Main

Inositol is ubiquitous in living organisms where it is largely present as a free sugar alcohol and also as a head group of membrane lipids. In addition, phosphoinositides and glycosylphoshatidylinositols have specific roles in signal transduction and in lipid–protein interactions (1,2,3,4). In its free form, inositol is a sugar alcohol present in human milk, widely available in the diet, and it has been classified as “generally regarded as safe” for enteral administration by the US Food and Drug Administration. In utero, early fetal serum inositol levels are 2–10 times higher than adult levels and decrease gradually toward term (5,6,7). Inositol has been supplemented in infant formulas since the late 1990s at ~44 mg/100 kcal (350 mg/l), yet its clearance has not been studied in extremely preterm newborns. If an infant is not receiving enteral milk feedings, serum levels substantially decrease below the in utero levels at a comparable age. (5,6).

Hallman et al. (8,9,10,11) administered myo-inositol intravenously (i.v.) and/or enterally to preterm infants with respiratory distress syndrome in two randomized trials and found increased survival, increased survival without bronchopulmonary dysplasia, reduced severe retinopathy of prematurity, reduced severe intraventricular hemorrhage (IVH), and no observed toxicity. An additional partially randomized trial of inositol-supplemented formula in preterm infants (12) was also considered in the Cochrane review of inositol. The review concluded the combined findings warrant further study, particularly in the extremely preterm infants who remain at highest risk for these morbidities (13).

The endogenous production and metabolism of inositol, combined with dietary intake, add complexity to a pharmacokinetic (PK) analysis. However, to inform dosing of i.v. inositol in extremely preterm infants, we need to improve our understanding of its disposition in this population.

Results

Population Demographics

Figure 1 shows the flow chart of screened, randomized, and analyzed infants between June 2006 and December 2007 at 10 participating centers. Consent was obtained for 79 infants, 76 infants were randomized, and 74 infants received study drug. Two infants did not complete the minimum of four specified blood samples (three postdrug infusion), and their randomizations were replaced with two additional enrollees from the same center and of the same gestational age (GA) stratum, per protocol. Available data from the two replaced infants were included in the PK and safety analyses. One infant received placebo instead of the assigned 120 mg/kg dose, and for the PK analysis, this infant’s serum and urine data were included in the placebo group. However, this subject’s data on adverse events and clinical outcomes were included as randomized (intention to treat). The baseline characteristics of the randomized infants were similar across all three groups. The median age at the time of study drug infusion was between 2.4 and 3.2 d ( Table 1 ).

Figure 1
figure 1

Consort diagram: flow of consented, randomized, and analyzed subjects. aTwo infants started feeding prior to randomization; one infant’s condition worsened before randomization resulting in loss of vascular access required for drug administration. bTwo infants identified as ineligible postrandomization (one had <600 g birth weight, one developed severe intraventricular hemorrhage postconsent). cOne infant received placebo in error, samples included as placebo in the pharmacokinetic (PK) analyses; however, clinical outcomes included as intention to treat.

PowerPoint slide

Full size image
Table 1 Baseline characteristics of randomized infants by treatment group
Full size table

Safety

During the infusions, heart rate, blood pressure, and respiration did not differ between placebo and inositol infants at either dose (data not shown). The incidence of at least one adverse event was lowest for the higher dose of inositol: 80% in the placebo, 84% in the 60 mg/kg, and 54% in the 120 mg/kg group (P = 0.05; Table 2 ). The frequencies of sepsis, IVH, need for supplemental oxygen, need for mechanical ventilation, and use of the specified medications did not differ significantly among the groups, nor did early specified clinical diagnoses during the first 4 d after infusion ( Table 2 ). Serious adverse events occurred in 29% of all subjects and were lowest in the 120 mg/kg group (17%; P = 0.11) (data not shown, tables available from the authors). During hospitalization, there were no significant differences between groups in the rates of expected preterm diagnoses as a whole ( Table 3 ), nor between groups in the upper or lower GA strata. Clinical event rates were similar to historic data observed in this population in the National Institutes of Child Health and Human Development Neonatal Research Network (14).

