The ECG data included in this study were obtained at nine clinical sites according to the standard operating procedures (SOPs) of the individual sites. There was no placebo treated subject with time-matched ECG-PK data available at the time of conducting this analysis. Hence, the placebo-adjusted and baseline-corrected QTcF change (ΔΔQTcF) could not be derived. There was no central ECG assessment for these studies. Thus, the consistency of ECG measurements across studies/sites and the sensitivity of pooled data to detect potential QT prolongation were to be confirmed during the analysis.

To address the potential bias from pooling, a separate C-QT analysis has been conducted based on study PTC923-MD-005-HV alone, for which data were obtained from 32 healthy adults from a single site and every subject received a low dose of 20 mg/kg/day and a high dose of 60 mg/kg/day sepiapterin, under either the fasted condition or with either a low-fat or a high-fat diet. A similar negligible trend of BH4-ΔQTcF relationship and magnitude of ΔQTcF (upper 90% CI of -3.04 ms at 60 mg/kg with a high-fat diet) was predicted based on this model (data on file). The concordance of the results indicated that there was consistent measurement across studies and sites and no obvious bias while pooling data across studies. Assessment of study ID as a covariate during the C-QTcF analysis of this study indicated that it was not significant. Thus, it was concluded that there was no apparent bias between sites and studies. Study ID was explored as a covariate during model development and was found not significant. Hence, it was not retained in the final model.

The hysteresis analysis was conducted for both BH4-ΔQTcF and sepiapterin-ΔQTcF and the data showed that no ΔQTcF values exceeded 5 ms at specific diurnal time points, and the time difference between Umax and Cmax was less than 1 h (Fig. 3). These results indicated no apparent hysteresis or time delay between BH4 or sepiapterin concentrations and changes in QTcF. The ΔQTcF occurrence and baseline (predose)-corrected ΔQTcF are sufficient to assess the risk of QTc prolongation for this situation.

Food intake is known to reduce QTcF by 5–10 ms, with the maximum effect at around 2 h and reduced to negligible by 8 h postprandially [20,21,22]. It has been proposed that the food effect on ECG could be used as a study sensitivity check for studies lacking a positive control. Most of the data collected for this analysis were post sepiapterin administered with food (49.4% with a low-fat diet and 36.3% with a high-fat diet). As presented in Table S5, at the dosage of 60 mg/kg/day sepiapterin, the reduction in QTcF was 1.79 ms for the low-fat meal condition and 0.82 ms for the high-fat meal condition, while there was an increase of 0.71 ms in QTcF for the fasted state. The food impact was also assessed using an LME model. Compared to the fasted condition, the ΔQTcF was reduced by 5.47 and 3.76 ms following a low-fat and high-fat diet, respectively (Table 3). A similar magnitude ΔQTcF reduction of 5.52 and 3.97 ms was obtained based on sepiapterin-ΔQTcF correlation for low-fat and high-fat diet, respectively. These results demonstrate that the pooled data and this analysis have sufficient sensitivity to detect potential QTcF prolongations and that the data are consistent across different sites.

This analysis included seven patients with PKU aged < 18 years old, of whom two had ages 2–5 years and four had ages 6–11 years. The Bazett’s (QTcB) method has been concepted to have better performance for pediatrics. Thus, QT interval corrected using Bazett’s method was explored with baseline RR, and a significant residual trend was identified (Fig. 5). There were no residual trends detected in QTcF or population corrected QT (QTcP) versus baseline RR. Hence, the final analysis was conducted with QTcF.

