Reinehr et al. [28].
MoMs were calculated according to the procedure described and were used to cross-validate the diagnostic value of the MoM cutoffs for 21OHD and 11OHD.
Hormone Analysis
The following hormones were selected for our study: proges- terone, deoxycorticosterone, corticosterone, 17-hydroxyproges- terone, 11-deoxycortisol, 21-deoxycortisol, cortisol, cortisone, an- drostenedione, and testosterone. The hormone analysis for the pa- tient samples was performed as previously described for the reference cohort [24, 25]. All steroid hormones were analyzed by the same UPLC-MS/MS method in the same lab using the same mass spectrometer as previously published. In brief, plasma ste- roid hormone concentrations were measured prior to initiation of any potential therapy. Plasma concentrations of all 10 steroids were determined simultaneously, using UPLC-MS/MS as previ- ously described [24, 25]. Aliquots of plasma samples, calibrator, and controls with a volume of 0.1 mL were combined with the in- ternal standard mixture to monitor recovery. All samples were ex- tracted using Oasis MAX SPE system plates (Waters, Milford, MA, USA). Hormone determination was carried out using a UPLC Quattro Premier/Xe system (Waters).
Molecular Analysis
DNA was extracted from the patients’ blood leukocytes using the Wizard Genomic Kit (Promega, Mannheim, Germany). A multiplex minisequencing method was employed to detect the most common mutations of the CYP21A2 gene [29] using an ABI 310 Sequencer (Applied Biosystems Inc., Foster City, CA, USA). In the case of a negative result, this was followed by CYP21A2 gene sequencing comprising the complete coding region, including in- tron-exon boundaries [30]. MLPA-multiplex ligation-dependent probe amplification (SALSA MLPA kit CAH, MCR-Holland, Am- sterdam, The Netherlands) was performed on all samples to iden- tify large CYP21A2 gene deletions. In order to determine CYP11B1 gene mutations, we performed direct DNA sequencing comprising the complete coding region of CYP11B1, including all intron-exon boundaries [31]. Samples were electrophoresed on an ABI 310 Se- quencer and analyzed with ABI Seq-Scape 1.1 software (Applied Biosystems Inc.) [31].
Results
Reference MoMs
Our previously published reference steroid hormone concentrations, including the gender- and age-specific median values [23, 24], were used to calculate the refer-
# 110HD patients
21OHD patients
100
log MoM
1
Progesterone MoM
Deoxycorticosterone MOM
Corticosterone MoM
17-Hydroxyprogesterone MOM
11-Deoxycortisol MoM
21-Deoxycortisol MOM
Cortisol MOM
Cortisone MoM
Androstenedione MoM
Testosterone MOM
| Hormone | 5% | 50% | 95% |
|---|---|---|---|
| Progesterone | 0.96 | 1 | 2.49 |
| Deoxycorticosterone | 0.65 | 1 | 3.60 |
| Corticosterone | 0.15 | 1 | 5.61 |
| 17-Hydroxyprogesterone | 0.24 | 1 | 3.03 |
| 11-Deoxycortisol | 0.30 | 1 | 4.55 |
| 21-Deoxycortisol | 0.90 | 1 | 5.58 |
| Cortisol | 0.28 | 1 | 2.46 |
| Cortisone | 0.33 | 1 | 2.00 |
| Androstenedione | 0.20 | 1 | 3.69 |
| Testosterone | 0.22 | 1 | 3.92 |
The range is given as the 5th, 50th, and 95th percentile interval. MoM, multiple of median.
ence MoMs. The calculated medians of the reference MoMs for each of the 10 hormones are, therefore, always equal to 1; the range for the reference MoM values is de- picted in Table 1.
MoMs in 21OHD and 11OHD Patients
The calculated MoMs for each of the 10 steroid hor- mones in the 23 CAH patients, comprising both the nine- teen 21OHD patients (21OHD cohort 1) and the five 11OHD patients, are depicted in Table 2 and Figure 1. All MoMs in Figure 1 are shown in the log10 scale.
