A simple step to improve fat and calorie delivery of human milk delivered via bottle-feeding pump, an experimental study.

Abstract

Background:

Enteral feeding-pump systems deliver decreased amounts of macronutrients in human milk to neonates. This study determined the macronutrient loss associated with a bottle-feeding pump system and the effect of manually mixing the human milk during extended feeds.

Methods:

Macronutrient content from samples of donor human milk was analyzed after simulated extended feeds with a bottle-feeding pump system using a human milk analyzer. Simulations were repeated using manual mixing of the bottle every 30 minutes during feeding. Percent of baseline was calculated and one-sample t-tests and ANOVA analysis compared the effect of manual mixing and duration of feeding on macronutrient delivery.

Results:

The delivery of fat and calories was lower over time but manual mixing considerably improved retention. Length of feeding impacted fat delivery with less fat delivered over time (p < 0.001). Manually mixing significantly increased fat delivery (p < 0.001). Similar results were found for calories with a significant reduction in calorie delivery over time (p < 0.001) and significantly more calories delivered with mixing (p < 0.001). Mixing and duration of feeding had minimal effect on protein or carbohydrate delivery.

Conclusions:

Bottle-feeding pump systems are associated with significant reduction in delivery of fat and calories of donor human milk. Manual mixing of donor human milk during prolonged feeds is a simple way to improve fat and calorie delivery to the neonate.

1. Introduction:

Human milk is widely acknowledged as the gold standard of nutrition for infants in the neonatal intensive care unit (NICU), especially very low birth weight and low birth weight infants.¹ Feeding pumps are often used to deliver human ilk via an orogastric or nasogastric tube to premature infants until they develop more mature oral-motor skills needed to feed orally.² A known challenge of enteral feeding pumps is fat deposition on the feeding circuit resulting in fat loss, which accounts for up to 55% of caloric content of human milk and provides the most energy dense calories.^3–5 The reduction in fat delivery is exacerbated when feeds are delivered over an extended period of time, which is a common practice in many NICUs due to feeding intolerance.^6,7 This fat loss can be detrimental to premature infants as insufficient caloric intake leads to postnatal growth failure and is associated with increased morbidity,^8,9 while improved caloric delivery is associated with improved neurodevelopmental outcomes in high-risk preterm infants.¹⁰ Many important nutrients and fat-soluble vitamins are also associated with milk fat, and thus decreased fat delivery may have further nutritional implications for neonates.¹¹ Given these findings, it is imperative to evaluate strategies to improve fat delivery.

Previous studies have investigated strategies to limit fat loss and have found decreased fat loss with shortened duration of feeds, shortened length of pump tubing, addition of cream emulsifiers and manual agitation of the syringe of human milk during a pump feed; NeoMed^® created an eccentric or “off center” nozzle syringe with a flattened plunger head and O-ring gasket in an effort to promote human milk flow and reduce fat adherence.^12–15 These studies and innovations have mainly focused on in-vitro laboratory simulations and have not translated to universally accepted changes to infant feeding systems in the NICU. Additionally, these studies focused on syringe feeding pumps, however bottle-feeding pumps or “screw cap” delivery pumps, as shown in Figure 1, are now increasingly used in many NICU settings given their ability to accurately deliver human milk in very low volumes and minimize milk left in the tubing; bottle-feeding pumps are less expensive than syringe pumps and incorporate several safety features that reduce the likelihood of tubing misconnections or programming errors, as compared to syringe feeding pumps¹⁶. To date, there is no published study evaluating the fat loss associated with a bottle-feeding pump and we identified no strategies to improve fat retention with a bottle-feeding pump in the current literature. Therefore, we performed the current experiment with a primary aim of determining the macronutrient loss of donor human milk associated with a bottle-feeding pump system. Our secondary aim focused on evaluating the effect of manually mixing the milk during simulated feeds as a method to potentially improve macronutrient delivery while using a bottle-feeding pump system.

Figure 1.

Depiction of enteral bottle-feeding pump system

2. Materials and Methods:

This experimental study utilized surplus pooled and previously pasteurized donated human milk (DBM) from The Milk Bank at Riley Hospital for Children. The milk was kept in a freezer stored at −20°C, per unit protocol; it was thawed prior to the experimental simulations via a Thermo Scientific Precision General Water Bath (Thermo Fisher Scientific, Newington, NH). A 3mL sample of DBM was warmed to 38°C in accordance with Miris Human Milk Analyzer ^™ (HMA, Miris AB, Uppsala, Sweden) protocol and homogenized by application of high frequency ultrasonic waves via the Miris Ultrasonic Processor^™. Calibration of the milk analyzer was performed via the check procedure with zero-level adjustment along with a validation of the internal calibration with Miris Calibration Control Kit. The DBM sample was then analyzed for baseline macronutrient content.