The in vitro hERG studies with sepiapterin showed minimal hERG inhibition (< 20%) at the highest tested condition, 30 µM (7,116 ng/mL) for sepiapterin and 866 µM (208,810 ng/mL) for BH4. These concentrations are 2524-fold and 609-fold, respectively, of the sepiapterin and BH4 plasma Cmax in patients with PKU receiving the highest clinical dose of 60 mg/kg/day. There was a probability of 0.89 or greater that sepiapterin would not cause QT prolongation clinically, based on a half-maximal inhibitory concentration (IC50) > 30 µM, and the probability of ≥ 0.87 based on the multiples of exposures from the retrospective analysis of other compounds [23]. Strong correlation of hERG inhibitory potency and drug-induced arrhythmia has been reported in many literature sources [24, 25]. A safety margin (hERG IC50/free therapeutic drug concentration) of 30-fold is frequently used to predict human cardiac risks. However, a safety margin greater than 30-fold does not rule out the potential for clinical QT risk. Multiple compounds with exceptions have been identified, such as terfenadine (93-fold), tacrolimus (700-fold), and amiodarone (1400-fold), due either to metabolic inhibition resulting in elevated plasma concentration or to non-hERG-mediated mechanisms. Hence, the much greater than 30-fold safety margin of sepiapterin in hERG analysis alone is not sufficient to rule out potential QT risks. However, in combination with clinical observations, and the negligible trend of QT prolongation with either BH4 or sepiapterin concentration demonstrated in this study, and an upper 90% CI of 0.25 ms QTcF prolongation at 2× the therapeutic peak plasma total concentration in patients (Table 4), collectively, these data strongly support the conclusion that there is very limited cardiovascular risk of QT prolongation with sepiapterin. This is further supported by clinical observations from all studies; among 388 subjects, there were no treatment-emergent adverse events (TEAE) in the cardiac system disorders following sepiapterin treatment (Table S5).

As proposed by the cardiac safety research consortium, QT prolongation could be assessed using C-QTc modeling based on early-phase clinical data, for example, first-in-human studies. This proposal was subsequently adopted in the ICH E14 guidance and accepted by regulatory agencies. The guidance elaborates that concentrations achieved in these studies should be a sufficient multiple (commonly 2×) above the exposure at the maximum therapeutic dose at steady state, which should reflect high clinical exposure scenarios, and that data could be pooled from multiple studies. The high clinical exposure scenario refers to situations “such as drug-drug and drug-food interactions, organ dysfunction, and/or genetically impaired metabolism.” The C-QTc analysis could serve as an alternative approach to the mandated thorough QT (TQT) assessment if such analysis demonstrated that the upper bound of the 90% CI was < 10 ms at the highest clinically relevant exposures. This alternative approach finds support in various publications. For example, the C-QTc model analysis of degarelix revealed no QT prolongation effects, which was later confirmed by a TQT study [26]. These findings advocate for integration of QT assessment into phase 1 dose escalation studies, potentially eliminating the need for dedicated TQT studies when no QT effect is observed.

The high clinical exposure scenario may not always be clearly defined at the time of conducting such C-QTc analysis, especially when there is no immediate need to conduct clinical studies in patients with renal/hepatic impairment. However, we believe that such scenarios have already been demonstrated in support of the present analysis, based on the overall absorption, metabolism, and elimination properties of sepiapterin and BH4, as well as the saturable absorption of sepiapterin observed in previous clinical studies [10]. In adult healthy volunteers, when the dose was increased 3-fold from 20 to 60 mg/kg, BH4 Cmax increased from 425 to 677 ng/mL, only 59%, and AUC0 − 24 h increased from 3070 to 5230 h*ng/mL, only 70%. This far less than dose proportional increase suggests that absorption was approaching saturation at doses above 20 mg/kg.

Clinical studies in patients with renal/hepatic dysfunction were not conducted at the time of this analysis. However, renal and hepatic dysfunction are not a comorbidity of PKU, and data indicate renal and hepatic clearance in PKU patients will not be the limiting factor: the estimated GFRs from all PKU patients (n = 157) in the phase 3 study were greater than or equal to 87 mL/min/1.73 m2, serum bilirubin [median(min, max)] was 0.516 (0.128, 2.52) mg/dL, and serum albumin was ≥ 3.6 g/dL.

The renal clearance of total radioactivity derived from 14C-sepiapterin and metabolites was 1.536 L/h (25.6 mL/min) following an oral administration of 4000 mg sepiapterin (containing 100 µCi 14C-sepiapterin) to adult male healthy subjects (data on file), which was less than the lower bound GFR of patients with moderate renal dysfunction (30 mL/min). Sepiapterin and BH4 metabolism are mediated both nonenzymatically and enzymatically by multiple enzymes (such as sepiapterin reductase, carbonyl reductase, dihydrofolate reductase, pteridine-4α-carbinolamine dehydratase, dihydropteridine reductase, aromatic amino acid hydroxylases, nitric oxide synthases, and xanthine oxidase) [27,28,29]. Metabolism can occur in multiple tissues, including liver, kidney, adrenal gland, muscle, and brain. Extensive metabolism of sepiapterin was observed with no major dominant metabolites other than BH4. Hence, it is not expected that significant increases in BH4 exposure will occur in patients with mild or moderate renal or hepatic dysfunction, and the risk of significant increase in BH4 exposure due to drug-drug interactions with sepiapterin/BH4 metabolism is limited, though this needs to be confirmed clinically in studies in patients with hepatic/renal dysfunctions.

