The increase in diaphragm thickness in preterm infants is related to birth weight: a pilot study

Diaphragm ultrasound (DU) has been used in adult and pediatric critical patients in relation to prediction of extubation success or to detect diaphragm dysfunction, but there is a lack of evidence in neonates. Our aim is to study the evolution of diaphragm thickness in preterm infants, as well as related variables. This prospective monocentric observational study included preterm infants born before 32 weeks (PT32). We performed DU to measure right and left inspiratory and expiratory thickness (RIT, LIT, RET, and LET) and calculated the diaphragm-thickening fraction (DTF) in the first 24 h of life and then weekly until 36 weeks postmenstrual age, death, or discharge. Using multilevel mixed-effect regression, we evaluated the influence of time since birth on diaphragm measurements, as well as bronchopulmonary dysplasia (BPD), birth weight (BW), and days of invasive mechanical ventilation (IMV). We included 107 infants, and we performed 519 DUs. All diaphragm thickness increased with time since birth, but the only additional variable that influenced this growth was BW: beta coefficients RIT = 0.00006; RET = 0.00005; LIT = 0.00005; and LET = 0.00004, p < 0.001. Right DTF values remained stable since birth but left DTF increased with time only in infants with BPD. Conclusion: In our population we found that the higher the BW, the higher diaphragm thicknesses at birth and follow-up. Contrary to the previously published findings in adult and pediatric settings, we were unable to describe a relationship between days of IMV and diaphragm thickness in PT32. The final diagnosis of BPD does not influence this increase either, but it does increase left DTF. What is Known: • Diaphragm thickness and diaphragm thickening fraction have been related to the time on invasive mechanical ventilation in adults and pediatric patients, as well as with extubation failure. • Very few evidence is yet available on the use of diaphragmatic ultrasound in preterm infants. What is New: • Birth weight is the only variable related to diaphragm thickness in preterm infants born before 32 weeks postmenstrual age. • Days of invasive mechanical ventilation do not influence diaphragm increase in thickness in preterm infants. What is Known: • Diaphragm thickness and diaphragm thickening fraction have been related to the time on invasive mechanical ventilation in adults and pediatric patients, as well as with extubation failure. • Very few evidence is yet available on the use of diaphragmatic ultrasound in preterm infants. What is New: • Birth weight is the only variable related to diaphragm thickness in preterm infants born before 32 weeks postmenstrual age. • Days of invasive mechanical ventilation do not influence diaphragm increase in thickness in preterm infants.


Introduction
Recent research on diaphragm monitoring has focused on the importance of this muscle in many respiratory disorders: diaphragm thickness and diaphragm-thickening fraction (DTF) have been related to the time on invasive mechanical ventilation (IMV) in adults [1,2] and pediatric patients [3], as well as with extubation failure [3][4][5]. Being able to reproduce this finding in premature infants may help to reduce extubation failure in this population [6]. The development of these lines of investigation has been possible due to the application of point-of-care ultrasound (POCUS) in intensive care units. POCUS is being increasingly spreading worldwide because it is easy to learn and reproducible, and such imaging can be performed at the bedside and repeated as many times as necessary without relevant side effects [7,8]. After many studies in this field, clinical guidelines based on the main evidence regarding this issue in pediatric and neonatal patients have been published [9]. Nowadays, the few studies performed on diaphragm ultrasound (DU) in neonatal patients show reference values for the diaphragm thickness [10,11] or thickening fraction [12], changes according to the level of noninvasive respiratory support [13], the relationship with extubation failure after initial respiratory distress syndrome [14], the decrease of diaphragm inspiratory thickness after intubation [15], or comparisons between preterm infants exhibiting bronchopulmonary dysplasia (BPD) at term corrected age and term infants without morbidities [16]. However, all these studies have a transverse design, but premature infants' diaphragms should grow with them during their Neonatal Intensive Care Unit (NICU) stay. Static measurements of the diaphragm at specific times may not consider the different variability of our patients; some of whom are twice the size of others, or remain for several weeks with mechanical ventilation.
This difference has led us to consider prospectively assessing the evolution of the diaphragm's thickness in premature patients from birth until 36 weeks postmenstrual age, death, or discharge, using bedside ultrasound, as well as assessing which variables are related to this evolution.

