This study provides the results of a pilot phase II RCT confirming that a single 10,000 IU/kg dose can rapidly and safely normalize plasma 25(OH)D concentrations in critically ill children identified as vitamin D deficient (< 50 nmol/L) during admission to the PICU, and that proceeding with a Phase III trial is feasible.
This study recruited vitamin D deficient children from 4 PICUs and 1 NICU at 4 academic centers in three countries (Canada, Austria and Chile). The median plasma 25(OH)D concentration of participants at screening was ~ 35 nmol/L, with evaluation of baseline patient characteristics demonstrating the study cohort to be severely ill, with 80% requiring mechanical ventilation and 50% receiving vasoactive agents. The loading dose regimen was observed to rapidly increase blood levels within 1 day, and peak on days 2 and 3, with 80% of participants exceeding the target blood threshold of 75 nmol/L. These findings are similar to those studies with comparable doses administered to critically ill adults with admission 25(OH)D levels under 50 nmol/L. For example, in the VITdAL study Amrein and colleagues reported an increase of 50 nmol/L following a 540,000 IU cholecalciferol dose, with 52% exceeding 75 nmol/L[32]. Similarly, in the VIOLET trial, an adult acute lower respiratory tract infection population at high risk for acute respiratory distress syndrome and ICU admission increased blood 25(OH)D concentrations from an average of 28 to 117 nmol/L (75% exceeding 75 nmol/L) with a 540,000 IU enteral dose[53]. While there are pediatric trials evaluating loading dose cholecalciferol regimens, none have evaluated a similar age or weight base dosing regimen on critically ill children. The most related study, published by Sahu and colleagues in 2019, evaluated the 10,000 IU/kg dose recommended from our systematic review, in a cohort of VDD children undergoing cardiac surgery for congenital heart disease (Tetralogy of Fallot)[54]. Administration of the cholecalciferol dose approximately two weeks prior to surgery produced significantly higher concentrations in the treatment arm relative to control (83.5 vs. 27.4 nmol/L) in blood collected immediately before the surgery. Given that blood 25(OH)D concentrations will peak ~ 72 hours following loading dose administration and then begin to decline[38], and the known half-life of vitamin D, blood 25(OH)D concentrations would have been higher if they had been measured closer to loading dose administration.
In addition, this study also assessed the feasibility of the weight-based loading dose regimen through an evaluation of vitamin D toxicity data. Loading dose cholecalciferol, particularly when using excessive or repeated doses, particularly involving infants and young children, has been linked to hypercalcemia, nephrocalcinosis and renal failure[55]. Critically ill patients may be at greater risk for adverse events due severe organ dysfunction, higher prevalence of genetic abnormalities, and potential interactions with common ICU medications[56, 57]. Despite these concerns, we did not document any cases of persistent hypercalcemia, consistent with the low hypercalcemia rate (2.6%) observed in a meta-analysis of data from pediatric interventional trials[42]. These findings also align with the low rates of hypercalcemia reported in the adult VITdAL-ICU (≤ 1%) and VIOLET (≤ 3%) trials, with no differences evident between control and intervention arms[32, 53]. Similarly, in their pediatric cardiac surgery trial, Sahu and colleagues did not report higher blood calcium levels in the arm receiving a 10,000 IU/kg pre-operative loading dose, when compared to the control group[54]. Evaluation of urine calcium and nephrocalcinosis data in our trial again failed to identify reason for concern with hypercalciuria rates similar between treatment (15%) and usual care arms (21%). While hypercalciuria rates were similar between groups, the baseline rate was considerably higher than the pooled 2.5% rate reported in the systematic review of pediatric vitamin D clinical trial data that did not include critically ill children[38]. While high in comparison to healthier populations, this difference was not unexpected in the pediatric critical care setting due to the significant presence of systematic inflammation, renal dysfunction, and medication use (e.g. diuretics). Again, findings were very similar to those presented in the adult VITdAL-ICU trial, where both arms had 25% hypercalciuria rates both before and following administration of the 540,000 IU cholecalciferol load[32]. Sahu and colleagues also reported reassuring findings with average post cardiac surgery calcium:creatine ratios being similar or potentially even lower in the cohort receiving the cholecalciferol load (2.0 vs 1.1, p = 0.16)[54]. All of this suggests that hypercalciuria is not an appropriate biomarker of excess vitamin D levels in the ICU setting. Finally, further suggesting the safety of this regimen, was the absence of any definitive cases of nephrocalcinosis in the treatment arm or serious adverse events potentially related to Vitamin D.