Table 2 Clinical outcomes through day 7 after infusion by treatment group
Full size table
Table 3 Clinical outcomes and adverse events by treatment group
Full size table

PK Analyses

The raw mean serum levels rose in proportion to the dose given, gradually returned to baseline ( Figure 2 ), and appeared consistent with the compartmental model under consideration.

Figure 2
figure 2

Serum inositol levels by dose group. Samples were collected within scheduled windows, plus additional scavenged samples as available. For this graph, collection times were clustered as follows to obtain mean values: 0 h (baseline) = obtained before infusion; 0.3 h (end of infusion) = 0–2-h postinfusion; 4 h = 2–6-h postinfusion; 8 h = 6–10-h postinfusion; 12 h = 10–14-h postinfusion; 24 h= 14–30-h postinfusion; 36 h = 30–42-h postinfusion; 48 h = 42–60-h postinfusion; 72 h = 60–82-h postinfusion; 96 h = >82-h postinfusion. Light gray squares, placebo; gray circles, 60 mg/kg; black triangles, 120 mg/kg; vertical bar, ±1 SD.

PowerPoint slide

Full size image

The PK analysis, which included endogenous production, revealed no improvement between one- and two-compartment models (P = 0.38), so a one-compartment model was used. In the measurement of residual error, a constant variance best fit the data. The relationships between the random effects were graphically studied by plotting uvi vs. uCli, uVi vs. uRi, and uCli vs. uRi for all infants (see Methods section for formulas and definitions). A strong linear relationship was observed between the random-effect estimates for clearance (Cl) and endogenous production (R) with no apparent relationship between the other two combinations of random effects. The random effects were then modeled only with correlation between Cl and R.

Table 4 presents the population-PK (Pop-PK) estimates for the one-compartment model excluding any covariates with the associated random-effect variance and correlation estimates in Table 5 . Derived values for the elimination rate, the half-life, and the apparent concentration associated with endogenous production are also shown in Table 4 . The model appears to provide a good fit to the data as shown in Figure 3 with the observed and individual predicted values being well aligned. The mass of data points at the low values are primarily comprised of predosing and placebo measurements combined with a smaller number of late time point measurements, all of which are expected to have low inositol concentrations. Residual plots comparing predicted and actual values, not included here, did not indicate any major model deficiencies.

Table 4 Population pharmacokinetic parameter estimates for a typical infant (fixed effects)
Full size table
Table 5 Population pharmacokinetic random effects variances and correlations
Full size table
Figure 3
figure 3

Observed vs. individual predicted inositol concentrations from the population pharmacokinetic analysis. Predicted values were calculated for each observed data point using the individual characteristics in the model described in the PK section of the Results section.

PowerPoint slide

Full size image

This model was used to assist the investigators to project the effect on serum levels of inositol from repeated dosing with every 12-h or every 24-h schedules using several different daily doses. As an example, Figure 4 shows the model’s prediction for 80 mg/kg/d divided into two doses every 12 h, for 36 h. Although this study did not directly evaluate multiple dose administrations, the model’s estimate represents probable serum levels over a short time period for planning future studies of repeated administration.

Figure 4
figure 4

Predicted serum inositol model. The model described in the PK results section (before covariates) was used to predict the pattern of serum levels for a typical infant given repeated doses of 80 mg/kg/d divided into 40 mg/kg every 12 h for 36 h.

PowerPoint slide

Full size image

The effect of the following covariates on each of the model parameters was tested: infant birth weight, allometric size at birth, GA strata at birth (23–26 vs. 27–29 weeks), postmenstrual age, postnatal age, creatinine clearance, and sex (15). Allometric size, a function of birth weight, entered the model for a parameter as a multiplicative scaling factor given by , where wi is the birth weight, in grams, for the ith infant, is the median birth weight of the infants in the study (997.5 g), and βsize is the estimated coefficient. The remaining covariates had an exponential multiplicative effect given bywhere z is one of the covariates. Table 6 presents the mean and SDs for the continuous covariates while the two categorical covariates were both equally divided with 50% in each of their two categories.