The Wilcoxon rank sum test was performed to calcu- late significant differences between CAH patients and the reference cohort (Table 3). Figure 1 and Table 3 show that in 21OHD patients, the MoMs for progesterone, 17-hydroxyprogesterone, 21-deoxycortisol, androstene- dione, and testosterone were significantly higher and for deoxycorticosterone and 11-deoxycortisol they were significantly lower than the reference MoMs. 11OHD
| Hormone | 21OHD 11OHD | |||||
|---|---|---|---|---|---|---|
| 5% | 50% | 95% | 5% | 50% | 95% | |
| Progesterone | 0.52 | 7.83 | 176.58 | 0.24 | 2.88 | 9.81 |
| Deoxycorticosterone | 0.11 | 0.64 | 2.47 | 10.78 | 40.64 | 231.03 |
| Corticosterone | 0.02 | 0.66 | 11.66 | 0.02 | 0.66 | 9.01 |
| 17-Hydroxyprogesterone | 10.75 | 48.48 | 999.00 | 2.20 | 9.77 | 26.04 |
| 11-Deoxycortisol | 0.04 | 0.31 | 22.53 | 33.76 | 354.38 | 621.21 |
| 21-Deoxycortisol | 6.59 | 64.57 | 710.52 | 0.30 | 0.53 | 1.08 |
| Cortisol | 0.26 | 0.99 | 3.07 | 0.05 | 0.77 | 1.77 |
| Cortisone | 0.30 | 1.04 | 2.21 | 0.08 | 0.62 | 1.94 |
| Androstenedione | 1.57 | 9.23 | 91.02 | 4.35 | 102.39 | 347.54 |
| Testosterone | 0.92 | 3.54 | 16.94 | 1.78 | 15.99 | 96.87 |
The range is given as the 5th, 50th, and 95th percentile interval. MoM, multiple of median; 21OHD, 21-hydroxylase deficiency; 11OHD, 11ß-hydroxylase deficiency.
| Progester- one | Deoxycortico- sterone | Corticoste- rone | 17-Hydroxy- progesterone | 11-Deoxy- cortisol | 21-Deoxy- cortisol | Cortisol | Cortisone | Androstene- dione | Testos- terone | |
|---|---|---|---|---|---|---|---|---|---|---|
| Control versus 21OHD | ||||||||||
| Area | 0.88 | 0.73 | 0.63 | 1.00 | 0.73 | 0.99 | 0.53 | 0.52 | 0.96 | 0.87 |
| SE | 0.07 | 0.07 | 0.09 | 0.00 | 0.08 | 0.01 | 0.07 | 0.07 | 0.03 | 0.04 |
| 95% CI | 0.74-1.01 | 0.60-0.87 | 0.46-0.81 | 0.998-1.001 | 0.57-0.90 | 0.98-1.01 | 0.38-0.67 | 0.38-0.67 | 0.90-1.02 | 0.79-0.95 |
| p value | <0.0001 | 0.0006 | 0.0470 | <0.0001 | 0.0006 | <0.0001 | 0.6662 | 0.7270 | <0.0001 | <0.0001 |
| Control versus 11OHD | ||||||||||
| Area | 0.63 | 1.00 | 0.71 | 0.96 | 1.00 | 0.89 | 0.63 | 0.69 | 1.00 | 0.94 |
| SE | 0.18 | 0.00 | 0.17 | 0.02 | 0.00 | 0.10 | 0.14 | 0.13 | 0.00 | 0.05 |
| 95% CI | 0.27-0.99 | 0.996-1.001 | 0.37-1.04 | 0.92-1.01 | 1.00-1.00 | 0.70-1.08 | 0.35-0.90 | 0.43-0.95 | 0.99-1.00 | 0.85-1.03 |
| p value | 0.2570 | <0.0001 | 0.0796 | 0.0001 | <0.0001 | 0.0009 | 0.2814 | 0.1024 | <0.0001 | 0.0003 |
Italic p values indicate significant differences of steroid MoMs between 21OHD and controls and 11OHD and controls, respectively. ROC, receiver operating characteristic; MoM, multiple of median; 21OHD, 21-hydroxylase deficiency; 11OHD, 11ß-hydroxylase deficiency; SE, standard error; CI, confidence interval.
patients demonstrate significantly elevated MoMs for deoxycorticosterone, 17-hydroxyprogesterone, 11-de- oxycortisol, androstenedione, and testosterone in the presence of mostly very low MoMs for 21-deoxycortisol (Fig. 1; Table 3).