An important study observation relates to significant variability in post loading dose 25(OH)D concentrations. Eight out of 10 of participants in the treatment arm exceed the 75 nmol/L threshold following administration of the single loading dose. However, while the group average 25(OH)D concentration achieved was successful, individual patient responses to the loading dose were variable with a small number of patients exhibiting minimal change and not reaching plasma 25(OH)D concentrations > 75 nmol/L (n = 7). In addition, a small number of patients achieved plasma concentrations > 200 nmol/L (n = 4) or > 250 nmol/L (n = 2). Significant variability in post-study drug concentrations has been previously recognized, with regression analysis of data from pediatric interventional vitamin D trial literature calculating the SD to average 42% of the mean 25(OH)D[38]. The calculated SD of ~ 63 nmol/L on a mean plasma 25(OH)D concentration of ~ 126 nmol/L and the observation that ~ 20% of study participants had concentrations either below 50 nmol/L or above 200 nmol/L is consistent with these previous findings. While less variability might have been anticipated with application of a weight-based single dose loading regimen, differing degrees of amounts of critical care malabsorption may have counterbalanced. These findings are also in line with the post-loading dose SD of 52 nmol/L (day 7) and 58 nmol/L (day 3) in the VITdAL and VIOLET trials, respectively[32, 53]. Both VITdAL (24%) and VIOLET (12%) also reported a significant number of non-responders, defined as post-drug 25(OH)D concentrations below 50 nmol/L. Comparing results with respect to elevated 25(OH)D is more challenging due to application of different thresholds across studies. In the VITdAL analyses, Amrein et al. applied a 150 nmol/L threshold, reporting 13% of study participants above this level[32]. Importantly, 150 nmol/L does not represent a level associated with acute toxicity, but represents a threshold above which it is difficult to achieve with sunlight and most diets. Amrein and colleagues report the two highest 25(OH)D concentrations in their treatment arm as ~ 265 nmol/L, further indicating this to be well below the 375 nmol/L threshold where risk of acute toxicity may begin to rise[58, 59]. With respect to the VIOLET trial, the upper target 25(OH)D concentration was set as 300 nmol/L, with only one patient exceeding that value[53]. Similar to these studies, the two highest clinical 25(OH)D measurements in our pilot study were 343 and 275 nmol/L, and importantly, there were no symptoms of vitamin D toxicity. Our protocol indicated a change in dose would be considered if > 10% of the participants receiving the loading dose achieved a 25(OH)D concentration above 250 nmol/L[47], and this was not exceeded with the clinical samples. As mentioned above, acute toxicity is not believed to be a problem until 25(OH)D concentrations exceed ~ 375 nmol/L[58, 59]. As the first evaluation of this dosing regimen in a high-risk population, we made the decision to use a more conservative threshold of 250 nmol/L. Although our aim was to avoid 25(OH)D concentrations > 250 nmol/L, we still recommend proceeding with this dosing regimen for a subsequent large-scale trial. Vitamin D toxicity is time-dependent, and transient levels > 200 nmol/L do not appear relevant[59]. Importantly, there were no cases of persistent hypercalcemia, clinically significant hypercalciuria, nephrocalcinosis or any other adverse events related to vitamin D supplementation observed in this pilot study. Further, reducing the dosing regimen would increase the number of patients who do not achieve or maintain post-supplementation 25(OH)D concentrations > 75 nmol/L and dilute the impact of the loading dose regimen on clinical outcome in a Phase III trial. This decision was also informed by observations by our group, and others that 25(OH)D concentrations peak 72 hours following loading dose administration and then rapidly begin to fall[38], with an average decline of 10 nmol/L week observed in our systematic review of pediatric loading dose trials[38]. For example, Thacher et al. found that 25(OH)D3 concentrations fell by 53% and 59% in rachitic and healthy children 14 days following administration of a loading dose of 50,000 IU[60].
