Urine Steroid Metabolomics as a Novel Tool for Detection of Recurrent Adrenocortical Carcinoma

Figure 1.. Study recruitment flow chart.

Figure 2.. (A) All longitudinal urine samples collected from patients who developed disease recurrence, plotted against time from surgery. Blue dots represent postoperative samples collected while the patient was considered disease-free according to their most recent clinical and radiological evaluation. Red dots represent samples collected after the first radiological manifestation of recurrent disease and before any second curative therapy. Purple dots represent samples that were collected preoperatively. (B) Heat-map visualization of longitudinal urinary steroid profile results in 5 representative patients who developed recurrent disease during follow-up. Arrows indicate the time of the first radiological manifestation of recurrent disease (Rec) and surgery for recurrence (Sx). Steroid numbers correspond to steroid metabolites as tabulated in Table 1.

Figure 3.. Comparison of preoperative (baseline) and post-recurrence samples (13 patients with a total of 15 recurrences). (A) Overlap between baseline urine steroid profile and the profile of the first post-recurrence sample provided by the same patient, when considering the 6 most elevated steroid metabolites in each sample. Steroid values were normalized to the upper limit of the reference range of the corresponding steroid metabolite in sex-matched healthy controls. (B) Frequency of inclusion of each steroid metabolite in the “top 6” most elevated steroid biomarkers. The presented panel consists of the 8 steroid biomarkers previously described as part of the “malignant steroid fingerprint” diagnostic for adrenocortical carcinoma (15).

Figure 4.. Assessment of longitudinally collected urine steroid profiling results by 3 expert clinicians (C1–C3). (A+B) Clinician assessment in the 32 patients who developed disease recurrence, grouped according to whether they had provided preoperative urine samples (n = 13) or not (n = 19) (A) or whether they had adjuvant mitotane treatment at the time of recurrence (n = 12) or not (n = 20) (B). (C) Clinician assessment of urine profiles from 31 patients who remained disease-free for at least three years post-operatively and provided at least two post-operative samples. (D) Time interval between the detection of recurrent adrenocortical carcinoma by imaging and the earliest detection by clinician assessment of urine steroid profiles. Each point corresponds to a single urine sample; no patient is represented by more than one sample. Negative values indicate that biochemical detection preceded radiological detection.

Figure 5.. Machine learning-based analysis of the urine steroid profile results. (A) Receiver operating characteristics curve analysis of the performance of random forest classification in distinguishing post-recurrence samples from samples provided by non-recurred patients. The performance of the three clinician assessors (C1–C3) has also been plotted for comparison. Steroid numbers correspond to Table 1. (B) Random forest assessment of the relative importance of the 19 steroid metabolites in distinguishing post-recurrence samples (n = 32) from postoperative samples provided by non-recurred patients (n = 39), quantifying the significance of each single steroid marker for the detection of adrenocortical carcinoma recurrence, with all significances adding up to the sum of 1.