Materials and methods
All patients' parents gave written informed consent to participate in the study. The study protocol was approved on January 2018 by the "Ethics and Investigation Committee in the province of Cádiz" (internal code number TFG-FASV-2018), with the name "Monitoring of diaphragmatic function measured by fractional shortening in very low birth weight newborns." The study was conducted following the STROBE statement guidelines [17]. The study was conducted following the ethical standards of the responsible regional committee on human experimentation and with the Helsinki Declaration of 1975.
We conducted an observational study with preterm infants born before 32 weeks (PT32) in our center from February 2018 to August 2021. The aim was to study diaphragm thickness progression from birth, weekly, until 36 weeks postmenstrual age, death, or discharge. We analyzed the right and left inspiratory and expiratory thickness (RIT, LIT, RET, LET) and left and right diaphragm-thickening fraction (DTF). We divided patients based on BW: Group 1 included infants born at less than 1000 g and Group 2 included infants born at PT32 with a BW of 1000 g or higher. Infants with chromosomal abnormalities or major malformations, those with clinical instability since birth that made DU difficult to perform, or infants who required sedation were excluded.
In our center, we do not provide routine sedation for preterm infants who need IMV. We only proceed if the patient is unable to adapt to the ventilator to avoid severe episodes of hypoxemia (approximately 1% of our annual PT32 receive sedation).
We gathered data on general clinical variables in both groups (GA, birth weight [BW], sex, prenatal steroids, type of delivery, SNAPPE-II [18], surfactant administration, length of admission, and death before discharge from the NICU), as well as other variables possibly related to diaphragm growth: late-onset sepsis (LOS) [19], necrotizing enterocolitis Bell's grade II-III (NEC) [20], days of IMV/noninvasive ventilation (NIV)/total respiratory The exams were recorded anonymously, and diaphragm thicknesses were measured in three different respiratory cycles. We used mean values for each patient. We also searched for diaphragm dysmotility patterns or paradoxical movements in the infants evaluated, as has been previously described in some children [23]. The measurements were performed offline after the end of the exam. The diaphragm thickness was measured to the nearest 0.1 mm, which is within the range of the accuracy of the transducer.

Statistical analysis
To compare variables between groups, we used the Wilcoxon rank sum test with Bonferroni correction for multiple comparisons and the chi square test or Fisher's exact test. We used a linear multilevel mixed-effect regression (LMLR) model to adjust for repeated measures to study the effect of time on the diaphragm thickness or thickening fraction.
To analyze which variables may also influence diaphragm growth, we first calculated bivariable LMLR models with the variables "time since birth," BPD, days of IMV, NIV or CPAP, BW, LOS, or NEC. We first eliminated from the model those variables with low tolerance (below 0.1). Beta coefficients with 95% confidence intervals (CIs) were calculated for each model. Only those variables with significant beta coefficients were included in a multivariable LMLR model. LMLR models were calculated for the whole sample and were not stratified according to BW. Different LMLR Fig. 1 Description of the technique to perform diaphragm ultrasound at the zone of apposition (between the anterior and the posterior axillary line and perpendicular to the ribs, between the eighth and the ninth intercostal spaces). A high-frequency (8-15 MHz) probe was used. A M-mode: note diaphragm thickening during inspiration (distance from line 3 and 4), which is the inspiratory thickness (IT), as well as the diaphragm shortening during expiration (distance from line 1 and 2), so-called expiratory thickness (ET). B B-mode: note the diaphragm as an hypoechogenic linear structure over the liver, delin-eated by two hyperechogenic lines: the parietal pleural line (above) and the parietal peritoneal line (below). IT and ET can also be measured in B-mode, freezing the image both during maximal expiration and maximal inspiration. All measurements are performed in three consecutive breaths, and the mean values are analyzed. We always selected one inspiration and the previous corresponding expiration. The diaphragm-thickening fraction (DTF) was calculated as follows: support, and BPD according to the definition by Jensen and coworkers [21]. Nasal-intermittent positive pressure ventilation (nIPPV) and NIV-NAVA were the modes included as NIV. Clinical sepsis was diagnosed using the Kiss method [22]: those cases that began before 72 h from birth were diagnosed as early-onset sepsis and otherwise as late-onset sepsis.

Ultrasound technique
All exams were performed with a portable ultrasound machine (General Electric Vivid IQ). The focus was positioned at 1 cm with the depth at 3 cm, and harmonics were turned on. All exams were performed by the same investigator (A. A. O.), always during sleep or with the patient breathing quietly, in the supine position and approximately 1 h after feeding. Parents or nurses collaborated with us to calm infants who were uncooperative during the examinations. During the procedure, all patients were monitored noninvasively using pulse oximetry for heart rate and arterial oxygen saturation. The ultrasound gel was heated beforehand to prevent neonatal hypothermia.
We performed the DU at the zone of apposition, between the anterior and the posterior axillary line and perpendicular to the ribs, between the eighth and the ninth intercostal spaces, with a high-frequency (8-15 MHz) linear probe to measure the RIT, RET, LIT, and LET, which we used to calculate the DTF, according to our previous publication [12], as it is shown in Fig. 1. models were compared using likelihood-ratio (LR) tests. All tests were considered statistically significant if p < 0.05.
We calculated the sample size for a mean estimation of diaphragm thickness with an α-error of 5%, a power of 90%, a precision of 0.3%, and an estimated standard deviation of 0.9 mm (according to a previous study in the NICU [12]), so we needed a total of 35 infants in each group. To maintain the usual proportion of infants in both groups, we kept a ratio of 1:2.