The 2 major subtypes of CAH analyzed here, 21OHD and 11OHD, therefore each show a disease-typical sig- nature using a MoM-transformed, age- and sex-in- dependent, unit-less, and hence largely simplified dis- play of the UPLC-MS/MS steroid hormone profiles (Fig. 1).
Identification of Diagnostic Single-Steroid MoMs for 21OHD and 11OHD
For diagnosing 21OHD and 11OHD, ROC curve anal- ysis was used to define the cutoff values for those hor- mones with a statistically confirmed significant p value when comparing CAH and reference cohorts (Table 3; online suppl. Table 1). First, the levels of sensitivity and specificity for the cutoffs for each hormone were com- pared with each other (online suppl. Table 1). The cutoff values with the highest sensitivity and the highest speci- ficity were chosen as potential “stand-alone” markers in-
ACC patients
Obese patients
21OHD patients
100
log MoM
1
Progesterone MoM
Deoxycorticosterone MOM
Corticosterone MOM
17-Hydroxyprogesterone MOM
11-Deoxycortisol MOM
21-Deoxycortisol MOM
Cortisol MOM
Cortisone MOM
Androstenedione MOM
Testosterone MoM
dicating the diagnosis of 21OHD or 11OHD without hav- ing to look at a complete steroid profile. We named these markers “diagnostic single-steroid MoMs.” We selected 21-deoxycortisol and 17-hydroxyprogesterone as diag- nostic single-steroid MoMs for diagnosing 21OHD. The corresponding cutoff values for these MoMs were 21-de- oxycortisol >12.28 (sensitivity 94.74%, specificity 100%) and 17-hydroxyprogesterone >3.78 (sensitivity 100%, specificity 98.83%). For diagnosing 11OHD, we selected 11-deoxycortisol as a diagnostic single-steroid MoM. The corresponding cutoff value was 11-deoxycortisol >13.18 (sensitivity 100%, specificity 100%).
Validation of Diagnostic Single-Steroid MoMs for 21OHD and 11OHD
We then validated the model of “diagnostic single-ste- roid MoMs” by using the above cutoff values for 21-de- oxycortisol and 17-hydroxyprogesterone (21OHD) as well as for 11-deoxycortisol (11OHD) in 3 independent
validation cohorts: (1) 8 newly diagnosed independent 21OHD patients (21OHD cohort 2); (2) 6 patients with documented ACC, and (3) 39 patients with documented obesity (online suppl. Table 2).
Seven out of 8 genetically proven 21OHD subjects were correctly detected by using only the diagnostic 21-deoxycortisol MoM, and all 8 patients were correctly detected by using only the diagnostic 17-hydroxyproges- terone MoM. No false-positive 11-deoxycortisol MoMs were noted. None of the obese patients were falsely as- sumed to suffer from 21OHD based on the 21-deoxycor- tisol MoM or to suffer from 11OHD based on the 11-de- oxycortisol MoM. However, the 17-hydroxyprogesterone MoM returned 3 positive results in the obesity cohort (7.7% of the individuals) (Fig. 2). Unfortunately, due to the use of anonymized data, we could not sequence the CYP21A2 gene to gain insight into the potential presence of late-onset CAH or heterozygosity of CYP21A2 muta- tions. In the ACC cohort, the diagnostic 21-deoxycortisol
MoM did not show any false-positive value. However, the 17-hydroxyprogesterone MoM gave 3 false-positive val- ues (50%), and 4 out of 6 (66.7%) ACC patients had an 11-deoxycortisol MoM higher than the cutoff value [32] (Fig. 2).