General Simulation Setup

The DBM feeding simulation was performed using our standard NICU tube feeding pump (ZEVEX Enteral Nutrition Delivery Systems, Infinity Orange^™, Utah USA), a 120mL plastic milk bottle (Abbott Nutritional Products, Abbott Park, IL) connected to an extension set (Enteral Pump Delivery Set with ENFit^® Connector, ZEVEX Enteral Nutrition Delivery Systems, Infinity Orange^™, Utah USA) with an attached nasogastric (NG) feeding tube (NeoMed PUR 6.5Fr, length 60cm, polyurethane, ref FTL6.5P-EO), as shown in Figure 1.

Feeding Setup and Analysis

The feeding pump was primed with 4mL of DBM warmed to 38°C per unit protocol. An additional 16mL of warmed DBM was placed in the bottle, which was then connected to the feeding pump; a volume of 16mL was used for all experiments as it represents a single feeding in an 800 gram neonate receiving full enteral feeds at 150 mL/kilograms/day. DBM was delivered through a NeoMed 6.5Fr polyurethane 60cm length NG tube over 15 minutes, 30 minutes, 60 minutes, 90 minutes and continuously over 4 hours or 240 minutes; separate 16mL samples of DBM were used for each simulation. Simulated feedings were infused until full volume of the bottle was delivered into the plastic centrifuge tube, at which time the milk was analyzed with the human milk analyzer. Simulations were repeated with the same volumes and same rates but for the 60-minute, 90-minute and 4-hour simulations, feeds were paused at 30 minute intervals, at which time the bottle of DBM was manually turned upside down and then righted a total of 10 times over a 15 second period. Feeds were then resumed at the previous rate and the process was repeated every 30 minutes until completion. The manually mixed and unmixed configurations were tested 9 times at each feeding rate to increase precision.

Nutrient Analysis

Macronutrient analysis was performed using the Miris HMA^™, which reports fat content, calorie content, true protein content, crude protein content, carbohydrate content. The 3mL sample of DBM collected prior to simulations was homogenized via the Miris Ultrasonic Processor and analyzed as baseline DBM macronutrient content prior to infusion using the HMA. The DBM collected at the end of each simulated feed, which represents the milk delivered to the neonate, underwent homogenization via the Miris Ultrasonic Processor and then macronutrient content analysis via the HMA.

Data Collection and Statistical Analysis

The total amount of fat, calories, protein and carbohydrates delivered was calculated and expressed as a percentage delivered relative to the baseline content given the variation in the macronutrient content of the different samples of pooled donor human milk; given that different samples were used for each simulation, there were no repeated measures to account for. To determine whether nutrient percentage differed from baseline in unmixed samples, we performed one-sample t-tests (comparing with 100), correcting for multiple comparisons using the Holm-Bonferroni method. Separate two-way ANOVAs were used to assess the effects of mixing (mixed vs. not mixed) and duration (60, 90 or 240 minutes) on percentage of fat, carbohydrates, and calories relative to baseline (100%). ANOVA results are presented where F is the ANOVA test statistic, the numbers in parentheses are degrees of freedom and η_p² is partial eta-squared, a measure of effect size. Significant main effects were followed up with Tukey’s post-hoc tests. Percentages of fat and calories between mixed and non-mixed samples at each time were compared with post-hoc tests using Holm-Bonferroni corrections. Protein values were not normally distributed, and the variance was quite different between groups, with two groups having no variance (all values were 100% of baseline). For these reasons, the impact of duration and mixing on protein concentration were assessed using separate Kruskal-Wallis tests (with Holm-Bonferroni-adjusted Dunn post-hoc comparisons). Kruskal-Wallis tests results are presented with an H value (the test statistic) and the effect size measurement of eta-square (η2) which was calculated using the formula (H – k + 1) / (N – k), where, k is the number of groups, and N is the total sample size.¹⁷ All analyses were performed using R (version 4.2.3).¹⁸ A p-value < 0.05 was considered statistically significant.

Results:

The baseline average donor human milk content was 4.17 ± 0.6 gm/100mL of fat, 8.1 ± 0.3 gm/100mL of carbohydrates, 75.6 ± 6.8 kcal/100mL of calories and 0.84 ± 0.1 g/100mL of protein. The percentage of fat, calories, carbohydrates and protein delivered over each time frame is summarized in Table 1.

Table 1.

Macronutrient content of non-mixed and mixed milk at various infusion rates, expressed as percentage of baseline macronutrient delivered.