Results
We included 129 PT32, but 22 were excluded because we could only perform one DU during their stay in our NICU. Ultimately, we analyzed 107 PT32, and we performed 519 DUs. The flow chart of patient inclusion is detailed in Fig. 2. Group 1 comprised 37 infants born at less than 1000 g, and Group 2 was composed of 70 patients born at 1000 g or higher. The patients' main clinical characteristics are described in Table 1. There were significant differences between groups in terms of GA, BW, CRIB, and SNAPPE-II at birth, days of IMV, days of NIV, total days of nCPAP, incidence of BPD, length of admission, and death. The median and interquartile ranks of RIT, LIT, RET, LET, and right and left DTF in both groups from birth until the end of the study are depicted in Figs. 3 and 4.

Diaphragm thickness
Time since birth was significantly related to growth in terms of diaphragm thickness in these infants, according to LMLR, compared to the linear model: RIT (LR chi 2 32.21, p < 0.001), LIT (LR chi 2 19.62, p < 0.001), RET (LR chi 2 26.78, p < 0.001), and LET (LR chi 2 17.48, p < 0.001). In the bivariable analysis, we studied the variables that could influence growth in terms of diaphragm thickness, but only BW (in all measurements) and length of IMV (in right thicknesses), together with time since birth, were significantly related (see the beta coefficients and 95% CI in Table 2). As only one patient suffered NEC, we could not conclude anything about this variable's effect.
Using multivariable analysis, we found that only BW and time were related to diaphragm growth in this population (see Table 2). The addition of BW to time since birth to predict diaphragm thickness increased the validity of the final model in all measurements compared to the model in which time was the only predictive variable: RIT (LR chi 2 47.65, p < 0.001), LIT (LR chi 2 30.23, p < 0.001), RET (LR chi 2 46.68, p < 0.001), and LET (LR chi 2 36.11, p < 0.001). The predicted progression of RIT, LIT, RET, and LET using LMLR analysis according to BW is described in the supplementary material (S- Fig. 1): infants in the Group 1 (BW less than 1000 g) had a lower increase in all measurements throughout the follow-up period. We also detected a decrease in all diaphragm thickness in Group 1 patients from the third day of life to the first and second week, which was not evident in patients from Group 2 (see Fig. 3).
The comparison of the simplest model to predict the evolution of the left DTF, which uses time since birth as the only variable, with the final model that also included any grade BPD, was statistically significant (LR chi 2 4.23, p = 0.04).