Table 6 Covariate means and SDs
Full size table

Each covariate was tested individually against the base model. Those covariates found to have a statistically significant effect on any of the three Pop-PK model parameters, R, Cl, and V (V = volume of distribution), were then combined into a single model. Birth weight, postmenstrual age, postnatal age, and sex had no significant effects on the model parameters. Allometric size, GA strata, and creatinine clearance were found to have a statistically significant effect on one or more of the three model parameters ( Table 7 ). The final model including all of the statistically significant covariates indicates that infants with a GA of 27–29 wk had a lower volume of distribution than infants with a lower GA of 23–26 wk, while clearance increased with increasing allometric size. The pathway of influence for creatinine clearance was less clear. When its effect was considered separately on each of the model parameters, V, Cl, and R, it was statistically significant for Cl and R (P = 0.044 and P = 0.014, respectively) with a direction of effect that was physiologically plausible, increasing for Cl and decreasing for R. However, when creatinine clearance was allowed to simultaneously affect both Cl and R, then its direction of effect on Cl was the opposite of that expected with increased creatinine clearance leading to a reduction in Cl. In addition, the model with creatinine clearance affecting R alone was not significantly different from the model with creatinine clearance affecting both Cl and R (P = 0.29). Therefore, the final covariate model included the effect of creatinine clearance solely on R. The estimated typical value for each of the three model parameters were affected as follows by the final covariate model:

Table 7 PK model parameter estimates for a typical infant (fixed effects)
Full size table

where V*, Cl*, and R* are the covariate-adjusted model parameters and I (GA) is an indicator function that equals 0 (zero) for GA stratum 23–26 wk and 1 for stratum 27–29 wk.

Urine Excretion of Inositol

Excretion in the placebo infants had a baseline rate of 26.6 ± 12.3 mg/kg/24 h (mean ± SD) and showed an upward trend with age over the 96-h study to 36.8 ± 28.8 mg/kg/24 h on day 4 of the study. Following inositol administration, excretion rose during the first 12-h collection and decreased in the second 12 h. Thereafter, excretion approached levels in the placebo group. Following the 120 mg/kg dose, average excretion was 67.3 ± 35.8 mg/kg in the first 12 h and 25.3 ± 14.4 mg/kg in the second 12 h. For the 60 mg/kg subjects, 31.9 ± 13.4 mg/kg was excreted in the first 12 h and 18.0 ± 9.6 mg/kg in the second 12 h.

We examined the volume of urine output for evidence of a potential diuretic effect of the excreted inositol and found none. The mean volume of urine in the first 12 h after drug dosing when excretion was the highest was 40.7 ml/kg/12 h in placebo infants, 37.8 ml/kg/12 h in the 60 mg/kg group, and 46.2 ml/kg/12 h in the 120 mg/kg group, respectively (P value = 0.1242; Kruskal–Wallis test).

Discussion

These findings demonstrate that among preterm infants of 23–29 weeks gestation, a single i.v. dose of 60 mg/kg or 120 mg/kg inositol given over 20 min results in an initial dose proportional increase in serum concentrations that persists between 24 and 36 h. Accommodating a parameter for endogenous production, the PK analysis showed a central volume of distribution of 0.5115 l/kg with a half-life of 5.22 h when excluding any covariate effects. Although increased urine inositol loss occurred initially, this did not result in a diuresis.

Previous controlled trials suggest that supplemental inositol is safe and beneficial for preterm infants. The serum level achieved in the key study demonstrating a good benefit was 126–153 mg/l during the first week of treatment (11). In the present study, inositol also appeared safe; however, this study’s power to detect differences in uncommon adverse events is limited by the small sample size, and infants received only a single dose. Although the higher rates of bronchopulmonary dysplasia and sepsis in the highest dose group were unexpected, have not been previously reported, and are not statistically significant, it will be important to monitor these event rates during future studies. To date, reported studies do not describe i.v. inositol supplementation for >8 d.

Hallman et al. (11) observed an apparent steady-state serum concentration after the first of 5 d of 80 mg/kg/d dosing. The fall in serum levels in the “wash out” period in the study by Hallman et al. (11) suggested a half-life of 4–5 d, which differs substantially from our estimate (5.22 h) and is not explained, although fully established enteral intake of human milk in their studies may have contributed to support the serum levels despite discontinuing i.v. dosing. A PK study conducted with repeated dosing and a wider range of doses should help resolve these findings.

