QTc Intervals in Former Preterm/Extreme Low Birth Weight Infants: a Pooled Study Proposal

Preterm birth and gestational age at birth are associated with an increased risk for cardiovascular morbidity and mortality. In a large population-based cohort study with an incidence of preterm birth of 5.4 %, the adjusted hazard ratio of all-cause mortality in former preterms (23-33 weeks) was 1.44 in young adulthood (<50 years), most pronounced for cardiovascular mortality (1.89), diabetes (1.98) and chronic lung disease (2.28). Interestingly, the increased death risk was found across gestational age up to the ideal term age (39-41 weeks), but most prominent for the most immature group.

Besides better understanding of the underlying mechanisms associated with the 'Barker' and 'Brenner' cardiovascular risk hypotheses, there is an obvious need to further explore biomarkers to detect potential higher risk besides the gestational age or birth weight to translate this knowledge to clinical practice or secondary prevention strategies.

In a recent meta-analysis on changes in the preterm heart from birth to young adulthood, it was concluded that former preterms have morphological and functional cardiac impairments, proportionally to the degree of prematurity. In contrast to the echocardiographic differences in left and right ventricular systolic functions in former preterm when compared to term controls, observations on electrocardiographic (ECG) findings are less conclusive. Whether preterm birth is associated with conduction or repolarization abnormalities later in life is less well explored.

At present and based on a recently conducted systematic search, there are conflicting data on the potential QT interval prolongation (all Bazett) in former extreme low birth weight (ELBW, <1000 g) infants (data reported chronological).

Bassareo et al. reported that corrected QT (QTc) intervals and QT dispersion (QTd) were significantly prolonged in 24 former ELBW cases compared to 24 term controls (mean, ± SD: 417, 23.6 versus 369, 19.5 ms, and 30.4, 14.1 versus 24.6, 8.2 ms, respectively) (age at assessment = 23.2, SD 3.3 years).

Gervais et al. could subsequently not confirm these findings in 49 preterm cases (<30 weeks) when compared to 53 term born controls. At rest, mean QTc were 408, ± 34 and 409, ±SD 31 ms respectively (age at assessment = 18-33 years). Post-hoc power calculations showed that this study had 80% power to detect a difference of at least 18 ms in QTc between the two groups.

Finally, based on the PREMATCH study (ClinicalTrials.gov NCT02147457, S 56577), data on QTc observations are also available. QTc and QTd were similar between 93 ELBW cases and 87 controls [409 (range 360-465) versus 409 (337-460); 40 (0-100) versus 39 (0-110)] ms. When expressed by mean and SD, 409 (23) ms in the full dataset, 409 (SD 22) in former ELBW cases, 409 (SD 24) in controls [Salaets et al, Pediatr Res 2022]. Age, height, weight or body mass index were not associated with the QTc interval, while female sex (median difference 11.4 ms) and lower potassium (r=-0.26) were associated with longer QTc interval. We could not observe any significant association between QTc interval and perinatal characteristics (age at assessment 8-14 years). Post hoc analysis showed that our study had a power of 90 % to detect a difference of 11,45 ms between both groups.

Consequently, if researchers truly want to assess the presence or absence of either a difference or a prolongation of QTc intervals in this specific population, pooling of published data is likely the most effective approach (potential number of cases = 24 + 49 + 93 = 166; potential number of controls in the same studies = 24 + 53 + 87 = 164), preferably based on individual data. Although the sample is to a large extent pragmatic (as available), the investigators hereby aim to target the 5 ms QTc prolongation applied by the authorities (FDA, EMA) in paired healthy adult volunteer studies as 'golden' standard as primary outcome variable [EMA guideline, FDA guidance]. In all 3 cohorts, perinatal characteristics (like gestational age, weight, or pre- or postnatal steroids) were explored on their potential impact on the QTc interval, but all cohorts were underpowered to draw firm conclusions, so that these analyses will be included as secondary analyses.

Descriptive statistics will be used to report on the pooled dataset, and will be reported by mean and standard deviation, or median and range, pending normal distribution characteristics or incidence (%), (cases and controls, or cases only, respectively).

Statistical analysis will be used (two-tailed or one-tailed) to explore the relation of the data to the underlying population(s), like Rank correlation, Mann Whitney U or t-test (cases to controls, and within cases respectively). Based on the individual cohorts published, it is reasonable to anticipate normal distribution.

Power exploration using Mann-Whitney-U tests at alfa = 0,05 In order to find a significant two-tailed difference of 5ms, with a power of 90 % and extrapolating from the data distribution of the Leuven cohort, it was estimated that 467 subjects are required in both case and control groups. Decreasing the power to 80 % still requires 349 subjects in each group. In order to find a significant one-tailed difference of 5ms, with a power of 90 % and using the data distribution of the Leuven cohort, it was estimated that 381 subjects are required in both case and control groups.

Decreasing the power to 80 % still requires 275 subjects in each group (when using the FDA and EMA criteria to exclude an increase of 5 ms in former preterm cases).

In order to find a significant two-tailed difference of 10ms, with a power of 90 % and extrapolating from the data distribution of the Leuven cohort, it was estimate 118 subjects are required in both case and control groups. Decreasing the power to 80 % requires 88 subjects in each group.

In order to find a significant one-tailed difference of 10ms, with a power of 90 % and using the data distribution of the Leuven cohort, it was estimated that 96 subjects are required in both case and control groups. Decreasing the power to 80 % requires 70 subjects in each group. Inversely, having 166 and 164 subjects in each group and extrapolating the Leuven distribution data, this will have 90 % power to significantly discriminate a difference of 8.43ms (two-tailed) or 7.60ms (one-tailed), or 80 % power to significantly discriminate a difference of 7.28ms (two-tailed) or 6.46ms (one-tailed). Due to differences in data distribution between the cohorts: the actual discriminative power will probably be lower (the Leuven cohort -used for this estimation- has the narrowest standard deviation, and the Italian cohort has a different mean).

The investigators intend to conduct a post hoc power analysis with a difference (hypothesis: the mean QTc is not >5 ms higher in former preterm neonates compared to controls).