Discussion
During recent years, LC-MS/MS-based methods have become the preferred technique for determining steroid hormones in serum or plasma [4-7]. This creates multi- dimensional information complexing the understanding and interpretation of steroid data. In particular, reference ranges are still needed, especially in childhood, since ste- roid hormones change dramatically with age and puber- tal stages and often differ between the sexes [14, 24, 25, 32]. Moreover, a lack of comparability of steroid hor- mone data due to differences in laboratory methods, even with various LC-MS/MS techniques [9, 33], hampers clinical and basic science studies in endocrine research. Therefore, standardized methods “simplifying” the dis- play and interpretation of complex steroid data could im- prove diagnosis and research in endocrinology.
Different strategies of standardizing steroid profiling data have been published. Arlt et al. [34] have introduced a color coding scheme based on a learning vector quanti- fication of log-transformed gas chromatography MS urine steroid values. An alternative strategy often used in urine steroid profiles is the application of enzyme ratios, which calculate the ratios between steroid precursors and their corresponding downstream hormones reflecting the function of a defined enzymatic step. This technique has been shown to be of high diagnostic value for defined monogenetic endocrine disorders, e.g., 21OHD [35], CYP17A1 [36], and P450 oxidoreductase deficiency (PORD) [37]. Such ratios are mostly used to discriminate between affected and nonaffected subjects [38].
In contrast to interpreting steroid metabolites in urine and targeting standardization of gas chromatography MS urine steroid profile data, we here focused on a normal- ized display of plasma steroid hormone concentrations. Both approaches, plasma steroid analysis and urine ste- roid analysis, deliver different types of endocrine infor- mation. Advantages and disadvantages of the 2 methods depend on the clinical context and the diagnostic ques- tion to be answered.
We here present a reliable and easy-to-establish ap- proach to circumvent reference ranges which continually change during childhood and adolescence. MoM values
are particularly appropriate for not normally distributed values, which is an advantage, since medical data are mostly not normally distributed.
To compare MoM values between different labs and different methods, e.g., for scientific reasons, there is a continued need to establish lab- and method-specific ref- erence ranges. This may restrict the usage of MoM values to only large enough steroid labs.
MoMs have gained wide appeal for reporting maternal biochemical parameters, i.e., maternal a-fetoprotein, se- rum placental growth factor, pregnancy-associated plas- ma protein A, and progesterone, during pregnancy as well as hypercholesterolemia screening in primary care [19-23, 39].
In order to define pathologically significant hormone MoMs, we tested whether 1 steroid MoM alone was suf- ficient to establish a diagnosis of 21OHD or 11OHD in the differential diagnosis of hyperandrogenemia, a common clinical question in pediatric endocrinology. Usually, for the hormonal assessment of CAH, several steroids have to be considered as a pattern in parallel, e.g., 17-hydroxyprogesterone, androstenedione, cortisol in 21OHD [28]. We here established MoM cutoffs for “diagnostic single-steroid MoMs” predicting 21OHD (17-hydroxyprogesterone >3.78; 21-deoxycortisol >12.28) and 11OHD (11-deoxycortisol >13.18).
A potential weakness of these cutoffs is that they rely only on the comparison of previously known and mo- lecularly proven (retrospective) CAH patients with healthy control individuals. Moreover, both the 21OHD and 11OHD cohorts comprised classical forms and non- classical late-onset forms. In the clinical setting, however, the cutoffs should ideally be sensitive and specific enough to distinguish all new CAH patients from patients with other diagnoses, including individuals with overlapping symptoms of hyperandrogenemia. We, therefore, decid- ed to validate the 3 cutoffs at 2 stages.
First, we validated the sensitivity of the cutoffs in new patients with CAH. Seven out of eight 21OHD subjects were correctly detected by the 21-deoxycortisol MoM and all 8 patients by the 17-hydroxyprogesterone MoM. Un- fortunately, no independent new 11OHD patients have since then been analyzed in our lab.
