We thank Drs. Levy and Singh for their interest in our study “Preterm Birth Is Associated with Altered Myocardial Function in Infancy,” recently published in JASE . Their overall message, that small observational studies may suffer from limitations, is well taken.
In their letter, Levy and Singh emphasize that the principal finding of our study, a difference in myocardial longitudinal strain between 6-months-old infants born very preterm and same-aged peers born at term, could have been confounded rather than reflective of “physiological maturational processes” after preterm birth. In observational studies such as ours, confounding cannot be excluded. For example, the association between very preterm birth and lower myocardial deformation at 6 months of postnatal age may have been confounded by a maternal condition or exposure that contributed to (1) an increased risk for very preterm birth and (2) altered cardiac development in the infant. Neonatal influences, introduced after birth, are less likely to represent true confounding, because a confounder should be related to both exposure (preterm birth) and outcome (myocardial function in infancy). Therefore, a neonatal influence may have (1) acted as mediator of the effect, occurring in the causal pathway from preterm birth to altered myocardial function in infancy; (2) represented an interacting covariate (on the association between preterm birth and outcome); or (3) had no relevance to outcome.
Our longitudinal study was planned as a test of the hypothesis that myocardial function in human infants born very preterm develops in a significantly different manner than in infants born at term (as suggested by animal data ). Given a demonstrated dose-response relationship between preterm birth and adverse outcome, with larger risks for hypertension, cardiac hypertrophy, and death of cardiovascular disease in adults born at gradually lower gestational age, we chose to study extremely to very preterm infants born 10 to 14 weeks before term. At such short gestations, maternal complications precede preterm delivery, and neonatal intensive care for their infants is mandatory. Some of these exposures might have confounded, mediated, or interacted with our findings. Therefore we listed important antenatal and neonatal exposures and morbidities and clearly discussed this limitation in the paper.
At the end point of our study, when the infants had passed the neonatal period, we choose, in addition to preterm birth, also to evaluate (in a multivariate regression model) any role of two potential confounders (maternal smoking during pregnancy and birth weight SD score as a proxy for fetal growth) and one potential covariate (body weight at follow-up) for myocardial outcomes. The choice of these independent variables was based on (1) significant group differences in univariate tests (Table 1 in our paper), (2) the literature, suggesting a role of smoking and smallness for date in lasting cardiovascular developmental programming, and (3) sample-size considerations. The perinatal influences, suggested to have significant effects on right ventricular and left ventricular (LV) deformation and listed by Levy and Singh, might have had a larger impact when the infants were younger and still underwent care in the neonatal unit. Being aware of limitations in power, we found no association between antenatal steroid exposure, gestational age, a neonatal diagnosis of patent ductus arteriosus or bronchopulmonary dysplasia, respectively, and myocardial function at 6 months of postnatal age. And some of the potential confounders or covariates highlighted by Levy and Singh have little if any relevance to our study: there were no mothers with diabetes mellitus or asthma, and only one mother had preeclampsia, which is unlikely to have had any significant influence on results or interpretations. In addition, there were no infants with pulmonary hypertension in our study.
As for the solution suggested by Levy and Singh, to study only cohorts of uncomplicated preterm infants instead of “lumping them (diseased and healthy) all together,” we argue that there is no such cohort. The “healthy preterm infant,” suitable for research on physiologic maturational patterns and completely disengaged from perinatal conditions and exposures, in our view does not exist. Also, mothers delivering moderately or even late preterm have complicated pregnancies and deliveries, and their babies experience neonatal complications.
The end point in our longitudinal study of myocardial function in preterm infants was at 6 months of postnatal age. At this point, all morbidities before and after birth had resolved, all preterm infants had been discharged from the hospital before term-equivalent age, all were considered clinically healthy without any need for medication, and they exhibited normal heart rate and blood pressures. This fits well with the suggestion by Levy and Singh that studies on longer term developmental processes of the preterm heart should, if and when possible, be based on as large cardiorespiratory healthiness as possible.
The aim of our study was not to provide reference values for physiologic myocardial performance at different gestational ages but rather to describe the development of myocardial deformation over time in a sample of very preterm infants without severe neonatal morbidity (defined as major brain injury, severe retinopathy of prematurity, severe bronchopulmonary dysplasia, or necrotizing enterocolitis) and without remaining disability. The role model Levy and Singh picked for a study assessing the evolution of physiologic postnatal hemodynamics and allegedly fulfilling their standards for preterm healthiness shares some of the limitations of our study and, in addition, introduces other limitations excluded by us. For example, in the cited study by Koestenberger et al. , maternal diabetes and preeclampsia were not exclusion criteria; extremely preterm infants with ongoing (at the time of cardiac assessment) respiratory disease were included; all infants at <29 weeks of gestation were on treatment with indomethacin (a powerful vasoconstrictor, not only of ductus arteriosus but also of systemic arteries and arterioles) while cardiac reference values were determined; the cardiac evaluation was performed only once and within 48 hours after birth at a time point when circulatory adaptation from fetal life (such as increased pulmonary vascular resistance) may still be ongoing; and examiners were not blinded to gestational age, which introduces a risk for bias. These issues clearly illustrate the complexity of designing clinical studies of preterm infants. And can this be interpreted as healthy preterm physiology completely separated from disease status?
