The diagnostic value of stimulated androgen ratios in 5-alpha reductase type 2 (SRD5A2) deficiency: a case series and review of the literature

Ethics

Institutional Review Board approval for retrospective data review was obtained from Birmingham Women’s and Children’s Hospital (BWCH) NHS Foundation Trust (reference: CARMS-31543).

Clinical data

All children with genetically established SRD5A2 in our DSD MDT service at Birmingham Children’s Hospital, UK between 2011 and 2023 were included. Clinical data include: mode of presentation, presenting age and sex of rearing. Genital ambiguity was graded according to external masculinization score (EMS) [15], assessed by a skilled physician of the DSD multi-disciplinary team (MDT).

hCG stimulation test protocol

The hCG stimulation test was performed by administering 1,500 units of hCG (Merck Sharpe and Dohme) intramuscularly daily for three consecutive days; venous blood samples were taken before the first injection (baseline/day 1) and 24 h after the third injection (day 4) for measurement of serum concentrations of androstenedione, T and DHT. Androgens were measured on validated liquid chromatography/tandem mass-spectrometry methods in our clinical service laboratory; in brief, steroids were MTBE-extracted from 250 mcL serum following the addition of deuterated internal standards and separated chromatographically using an isocratic elution profile, ionised using positive atmospheric pressure chemical ionisation and detected according to compound-specific transitions. A stimulated T:DHT ratio >10 was considered as the cut-off for establishing SRD5A2 deficiency biochemically [1], 4], 9], 14].

Urinary steroid profiling (USP)

Urine steroid profiling (gas chromatography/mass spectrometry) was carried out from spot urines in a clinical service laboratory; a diagnosis of SRD5A2 deficiency was indicated by the presence of decreased 5α-reduced steroid metabolites relative to their 5β-reduced counterparts, mainly tetrahydrocortisol (THF) and tetrahydrocorticosterone (THE). All USPs carried out in the early infancy were repeated after 3 months of age.

Molecular genetics

Karyotype analysis was performed in all children prior to molecular genetic investigations. The NHS England national genomic DSD screening panel R146.2 was utilised for molecular genetic analysis, including the SRD5A2 gene. Exome sequencing analysis including the coding regions of clinically relevant genes was carried out by next generation sequencing (Twist comprehensive exome on Illumina NovaSeq instrument). Test sensitivity for detection of small variants (SNVs and small indels) is greater than 99 % at a read depth of equal or greater than 20. SRD5A2 gene (transcript) of this gene panel and its coverage (coding regions +/− 5 bp) is SRD5A2 (NM_000348.4). Nevertheless, direct sequencing of SRD5A2 was performed when testing for a known familial mutation.

Data analysis

Non-parametric data were described as median with the range (minimum and maximum) or with percentages when appropriate. The correlations were analysed with Spearman correlation, and a p-value less than 5 % was considered statistically significant.

Diagnostic hallmarks: the role of stimulated T:DHT ratios

Raised stimulated testosterone with relatively lower DHT levels, reflected in a high T:DHT ratio, is a biochemical hallmark in SRD5A2 deficiency. The hCG stimulation test remains a key diagnostic tool in the diagnostic work-up of DSD [13], 15], 17], [23], [24], [25]. There is, however, ongoing debate on the accuracy and value of hCG-stimulated androgens in the diagnostic work-up of DSD. In the context of SRD5A2 deficiency, there is no agreed diagnostic cutoff for the stimulated T:DHT ratio with expert suggestions ranging from 10 to 100, although a cutoff of 10 has been widely employed [9], 11], 13], 26]. Maimoun et al. demonstrated that 72 % of their cohort exceeded a ratio of 10 [9] whilst a large Chinese study involving 103 patients reported nearly 90 % above 10 [1]. Our cohort’s T:DHT ratio of 80 % falls within these previously reported ranges.

Lower cut-off values, such as 8.5 in infancy, have been suggested to improve the diagnostic sensitivity [26]. However, this lower threshold has failed to show an added advantage in establishing the diagnosis in our cohort and previous studies [27]. Additionally, comparatively lower ratios have been observed during puberty, which is thought to be due to increased pubertal steroid 5α-reductase type1 expression. Equivocal results have also been reported, highlighting the diagnostic difficulties [9], 14], 26], 28], 29].

Lower hCG-stimulated T levels might lead to a lower T:DHT ratio, exemplified by patient 3 in our cohort (Table 1). Hiort and coworkers have proposed prolonged hCG stimulation, yielding higher peak T levels and consequently higher diagnostic value of the T:DHT ratios [30]. Nevertheless, patient 14 in our cohort who achieved a high testosterone level of 24.9 nmol/L did not exhibit a correspondingly elevated T:DHT ratio.

Lastly, our findings revealed no correlation between stimulated androgen ratios and phenotypic severity, aligning with previously reported data [16], 20].

Table 2 provides an overview of case series and larger cohorts of 46, XY DSD patients, evaluating the utility of hCG-stimulated T:DHT ratios in distinguishing SRD5A2 deficiency from other DSDs, such as androgen insensitivity syndrome (AIS). Most studies have concluded that stimulated T:DHT ratios lack specificity for differentiating SRD5A2 deficiency from other 46, XY DSDs, as there is significant overlap which can lead to misdiagnosis in some cases [27], [31], [32], [33]. Sensitivity for a cutoff ratio >10 has been reported to be below 80 % [9], likely due to the absence of robust, age- and sex-specific reference values for androgen ratios. The condition’s rarity and small sample sizes further limit the validation of these findings. Further limitations to the accuracy, reproducibility and comparability of the tests are possibly related to the different test protocols that exist [27], and the various steroid hormone assays employed. Immunoassays can be particularly inaccurate in patients outside (mini-)puberty [16]. The measurement of steroid hormones with mass spectrometry-based assays are more reliable, particularly in small sample volume [17]. In our study, androgens were analysed with LC-MS/MS, deemed to be most accurate for all steroid hormone measurements.