Brown et al. (16) found that the plasma rate of appearance of inositol in term and late-preterm infants was 121.7 mg/kg/d, well in excess of the usual daily enteral intake from preterm formulas or human milk (~54 mg/kg/d). Endogenous production from glucose occurs in many tissues, and enterally received inositol is actively transported across the intestinal mucosa where it is absorbed almost completely (17,18). The catabolic enzyme for inositol, myo-inositol oxidase, is localized almost exclusively to the renal cortex (19), but the activity of this enzyme is very low in the fetal and newly born renal cortex. It rises rapidly after birth as the renal blood flow redistributes from the renal medulla to the cortex. Therefore, urinary losses of intact inositol are high in both the term and preterm newborn but gradually fall to very low levels in the weeks after birth (4,20). Despite these developmental physiologic mechanisms that affect inositol concentrations, serum levels remain responsive to dietary intake decreasing in both tissue and serum when diets are low in inositol and rising with supplemented diets (4).

We speculate that immature renal tubular transport at these stages of development is inadequate to reabsorb the amount of inositol filtered by the glomerulus, similar to glucose in extremely-low-birth-weight newborns. The amount of filtered inositol reaching the renal tubules over a short period of time exceeded renal tubular reabsorption so the higher dose caused a greater renal loss. This pattern of renal excretion is expected to change rapidly postnatally and could lead to reabsorption of a greater percentage of the dose and higher circulating concentrations rather than to faster clearance and shorter half-lives which occurs with most drugs administered to preterm newborns during maturation. With the expected increase in the catabolic myo-inositol oxidase enzyme in the renal cortex in the weeks after birth, remaining filtered inositol will likely be converted to D-glucuronic acid, and no longer appear in the urine (19). These observations and remaining uncertainties are important to the design of future multidose efficacy trials. We plan to examine daily administration of inositol over longer periods of time. Different doses will be evaluated, and divided doses will be used in an attempt to reduce peak serum levels and therefore reduce urine losses. As we increase our understanding of how to influence serum levels, we can better test inositol’s safety and potential to benefit extremely preterm infants.

Methods

Design

A randomized, double-masked, placebo-controlled PK trial with sparse blood sampling was conducted by the National Institutes of Child Health and Human Development Neonatal Research Network. Ten of the Neonatal Research Network Centers participated and enrolled subjects between June 2006 and December 2007. Randomization was performed centrally via computer within two prespecified GA strata (230/7–266/7 wk, and 270/7–296/7 wk), and infants were monitored prospectively for toxicity and clinical outcomes during hospitalization.

Population

Eligible subjects were of 230/7–296/7 wk gestation and ≥600 g birth weight, had no major congenital anomalies, were between 12 h and 6 d of age at randomization, and had received no human milk or formula feedings since birth.

Intervention

Inositol was given as a single low (60 mg/kg) or high (120 mg/kg) dose of 5% myo-inositol (referred to as “inositol” in the remaining text) i.v. over 20 min in a 1:1:1 randomization with placebo delivered in one of two volumes to maintain masking (5% glucose, United States Pharmacopeia, for i.v. administration). Drug or placebo was dispensed from the respective pharmacies in unit doses labeled as “inositol study drug,” and all clinical and research personnel except the pharmacist were masked to the study group. Inositol (hexahydroxycyclohexane) is manufactured from corn or rice bran and was provided as an isotonic, pyrogen- and preservative-free, sterile 5% solution of myo-inositol in water containing 0.5 g sodium chloride per liter (8.55 mmol/l), with pH ranging between 6.5 and 7.5.

Outcome Variables

Infants were continuously observed during the infusion by study personnel, and vital signs were recorded every 5 min for the first 30 min and thereafter at 15–30-min intervals for 2 h. Six prespecified clinical conditions and seven concomitant medications were recorded for 4 d after infusion (clincical conditions: culture-positive sepsis, patent ductus arteriosus, clinically diagnosed IVH, imaged IVH, supplemental oxygen, and mechanical ventilation; and medications: surfactant, dopamine/dobutamine, antibiotics, indomethacin as prophylaxis for IVH, indomethacin as a treatment for PDA, systemic steroids, and diuretics).