Levy and Singh comment that our calculations of sample size and power “might be appropriate only … with one outcome variable (i.e., a cohort of uncomplicated preterm-born infants).” Reading this, we unfortunately suspect some misunderstanding. As outlined in the statistical methods in our paper (our apologies for any lack of clarity), preterm birth was defined as main exposure , and the predefined outcomes were speckle-tracking variables, such as longitudinal strain, strain rate, and myocardial velocity at the end point of our study. All outcomes were continuous variables, and with 25 infants in each group, we would be able to detect a group difference in any of these outcome variables of ≥0.8 SDs.
We approximated global ventricular function from free wall strain measurements but did not suggest that our approximation should substitute for other approaches. As Levy et al. pointed out in another publication, there is at present no consensus on which approach on global strain estimations is more accurate or correlates more efficiently with health and disease outcomes than the other. The suggestion that LV global longitudinal strain always should be measured by averaging all values from 17 subsegments of the heart may not be as critical in very preterm infants (with very small hearts) as in adults. Although the vendors provide the software for this algorithm, the method was originally developed for studies of regional myocardial changes due to ischemic heart disease in adults. In neonates, regional functional impairment is very rare. This might be the reason why also other investigators (and clinicians) choose the same apical four-chamber view as we did for LV longitudinal strain measurements. And as for pediatric point estimates of LV global longitudinal strain, we note that calculations based on the combined measurements from four-, three-, and two-chamber views (mean global longitudinal strain, −20.2%; 95% CI, −19.5% to −20.8%) were found to be similar to those based on an apical four-chamber view only (mean free wall longitudinal strain, −20.4%; 95% CI, −19.8% to −21.7%).
So what do our findings mean? We thank Levy and Singh for pointing out the relative stability in our point estimates of LV strain in preterm infants despite individual variations, heterogeneity of perinatal risk factors, sample size, imaging, and power calculation issues; we find this reassuring in terms of study design and methodology. But whereas Levy and Singh chose to comment only on the “stable maturational pattern” of LV strain in the preterm infants investigated by us, we would also put it into perspective of the pattern in healthy, term infants. There was a significant relative increase in LV strain by 11% from birth to 3 months of age in term infants (absolute change from term to 3 months of age [ΔLV strain], +2.5%; P < .001), but not in preterm infants, who exhibit a slower, nonsignificant increase during the first (preterm ΔLV strain, +0.8%; P = .14) and second (preterm ΔLV strain, +1.3%; P = .05) 3 months periods of postnatal life. These findings indicate deviating postnatal trajectories of myocardial deformation for preterm and term infants (a similar group difference emerged also for LV early diastolic myocardial velocity). The effect amounted 0.6- to 0.7-SD lower mean LV longitudinal strain in 6-month-old preterm infants than in those infants who were born at term. We think that the clinical relevance of these findings remains to be explored and understood, much in the same way as research on blood pressure in people born small or too early has been performed and understood. What seemingly looks like small differences early in life may reflect greater disease susceptibility later in adult life.
Finally, we investigated in total 55 infants (25 very preterm and 30 term) and performed repeated echocardiographic examinations (in total 130 examinations) at prespecified postnatal ages. On the basis of this sample size and study design, we find no reasons to disqualify comparisons of speckle-tracking echocardiographic outcomes in infants born preterm and in infants born at term. Our sample size is in the same range as several other observational studies in this area of research. In the recently published meta-analyses by Levy et al. (which the authors should be commended for) on LV and right ventricular myocardial strain in children at different ages (cross-sectional data), 24 of 43 studies (56%) judged as valid had ≤40 participants. Further and larger studies are warranted on different aspects of the developing cardiovascular system after preterm birth. We are very much looking forward to take part of the results of the larger and so far unpublished study that Levy and Singh mentioned, in which myocardial performance has been tracked from birth to 1 year in extremely preterm infants and in which “all potential confounding factors have been adjusted for.”
This work was supported by grants from The Swedish Heart-Lung Foundation and the Childhood Foundation of the Swedish Order of Freemasons (grant number 20090380 ) and by a regional agreement on medical training and clinical research (ALF) between the Stockholm County Council and the Karolinska Institutet.