Secondly, we validated the specificity of the 3 cutoffs based on a previously published obesity cohort [28] and on a cohort of 6 patients with documented ACC. With respect to 21OHD, the 21-deoxycortisol MoM appeared to be more specific than the 17-hydroxyprogesterone MoM, since none of the obese and ACC patients showed a 21-deoxycortisol MoM above the 12.28 cutoff (no false
positives). In contrast, in both the obesity and the ACC cohort, 3 patients each had 17-hydroxyprogesterone MoMs higher than the cutoff. At the endocrine level, this may either be the result of unknown 21OHD in the obe- sity cohort or due to the complexity of autonomously ex- aggerated steroid synthesis in ACC, including high 17-hy- droxyprogesterone. Accordingly, Arlt et al. [34] showed that there are elevated levels of 17-hydroxyprogesterone in some cases of ACC. With respect to 11OHD, none of the obese subjects had an 11-deoxycortisol MoM higher than the cutoff value. However, 4 (66.67%) patients with ACC had an 11-deoxycortisol MoM higher than the se- lected 11OHD cutoff. This finding probably results from a functional lack of 11ß-hydroxylase activity in the adre- nal tumor, which is consistent with earlier observations by Doerr et al. [40] and by Arlt et al. [34].
In essence, the 21-deoxycortisol MoM cutoff appears to be a sensitive and specific stand-alone diagnostic mark- er for the diagnosis of 21OHD. The high diagnostic value of 21-deoxycortisol has previously been observed by oth- ers [41] and by us [42], confirming our conclusion. In contrast, the 17-hydroxyprogesterone MoM cutoff alone - although sensitive - is not specific for 21OHD, since it can be elevated in ACC and in obesity as well. The 11-deoxycortisol MoM is of limited specificity for 11OHD due to elevated values in ACC. However, it must be noted that hormone data interpretation should never be based on one concentration or on one MoM alone, but should always depend on careful clinical evaluation narrowing down the pre-test probability of 21OHD, 11OHD, or any other diagnosis with hyperandrogenemia. Currently, we propose that MoMs for 17-hydroxyprogesterone, 21-de- oxycortisol, and 11-deoxycortisol can be sensitive tools in the initial differential diagnosis of hyperandrogenemic conditions. However, once they are above the cutoff, they need careful re-evaluation for other differential diagnoses due to limited specificity. This is particularly relevant for diagnoses like PORD, which has not been evaluated in our study and which shows functional overlap with 21OHD. As a condition of potential hyperandrogenemia, premature adrenarche would have been a very meaning- ful validation cohort as well [43]. However, the samples used in this study were retrospective and anonymous. In case of premature adrenarche, we could not sufficiently exclude other underlying diseases, and, therefore, we could not include a sufficiently defined cohort in the pres- ent study. In contrast, the used validation cohorts were proven at the molecular level in 21OHD, by clinical doc- umentation in ACC, and at clinical and biochemical lev- els in obese subjects [28]. Prospective validation of single-
steroid MoMs would profit from patients with defined premature adrenarche.
In summary, we here show that it is possible to stan- dardize complex steroid data using MoMs and that MoM calculation leads to a didactically simplified display of hormone data without age, sex, and units in routine pe- diatric endocrinology. Therefore, MoM transformation has promising potential for enhancing and simplifying the exchange of plasma steroid data nationally and inter- nationally, e.g., in clinical and basic science studies.
Diagnostic single MoMs and their cutoffs provide a distinction between pathological values and values that are only elevated. We propose that diagnostic cutoffs for specific endocrine conditions, e.g., 21OHD, 11OHD, and other diagnoses, should be discussed further and should be defined by international consensus commissions with- in endocrine societies, within international research da- tabases (e.g., I-DSD, I-CAH), or in clinical reference net- works like Endo-ERN. Such networks are able to provide the necessary high number of patients with rare endo- crine disorders needed for an improved validation of MoM cutoffs. We finally suggest that MoM transforma- tion of steroid data is a potential and promising tool con- tributing to the harmonization of laboratory assessment in pediatric endocrinology in the future [44].