Adverse Events

Adverse events were recorded for 7 d, and severity determined using a toxicity table modified for neonatal use from the National Cancer Institute. Baseline characteristics and demographic data as well as neonatal morbidities from birth through hospital discharge (or 120 d if sooner) were recorded from the medical record using the definitions of the Neonatal Research Network generic database. Bronchopulmonary dysplasia was defined as receiving oxygen at 36 wk postmenstrual age (or at discharge if discharged before 36 wk postmenstrual age); IVH was classified according to Papile et al’s study (21); late-onset sepsis was defined as a positive culture from a normally sterile body fluid obtained after 72 h of age; and necrotizing enterocolitis was defined as either Modified Bell’s IIA or worse not needing surgery, or requiring surgery (IIIB) (22).

Blood Sampling

Each subject contributed four blood samples (one predose plus three postdose samples) according to a sparse sampling, Pop PK design (23,24). The predose sample was within 4 h before infusion, and the three postdose samples were distributed among nine sample windows: within 4 min of the end of infusion, 4 ± 1 h, 8 ± 1 h, 12 ± 1 h, 24 ± 2 h, 36 ± 2 h, 48 ± 2 h, 72 ± 2 h, and 96 ± 2 h after the start of the infusion. In addition, some scavenged serum samples from the clinical laboratories were obtained, with consent, to supplement the scheduled samples. The exact time of sampling in relation to the end of drug infusion was recorded for all samples. Blood (200 µl) was collected in serum separator microtubes, refrigerated until centrifugation, up to 60 h after sampling, and frozen at −70 °C to −80 °C until analyzed. (Stability data at room temperature for up to a week are available from the authors upon request.)

Urine Collections

Six consecutive urine collection procedures were carried out starting from 8 to 12 h before the infusion of study drug and continued for 96 h postinfusion (the second two for 12 h each, then three for 24 h each). Urine volume was determined by the change in weight of each preweighed non–gel-containing diaper with cotton balls within the given 12- or 24-h time period, assuming 1 g = 1 ml. Urine was extracted from the diapers and cotton balls and pooled within each timed collection, and inositol and creatinine were determined on an aliquot of the pooled sample. Serum creatinine was measured at least twice during the 96 h by each institution’s clinical laboratories.

Enteral Inositol Intake

Following drug infusion, enteral intake was permitted when the infant was considered as “ready to begin feedings” by the clinical team. Enteral intake (type and volume each day) was recorded for 4 d beginning with the day of study drug infusion. Samples of ingested milk were assayed for inositol to permit estimation of additional inositol intake during the sampling period.

Assay

Inositol isomers (myo-inositol, 1,5-anhydro-D-sorbitol, and D-chiro-inositol) were quantified from 25 or 50 µl samples using high-performance liquid chromatography with electrochemical (pulsed amperometry) detection (method and validation documents available upon request from http://www.ttuhsc.edu/sop/research/internalgrants/CPET.aspx., Leff RD). Results of validation experiments showed the following assay performance parameters: lower limit of quantitation 1.0 mg/l; coefficient of variation from low to high concentrations was 10.2–13.4% between days, and 1.9–2.3% within a day. Validated test samples included serum; plasma collected using EDTA, sodium heparin, or lithium heparin; urine extracted from any of five common nongel preterm infant diaper types with or without cotton balls; human milk and infant formulas. Serum, human milk and formulas contained only nondetectable or only barely detectable levels of the D-sorbitol and D-chiro-inositol isomers. Therefore, only myo-inositol levels are reported. The concentration of inositol in formulas was 356 (67) mg/l (m (±SD)) and in human milk was 287 (107) mg/l.

Statistical Analyses

The planned sample size was at least 36 infants in each GA stratum who received study drug and completed four blood sampling procedures. Baseline characteristics and clinical outcomes for all randomized infants were analyzed by study group using χ2 tests, Fisher’s exact tests, ANOVA, or Kruskal–Wallis tests, where appropriate.

PK Analysis

Pop-PK models were fit to the data using the nonlinear mixed-effects approach in Monolix (Monolix version 3.2. LIXOFT http://www.lixoft.com/, Paris, France). This approach accounts for the variability between infants in the model parameters, the correlation between measurements on the same infant at different occasions, and the residual unexplained variability in serum concentrations (25).