Discussion
We have shown that growth in terms of diaphragm thickness can be monitored in preterm infants using DU. To the best of our knowledge, this is the study with longer followup of monitoring diaphragm thickness in this population by ultrasound.
Although diaphragm thickness and DTF have been related to time on IMV in adults [1,2] and pediatric patients [3], contrary to what we might expect, the duration of IMV does not seem to influence diaphragm thickness growth in premature patients, according to our results. We believe that it is challenging to monitor diaphragm thickness, or its normal values, in patients whose growth is exponential during the first weeks of life, unlike to other ages: critical adult or pediatric patients will not exhibit increased diaphragm thickness in 1-month time, so it is easier to compare these values before and after the use of IMV. However, as we have demonstrated, the preterm infant's diaphragm does grow after birth, while in parallel, the patient is receiving IMV. This is probably one reason why the use of IMV does not influence this infants' diaphragm increase in thickness.
Another explanation for this difference between preterm infants and older children or adults could be the fact that many NICU patients do not receive sedation and analgesia while on IMV; in our center, we do not use it routinely, and we even excluded those patients who have received it during their admission to our unit. However, sedation and analgesia are common treatments in critical adult and pediatric patients. It is possible that it is the sedation itself that influences the loss of diaphragm thickness in these patients by preventing the muscle from continuing to exercise during IMV [24]. In any case, with the design of our study, we cannot rule out that diaphragmatic function or growth is reduced during IMV in premature infants who receive it; nonetheless, it is possible that after IMV withdrawal, a "catch-up period" occurs that allows the diaphragm function and thickness to recover to values close to those of the remaining premature patients without IMV. This could be the explanation of the differences found with a very recent study performed by Dr. Hoshino and coworkers [15]: they performed DU in infants born before 28 weeks that required IMV within the first 6 h of life and they showed a decrease in RIT and right DTF of 30% 24 h after the intubation, but with similar values for the rest of follow-up (4 days). Although these findings  can be consistent with the existence of diaphragm dysfunction in this population, the authors recognize the lack of nonintubated infants, as well as the shortness of follow-up. As the higher decrease was seen in the first hours, but not afterward, it is also plausible that this decrease in diaphragm measurements would be related to a decrease in work of breathing in a subgroup of infants with severe respiratory distress syndrome after the intubation, or maybe with the use of sedatives. Extremely low BW infants showed thinner diaphragms since birth, but also slower increase in all diaphragmatic thickness, compared to infants born with a BW between 1000 and 1500 g. The aforementioned difference in these variables since birth is easily understood, as the variable that differentiates the two groups was their BW: our group demonstrated previously that diaphragm thickness at birth was higher in term compared to preterm infants [12], so diaphragm size measured by ultrasound seems have a positive relationship with BW. However, the difference in the increase of diaphragmatic thickness during the followup, mainly in expiratory measurements, had never been described before: the explanation of this finding is out of the scope of this manuscript, but could be related to the lower increase of BW since birth in infants born with extremely low BW [25], or with different respiratory settings in both groups [13]. In any case, neither of these two hypotheses could be analyzed with our data since we did not gather the progression of our patients' weight or respiratory support, which could be interesting to assess in subsequent studies.
Another, but slight difference between both groups was detected in the first weeks of life: infants in Group 1 showed a decrease in all thicknesses from the first day and the third, compared with the following measurements at 1 and 2 weeks of life. This observation may be related with the greater weight loss experienced by these children compared to those with higher birth weight [25], or with higher respiratory settings required during the first days of life [26]. However, as we described above, it is very difficult to establish causality with the design of this study.
We have also found a significantly different evolution of left and right DTF: however, these results do not have a biologically plausible explanation. Although small variations in thickness are accepted between hemidiaphragms [27], both hemidiaphragms should not have a differential pattern of growth except for the existence of unilateral diaphragm paralysis or paresis. This variable, being a quotient between the difference in diaphragm thickness and the expiration thickness of the same patient, tends to remain relatively stable over time and in different age groups [12,28].
Another variable that does seem to influence diaphragmatic function is BPD: BPD is related to higher DTF values in our patients, as was previously shown by another group [16]. From our results, we can see that it may not be the duration of IMV that causes patients with BPD to have greater DTF: as Dr. Yeung and coworkers explained in their discussion [16], it is possible that the lung disease itself imposes greater work of breathing in these patients, which translates into a similar diaphragm thickness but greater DTF compared with those in other preterm infants without BPD or those in term neonates.
DTF is also affected in adult patients who require abdominal or liver surgery [20], and for this reason, we studied whether NEC, the main abdominal pathology in premature infants, could be related to diaphragm thickness or DTF, without definitive results as only one patient in the sample suffered this disease. Although sepsis has also been related to early diaphragmatic dysfunction and atrophy in adult patients [19], in our population, LOS did not affect diaphragm growth or DTF evolution.
In any case, the results of this study must be considered with caution. First, DU is a measure that has moderate interobserver agreement [12] and that requires extensive training to perform it. To try to overcome this limitation, all the exams were performed by an expert operator in neonatal diaphragmatic ultrasound. Furthermore, the configuration of the premature infant's diaphragm is different from that in adults: the shape is flatter, which makes the apposition zone smaller with a greater angle of insertion [29]. This fact makes measurements in this region more difficult and less reproducible as well.
Second, any patient included received sedation during their stay in our unit: the main reason was to avoid the influence of sedation on the diaphragm, and to gather a more homogeneous sample of infants with more strict inclusion criteria. However, this selection may have increased the incidence of patient-ventilator asynchronies in our sample, which was not monitored during the follow-up, and it may have influenced in the diaphragmatic measurements.
And lastly, this was a single-center study, in which we lacked the variability of different NICUs. The fact that DU requires more training than lung ultrasound, for example, limits the ability to carry out studies involving different units. We also acknowledge that our patients had low values of SNAPPE-II which may reduce the generalization of the study data. Further research in this field is warranted, mainly in the field of extubation failure of preterm infants, or patients needed long periods of noninvasive ventilation in NICUs.

Conclusions
In our population we found that the higher the BW, the higher diaphragm thicknesses at birth and follow-up. Contrary to the previously published findings in adult and pediatric settings, we were unable to detect an association between days of IMV and increase of diaphragm thickness in this population. The only variable related to left DTF was the diagnosis of BPD.