Acknowledgement
The authors thank Tanja Stampe, Brigitte Karwelies, Gisela Hohmann, Susanne Olin, Sabine Stein, and Silke Struve for excel- lent technical assistance.
Disclosure Statement
The authors have no conflicts of interest to disclose.
Funding Sources
This research did not receive any specific grant from any fund- ing agency in the public, commercial, or not-for-profit sector.
References
1 Honour JW. Diagnosis of diseases of steroid hormone production, metabolism and action. J Clin Res Pediatr Endocrinol. 2009;1(5):209- 26.
2 Miller WL, Auchus RJ. The molecular biolo- gy, biochemistry, and physiology of human steroidogenesis and its disorders. Endocr Rev. 2011 Feb;32(1):81-151.
3 Soldin SJ, Soldin OP. Steroid hormone analy- sis by tandem mass spectrometry. Clin Chem. 2009 Jun;55(6):1061-6.
4 Kulle AE, Welzel M, Holterhus PM, Riepe FG. Principles and clinical applications of liquid chromatography-tandem mass spectrometry for the determination of adrenal and gonadal steroid hormones. J Endocrinol Invest. 2011 Oct;34(9):702-8.
5 Carvalho VM. The coming of age of liquid chromatography coupled to tandem mass spectrometry in the endocrinology laborato- ry. J Chromatogr B Analyt Technol Biomed Life Sci. 2012 Feb;883-884:50-8.
6 Keevil BG. Novel liquid chromatography tan- dem mass spectrometry (LC-MS/MS) meth- ods for measuring steroids. Best Pract Res Clin Endocrinol Metab. 2013 Oct;27(5):663- 74.
7 Handelsman DJ, Wartofsky L. Requirement for mass spectrometry sex steroid assays in the Journal of Clinical Endocrinology and Metabolism. J Clin Endocrinol Metab. 2013 Oct;98(10):3971-3.
8 Kulle A, Krone N, Holterhus PM, Schuler G, Greaves R, Juul A et al. Steroid hormone anal- ysis in diagnosis and treatment of DSD: posi- tion paper of EU COST Action BM 1303 “DSDnet”. Eur J Endocrinol. 2017 May; 176(5):P1-9.
9 Wudy SA, Schuler G, Sánchez-Guijo A, Hart- mann MF. The art of measuring steroids: principles and practice of current hormonal steroid analysis. J Steroid Biochem Mol Biol. 2017 Sep 28;S0960-0760(17)30250-9. Epub 2017 Sep 28.
10 Moal V, Mathieu E, Reynier P, Malthièry Y, Gallois Y. Low serum testosterone assayed by liquid chromatography-tandem mass spec- trometry. Comparison with five immunoas- say techniques. Clin Chim Acta. 2007 Nov- Dec;386(1-2):12-9.
11 Marshall W. The interpretation of biochemi- cal data. In: Livingstone C, editor. Clinical biochemistry: metabolic and clinical aspects. Philadelphia: Elsevier; 2008. p. 17-27.
12 Kraus D. Interpreting pediatric laboratory data. In: Lee M, editor. Basic skills in inter- preting laboratory data. Bethesda: American Society of Health-System Pharmacists; 2009. p. 495-519.
13 Green SF. The cost of poor blood specimen quality and errors in preanalytical processes. Clin Biochem. 2013 Sep;46(13-14):1175-9.
14 Konforte D, Shea JL, Kyriakopoulou L, Col- antonio D, Cohen AH, Shaw J et al. Complex biological pattern of fertility hormones in children and adolescents: a study of healthy children from the CALIPER cohort and es- tablishment of pediatric reference intervals. Clin Chem. 2013 Aug;59(8):1215-27.
15 Karvanen J. The statistical basis of laboratory data normalization. Drug Inf J. 2003;37(1): 101-7.
16 Horn PS, Pesce AJ, Copeland BE. A robust ap- proach to reference interval estimation and evaluation. Clin Chem. 1998 Mar;44(3):622- 31.