As noted, inositol is produced endogenously and is present in human milk and infant formulas (26). Hence, the Pop-PK model had to account for an endogenous serum concentration of inositol not related to treatment. Apparent endogenous inositol production was included in the Pop-PK model similarly to Hayashi (27). The final model combines apparent endogenous production with a one-compartment i.v. infusion model with linear elimination. The model for serum concentration is then

where for the i-th infant, Ci (t) is the serum concentration at time t, Ri the apparent rate of inositol infusion due to the combination of endogenous production and feeding, Cli the clearance, and Vi the apparent volume of distribution. T is the duration of the infusion period and t is the time after the start of the infusion, both in hours; finally, is the residual error at time t. The steady-state endogenous concentration is then given by for the ith infant, which is used to measure the combined effect of endogenous inositol production and inositol intake from enteral feeding. It was not possible to separate these two sources of inositol since enteral intake was measured as a total amount fed over a day and not the amount fed at each occasion (feedings are given eight times throughout the 24-h day). In addition, the enteral intake of inositol was very low during the 4-d study. There was a minimal enteral intake of inositol in the first 12 h after the i.v. dose, and a maximum of 3.6 mg/kg in the first 24 h. Only 49% of subjects received any enteral intake in the second 24-h period (mean intake of inositol: 1.8 mg/kg/d; range: 0–20), and the mean intakes on days 3 and 4 were 4.5 and 5.5 mg/kg, respectively. Because the calculated enteral intake of inositol from human milk or formula among those able to receive some feedings was so low during the first two study days; measured enteral intake was not included as a separate source in these models.

The between-infant variability in the Pop-PK model parameters, Ri, Cli, and Vi is modeled using random-effect variables (uR, uCl, and uV) that approximate the individual trajectory over time of each infant’s concentration. The random effects are assumed to be normally distributed with means of 0 (zero) and variances and correlations that will be estimated. For example, the clearance for the i-th infant is modeled as , where Cl is the fixed-effect common to all infants and is the random effect unique to the i-th infant. A similar formulation is used for Ri and Vi. Thus, the three model parameters are log normal. Individual specific parameter estimates were obtained as the conditional modes, or the maximum a posteriori, of the Bayes estimates of the parameters. The fixed effects, R, Cl, and V, are the median values of the parameters and are called the typical values for the population from which each infant’s parameters are derived. The residual error, , is assumed to be uncorrelated with the random effects and normally distributed with mean 0 (zero) and a variance that will be estimated.

The quality of fit of a Pop-PK model was judged by visual examination of plots of observed vs. individual predicted concentrations and of residuals vs. individual predicted concentrations. Nested models were compared by referencing the improvement in the objective function (−2 log likelihood) against the χ2 distribution with degrees of freedom equal to the difference in the number of estimated parameters between two models. Statistical significance was assessed at the level of P ≤ 0.05.

Ethical Oversight

The institutional review boards of each center approved the protocol, and written informed consent was obtained for each participant. An independent Data Safety Monitoring Committee approved the protocol and monitoring plan before the study began and monitored the accumulating safety data at regular intervals. The US Food and Drug Administration approved protocol (IND: 70510) was registered with ClinicalTrials.gov (NCT00349726).

Statement of Financial Support

The National Institutes of Health provided grant support through the Eunice Kennedy Shriver National Institute of Child Health and Human Development’s (NICHD) Neonatal Research Network and the Pediatric Pharmacology Research Units Network, with cofunding from the National Eye Institute and with support from the National Center for Research Resources, and the National Center for Advancing Translational Sciences.

The content of the publication is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Additional participants and grant numbers are cited in the Supplementary Material online.

Disclosure:

None of the authors report any commercial, proprietary, or financial interest in any of the products described in this article. Abbott Nutrition Division, Abbott Laboratories, supplied the inositol drug used in the study. NICHD is the sponsor of the study and holds the investigational new drug (IND) number. Portions of this study were presented at the 2010 Pediatric Academic Societies Annual Meeting, Vancouver, Canada, May 1–4, 2010.