BH4 Cmax (677 ng/mL) in healthy subjects who received 60 mg/kg/day sepiapterin, administered with a high-fat diet, was 1.97-fold of that in patients with PKU (Cmax 343 ng/mL) who received 60 mg/kg/day sepiapterin with food (Phe-restricted diet specifically required for patients with PKU). This is most likely due to the food effect. Coadministration of food with sepiapterin increases sepiapterin and BH4 exposure. BH4 Cmax was 1.72-fold and 2.21-fold higher when sepiapterin was administered with a low-fat and a high-fat diet, respectively, compared with the fasted condition at dose 60 mg/kg [10]. All patients were instructed to take sepiapterin with food in the phase 2 and phase 3 clinical studies. However, PKU patients are all managed through a stringent Phe restriction diet, which approximately provides 26% energy from fat, similar to the standard low fat diet [14]. This provides the 1.97-fold margin of BH4 concentration for this study to assess potential QT prolongation risk in patients with PKU at the highest planned therapeutic dose of 60 mg/kg/day.

In light of the aspects discussed above, it is not expected that a significant elevation of plasma BH4 concentration would occur as a result of metabolic inhibition or moderate renal or hepatic dysfunction. Nevertheless, dedicated clinical studies would be required to confirm such predictions. If a significant increase in BH4 exposure is observed from patients with renal/hepatic dysfunction, a dose adjustment would be recommended to ensure similar BH4 exposure.

Additionally, BH4 Cmax increased far less than dose proportional at a dose above 20 mg/kg/day [10]. A linear extrapolation based on data obtained in the dose range 20–60 mg/kg/day predicted that increasing the dose from 60 to 120 mg/kg would increase BH4 Cmax by only an insignificant 22%, from 677 to 828 ng/mL. This is likely the highest feasible dose, which requires to suspend 48 g of drug product into 240 mL water for administration for an adult with the typical body weight 70 kg with a high fat diet. Therefore, we think that a sufficiently high clinical exposure scenario for BH4 has already been demonstrated, although there is a lack of renal/hepatic impairment clinical data. This is further strengthened by the negligible but slight negative slope (0.00071 ms/ng/mL) of the central trend of BH4 C-QTcF relationship and the upper bound 90% CI for ΔQTcF of 0.25 ms in healthy volunteers at BH4 Cmax, which was 1.97-fold the mean BH4 Cmax in patients with PKU (Tables 3 and 4, and 5).

It will be interesting to see more related reports and discussions regarding the practical definition of high clinical exposure scenario and demonstration of sufficiently high multiples of clinical exposure in additional clinical development programs.

During this study, the parent compound sepiapterin-QTcF correlation was also investigated per request from the US FDA, although the Cmax and AUC of sepiapterin were < 2% of those of BH4. The results indicated that there was a negligible but slightly negative central trend with a slope of 0.193 ms/(ng/mL) (Table S4) and an upper bound of 90% CI for ΔQTcF of -1.77 ms at the sepiapterin Cmax of 4.66 ng/mL in healthy volunteers (Table 5). Since exposure to sepiapterin (Cmax and AUC0 − 24 h) was < 2% that of BH4 in both healthy volunteers and patients with PKU and there was no clinical observation of any subject with > 450 ms QTcF or QTcF change from baseline > 30 ms, such analysis provides very limited information when no QT prolongation has already been demonstrated in the analysis of the major active metabolite, BH4. The Metabolites in Safety Testing (MIST) guidance requires only those metabolites present at greater than 10% of total drug-related exposure at steady state be assessed [30]. A similar approach, using either the same (10%) or a different threshold, should be considered when taking the concentration-response approach to assess cardiac risks.