17 Elveback LR, Taylor WF. Statistical methods of estimating percentiles. Ann N Y Acad Sci. 1969 Sep; 161(2):538-48.
18 Harrell F, Davis CE. A new distribution-free quantile estimator. Biometrika. 1982;69(3): 635-40.
19 Wald NJ, Cuckle H, Brock JH, Peto R, Polani PE, Woodford FP. Maternal serum-alpha-fe- toprotein measurement in antenatal screen- ing for anencephaly and spina bifida in early pregnancy. Report of U.K. collaborative study on alpha-fetoprotein in relation to neural- tube defects. Lancet. 1977 Jun; 1(8026):1323- 32.
20 Wald DS, Bestwick JP, Morris JK, Whyte K, Jenkins L, Wald NJ. Child-parent familial hy- percholesterolemia screening in primary care. N Engl J Med. 2016 Oct;375(17):1628-37.
21 Vasista A, Stockler MR, West T, Wilcken N, Kiely BE. More than just the median: calculat- ing survival times for patients with HER2 pos- itive, metastatic breast cancer using data from recent randomised trials. Breast. 2017 Feb;31: 99-104.
22 Sung KU, Roh JA, Eoh KJ, Kim EH. Maternal serum placental growth factor and pregnan- cy-associated plasma protein A measured in the first trimester as parameters of subsequent pre-eclampsia and small-for-gestational-age infants: a prospective observational study. Obstet Gynecol Sci. 2017 Mar;60(2):154-62.
23 Wells G, Bleicher K, Han X, McShane M, Chan YF, Bartlett A et al. Maternal diabetes, large-for-gestational-age births, and first tri- mester pregnancy-associated plasma protein- A. J Clin Endocrinol Metab. 2015 Jun; 100(6): 2372-9.
24 Kulle AE, Riepe FG, Melchior D, Hiort O, Holterhus PM. A novel ultrapressure liquid chromatography tandem mass spectrometry method for the simultaneous determination of androstenedione, testosterone, and dihy- drotestosterone in pediatric blood samples: age- and sex-specific reference data. J Clin En- docrinol Metab. 2010 May;95(5):2399-409.
25 Kulle AE, Welzel M, Holterhus PM, Riepe FG. Implementation of a liquid chromatography tandem mass spectrometry assay for eight ad- renal C-21 steroids and pediatric reference data. Horm Res Paediatr. 2013;79(1):22-31.
26 New MI. Inborn errors of adrenal steroido- genesis. Mol Cell Endocrinol. 2003 Dec; 211(1-2):75-83.
27 Menabò S, Polat S, Baldazzi L, Kulle AE, Holt- erhus PM, Grötzinger J et al. Congenital adre- nal hyperplasia due to 11-beta-hydroxylase deficiency: functional consequences of four CYP11B1 mutations. Eur J Hum Genet. 2014 May;22(5):610-6.
28 Reinehr T, Kulle A, Wolters B, Lass N, Welzel M, Riepe F et al. Steroid hormone profiles in prepubertal obese children before and after weight loss. J Clin Endocrinol Metab. 2013 Jun;98(6):E1022-30.
29 Krone N, Braun A, Weinert S, Peter M, Roscher AA, Partsch CJ et al. Multiplex minisequencing of the 21-hydroxylase gene as a rapid strategy to confirm congenital adrenal hyperplasia. Clin Chem. 2002 Jun;48(6 Pt 1): 818-25.
30 Krone N, Roscher AA, Schwarz HP, Braun A. Comprehensive analytical strategy for muta- tion screening in 21-hydroxylase deficiency. Clin Chem. 1998 Oct;44(10):2075-82.
31 Krone N, Riepe FG, Götze D, Korsch E, Rister M, Commentz J et al. Congenital adrenal hy- perplasia due to 11-hydroxylase deficiency: functional characterization of two novel point mutations and a three-base pair deletion in the CYP11B1 gene. J Clin Endocrinol Metab. 2005 Jun;90(6):3724-30.