The role of urinary steroid profiling

Urine steroid profiling (USP) by GC/MS is a reliable method of diagnosing SRD5A2 deficiency, identifying 5-alpha reduced steroid metabolites including tetrahydrocorticosterone and tetrohydrocortisol with low 5α to 5β reduced metabolite ratios [16]. When conducted after three months postnatally, USP has demonstrated high diagnostic accuracy [6], 7], 12], 34], although variations within reference ranges have been reported in different studies, largely due to different assay methodologies and ethnic variations. In our cohort, USP was confirmatory in all where applied, including those where the stimulated T:DHT ratio was below 10. High accuracy was reported in published case reports or series, but USP has not been tested systematically in larger samples, likely due to limited availability and analytical expertise in most healthcare (and research) settings [7], 14], 34]. It has its limitations in the neonatal and early infancy period, as the altered steroid metabolome is challenging to interpret due to the influence of foetal adrenal-derived 11-oxocortisone metabolites [35]. Although its availability is limited, it may be a more accurate, less invasive and less burdensome first-line test to establish the biochemical diagnosis of SRD5A2 deficiency.

Diagnostic pathways: current challenges and future directions

In children with 46, XY DSD, reaching the correct diagnosis in a timely manner is crucial to facilitate personalised management. The ‘traditional’ diagnostic pathways for investigation of DSD [15] begin with phenotypic characterisation, followed by karyotyping to define the DSD category, then biochemical testing and, lastly, targeted molecular genetic testing based on the interpretation of clinical and hormonal data. Traditionally, these stratified approaches result in a molecular diagnosis in only about half of individuals with 46, XY DSD [36], 37]. There are no widely published data on the average time a stratified approach takes in patients with a finally confirmed diagnosis; but delays are evident. For instance, a Turkish single-centre study reported an average diagnostic time of 5.9 years and 27 clinic visits, while an Italian cohort of children with SRD5A2 deficiency showed a mean diagnostic delay of 9.1 years. These delays have significant implications for affected families and healthcare systems [4], 38]. Consequently, studies have suggested that targeted genetic testing may facilitate earlier and more accurate diagnoses and widen the mutational spectrum of causative genes, warranting further exploration in developing new diagnostic strategies [4], 12], 39]. The time from gene extraction to finalising data analysis, which includes a complex but largely automated stratification of the potentially damaging effects of identified sequencing variants, takes approximately 6–8 weeks in our centre but may be faster in other settings. The time and cost are expected to decrease with advancements in bioinformatics. Importantly, the overall economic burden of hospital appointments may be reduced if a definitive diagnosis is achieved earlier [40].

Nevertheless, the inconsistent association between clinical phenotypes, molecular findings and diagnosing variants of unknown significance (VUS) still necessitates complementary hormonal and biochemical evaluations. Considering the less accurate results, poor patient experiences due to the painful and prolonged 4-day procedure, and the economic burden on healthcare resources and families attending multiple appointments, we discourage relying on the hCG-stimulation test as the primary biochemical diagnostic tool. Instead, we advocate for leveraging the mini-puberty window to assess endogenously stimulated androgen levels and promote simultaneous clinical, biochemical, and genetic workups as the preferred approach to accelerate the diagnostic work-up in DSD.

Genetic insights and mutational spectrum

Over 100 pathogenic mutations in the SRD5A2 gene have been described, with new variants continuing to emerge in the literature [4], 16], 41], of which about 65 % are familial homozygous variants [5], 9], 11], 40]. Some mutations have been reported more commonly in specific ethnic groups, such as p.(Gly34Arg) in Egypt and p.(Pro212Arg) in Mexican patients [41], 42]. Nevertheless, variable inheritance patterns, including uniparental disomy, have been reported [43]. In cases where heterozygous variants were identified along with biochemical evidence of raised stimulated DHT:T ratios, a dominant effect or the possibility of unidentified mutations (i.e. deep intronic or within the promoter regions) has been suggested [4], 23], 42]. Some genetic investigations can be offered antenatally in suspected pregnancies, but comprehensive postnatal assessment is crucial to establish the diagnosis [44].

In our cohort, nine distinct mutations were identified with homozygosity rates consistent with the literature: 71 and 28 % for homozygous and compound heterozygous variants, respectively [9], 29]. The most commonly identified variant in our cohort was p.(Arg246Gln), followed by p.(Glu200Lys), and p.(Arg227*). The p.(Arg246Gln) variant has been previously predominantly reported in Indian patients [39], 45] whereas p.(Glu200Lys) has been observed less frequently in other studies [16], 46]. Notably, our cohort is also of Asian ethnic origin. One had a novel variant, p.(Cys88Trp) which was initially reported as VUS and later reclassifies as likely pathogenic with the clinical phenotype and biochemical results. Understanding the spectrum of SRD5A2 mutations is essential for refining diagnostic approaches and developing more targeted therapies. Despite biallelic nonsense mutations consistently thought to be associated with a severe phenotype, most studies have failed to identify a clear genotypic and phenotypic correlation [9], even with the same mutations such as p.(Gly34Arg) and p.(Leu55Gln) [9], 41]. Conversely, recent larger studies have shown some phenotypic correlation for the p.Arg227Gln variant being associated with a milder phenotype (higher EMS and AMH levels), whereas a severe phenotype is found in individuals with homozygocity for p.(Arg246Gln), p.(Gly203Ser) or p.(Gln6*) [1], 20].