The results of this phase II trial support the feasibility of large-scale, multicentre phase III trial powered for clinical outcome. We met our a priori established feasibility criteria for protocol adherence, blinding, study withdrawal and patient accrual. The patient accrual rate in this study was 3.4 patients per month (~ 1.9 patients/month/site). However, with the exception of the lead site, sites in this study were only recruiting for ≤ 5 months, while peak recruitment in critical care RCTs is generally not achieved until at least 7 months after a site initiates recruitment[61]. It is likely then that the accrual rate in the sub-sites site would have been trended higher over time if these sites had been active for a longer period of time. Therefore, we believe that a reasonable expected accrual rate for a multi-year, multi-site RCT would be 2 patients/month/site, which is well within the range observed in other completed large, multi-centre PICU RCTs performed in the last decade by large research consortiums[62–69]. Of note, this expected accrual rate is dependent on a VDD prevalence consistent with that observed during this pilot study. Our ability to collect blood samples at Day 7 was slightly lower than anticipated (84% vs > 95%). In this trial, the Day 7 blood sample was essential to establish the efficacy of the dosing regimen, and to evaluate for toxicity; however, sample collection would be less important for a more pragmatic Phase III trial focused on clinical outcome. Given that 25(OH)D concentrations peak ~ 72 hours following loading dose administration, adjusting the protocol to allow post-drug sample collection anytime between Day 2 to Day 7 would increase the frequency of blood sample collection, as clinical bloodwork frequency tends to be higher earlier during a PICU admission. Importantly, this would allow a future Phase III trial to also perform a metabolomic sub-study to supplement the existing adult literature[70–72].
Recognizing the importance of patient-centred outcomes in clinical trials, we engaged with families participating in this pilot study to seek input on the primary outcome for the Phase III trial. This was done concurrently with a survey of healthcare providers and caregivers by Merrit and colleagues to identify the primary outcome for a multi-centre RCT in pediatric septic shock[73]. Feedback from families participating in this trial and the survey by Merrit et al indicates that health-related quality of life is the most important outcome for families in a Phase III clinical trial in critically ill children[73]. Historically, PICU RCTs have used hospital-based outcomes (e.g. length of stay in PICU, hospital mortality) that allow primary outcome data collection for 100% of enrolled patients. In contrast, previous observational studies evaluating health-related quality of life in PICU patients at ≥ 1 month have reported completion rates of 52 to 79%[74–76]. Using a primary outcome that will be collected following hospital discharge for the Phase III RCT means that some participants will be lost to follow-up, and that this will need to be accounted for the sample size calculation. Further, strategies to maximize follow-up will be essential, such as collecting multiple points of contact information, including locators; pre-notifying participants of an upcoming follow-up visit; making additional attempts to contact if the first attempt is not successful; monitoring loss to follow up to identify concerns early; and the addition of incentives for completing questionnaires[77–79].
CONCLUSION
A single 10,000 IU/kg dose can rapidly and safely normalize plasma 25(OH)D concentrations in critically ill children with VDD during admission to the PICU, but with significant variability in post loading dose 25(OH)D concentrations. We established that proceeding with a Phase III multicentre trial to evaluate the impact of this loading dose regimen on clinical outcomes is feasible. Importantly, using a patient oriented primary outcome determined after hospital discharge such as health-related quality of life will require thoughtful implementation of strategies to minimize loss to follow-up. Based on the results of this pilot study, we launched the Phase III trial in June 2019 (NCT03742505), and as of Febraury 2023, have enrolled 263 patients across 9 participating Canadian PICUs.