32 Kushnir MM, Blamires T, Rockwood AL, Roberts WL, Yue B, Erdogan E et al. Liquid chromatography-tandem mass spectrometry assay for androstenedione, dehydroepian- drosterone, and testosterone with pediatric and adult reference intervals. Clin Chem. 2010 Jul;56(7):1138-47.
33 Büttler RM, Martens F, Ackermans MT, Da- vison AS, van Herwaarden AE, Kortz L et al. Comparison of eight routine unpublished LC-MS/MS methods for the simultaneous measurement of testosterone and andro- stenedione in serum. Clin Chim Acta. 2016 Feb;454:112-8.
34 Arlt W, Biehl M, Taylor AE, Hahner S, Libé R, Hughes BA et al. Urine steroid metabolomics as a biomarker tool for detecting malignancy in adrenal tumors. J Clin Endocrinol Metab. 2011 Dec;96(12):3775-84.
35 Kamrath C, Hartmann MF, Boettcher C, Zimmer KP, Wudy SA. Diagnosis of 21-hy- droxylase deficiency by urinary metabolite ra- tios using gas chromatography-mass spec- trometry analysis: reference values for neo- nates and infants. J Steroid Biochem Mol Biol. 2016 Feb; 156:10-6.
36 Tiosano D, Knopf C, Koren I, Levanon N, Hartmann MF, Hochberg Z et al. Metabolic evidence for impaired 17alpha-hydroxylase activity in a kindred bearing the E305G muta- tion for isolate 17,20-lyase activity. Eur J En- docrinol. 2008 Mar; 158(3):385-92.
37 Koyama Y, Homma K, Fukami M, Miwa M, Ikeda K, Ogata T et al. Classic and non-classic 21-hydroxylase deficiency can be discrimi- nated from P450 oxidoreductase deficiency in Japanese infants by urinary steroid metabo- lites. Clin Pediatr Endocrinol. 2016 Apr; 25(2):37-44.
38 Caulfield MP, Lynn T, Gottschalk ME, Jones KL, Taylor NF, Malunowicz EM et al. The di- agnosis of congenital adrenal hyperplasia in the newborn by gas chromatography/mass spectrometry analysis of random urine speci- mens. J Clin Endocrinol Metab. 2002 Aug; 87(8):3682-90.
39 Lee J, Eklund EE, Lambert-Messerlian G, Palomaki GE, Butterfield K, Curran P et al. Serum progesterone levels in pregnant wom- en with obstructive sleep apnea: a case control study. J Womens Health (Larchmt). 2017 Mar;26(3):259-65.
40 Doerr HG, Sippell WG, Drop SL, Bidling- maier F, Knorr D. Evidence of 11 beta-hy- droxylase deficiency in childhood adrenocor- tical tumors. The plasma corticosterone/11- deoxycorticosterone ratio as a possible marker for malignancy. Cancer. 1987 Oct; 60(7):1625-9.
41 Costa-Barbosa FA, Tonetto-Fernandes VF, Carvalho VM, Nakamura OH, Moura V, Ba- chega TA et al. Superior discriminating value of ACTH-stimulated serum 21-deoxycortisol in identifying heterozygote carriers for 21-hy- droxylase deficiency. Clin Endocrinol (Oxf). 2010 Dec;73(6):700-6.
42 Kulle AE, Riepe FG, Hedderich J, Sippell WG, Schmitz J, Niermeyer L, Holterhus PM: LC- MS/MS based determination of basal- and ACTH-stimulated plasma concentrations of 11 steroid hormones: implications for detect- ing heterozygote CYP21A2 mutation carriers. Eur J Endocrinol. 2015 Oct; 173(4):517-24.
43 Utriainen P, Laakso S, Liimatta J, Jääskeläinen J, Voutilainen R. Premature adrenarche - a common condition with variable presenta- tion. Horm Res Paediatr. 2015;83(4):221-31.
44 Hannema SE, de Rijke YB. Improving labora- tory assessment in disorders of sex develop- ment through a multidisciplinary network. Sex Dev. 2018;12(1-3):135-9.
Horm Res Paediatr DOI: 10.1159/000488028