Macronutrient Delivery Efficiency	15 Min No Mixing	30 Min No Mixing	60 Min		90 Min		240 Min		Significant Effect?
			No Mixing	Mixing	No Mixing	Mixing	No Mixing	Mixing	Effect of Mixing	Effect of Time	Inter-action
% Fat Delivered	96.5 ± 1.8%	82.4 ± 3.4%	60.1 ± 9.3%	77.6 ± 11.2%	57.8 ± 10.6%	85.5 ± 4.0%	42.2 ± 5.7%	69.9 ± 4.5%	Yes	Yes	No
%Calories Delivered	96.8 ± 1.2%	90.6 ± 2.2%	78.9 ± 6.7%	86.9 ± 6.8%	77.8 ± 6.1%	88.3 ± 2.8%	70.9 ± 2.4%	84.9 ± 2.3%	Yes	Yes	No
%Carbohydrates Delivered	98.1 ± 1.5%	99.1 ± 0.8%	99.6 ± 0.6%	99.6 ± 0.9%	99.5 ± 1.5%	99.6 ± 1.1%	99 ± 0.5%	99.6 ± 0.6%	No	No	No
%Protein Delivered	97.2 ± 5.5%	88.4 ± 7.4%	100%	97.2 ± 5.5%	91.8 ± 4.9%	91.9 ± 6.2%	100%	100%	No	Yes	No

Fat content of unmixed donor human milk was significantly lower than the baseline 100% for all time periods assessed (each p < 0.001) (Figure 2). While there was a relatively high amount of fat delivered at 15 min (96.5 ±1.8%), this percentage was much lower at 30 min (82.4 ±3.4%) and further diminished at longer durations: 60 min (60.1 ±9.3%), 90 min (57.8 ±10.6%), and 4 hour (42.3 ±5.7%). Calories delivered in unmixed samples yielded similar results, with all time periods having percents lower than 100% (each p < 0.001) (Figure 2). Calories delivered steadily dropped as duration increased from 15 min (96.9 ±1.2%), 30 min (90.6 ±2.2%), 60 min (77.8 ±6.1%), 90 min (77.8 ±6.1%), and up to 4-hour feeds (70.9 ±2.4%). Carbohydrates delivered in unmixed milk was close to baseline at all time points—with mean delivery rates > 98% for each group—but values were significantly lower than baseline for simulated feeds of 15 min (98.1 ±1.6%, p = 0.024), 30 min (99.1 ±0.8%, p = 0.024), and 4 hour (99.0 ±0.6%, p = 0.004) (Figure 2). Carbohydrates delivered were not significantly different from 100% at 60 min (99.6 ±0.6%, p = 0.161), and 90 min (99.5 ±1.6%, p = 0.347). Protein delivery in unmixed samples was typically close to baseline levels but was significantly lower than 100% at 30 min (88.5 ±7.4%, p = 0.005) and 90 min (91.8 ±5%, p = 0.005) (Figure 2). The percentage of protein delivered did not significantly differ from baseline at 15 min (97.2 ±5.5%, p = 0.169), or 4-hour (99.6 ±0.6%, p = 0.161). At 60 min, all protein values were 100% of baseline and a statistical test could not be performed.

Figure 2.

Macronutrient Delivery Over Time. Values expressed as percentage of baseline delivered. **: significant change (p < 0.001) observed in delivery efficiency as compared to baseline

While the delivery of fat and calories was lower for longer durations, our data showed mixing considerably improved retention (Figure 3). There was a significant main effect of mixing (F(1, 48) = 117.387, p < 0.001, η_p² = 0.71) on percentage of fat delivered with fat retention improving from 60.1% to 77.6% (p < 0.001) with mixing for feeds over 60 min, from 57.8% to 85.5% (p < 0.001) for feeds over 90 min and from 42.2% to 69.9% (p < 0.001). A significant main effect of duration on percent fat delivered was found (F(2, 48) = 17.885, p < 0.001, η_p² = 0.43). Post-hoc tests revealed that fat concentration was significantly lower at 240 min compared to 60 min (p < 0.001) and 90 min (p < 0.001), but there was no significant difference in fat delivered between 60 and 90 min (p = 0.762). The interaction between mixing and duration on delivery of fat was not significant (F(2, 48) = 2.032, p = 0.142, η_p² = 0.08), suggesting that that the impact of mixing on fat retention was not significantly influenced by time. Similar results were found for calories with significantly more calories delivered in the mixed condition (F(1, 48) = 65.216, p < 0.001, η_p² = 0.58), with calorie retention improving from 78.9% to 86.9% (p = 0.001) with mixing for feeds over 60 min, from 77.8% to 88.3% (p < 0.001) for feeds over 90 min and from 70.9% to 84.9% (p < 0.001) during 4 hour feeds also a significant reduction in calorie delivery over time (F(2, 48) =6.328, p = 0.004, η_p² = 0.21). By performing post-hoc analyses, we found that calories delivered were lower after 4 hour feeds compared to either 60 (p = 0.010) or 90 min (p = 0.009), but there was no significant difference in calories measured between 60 and 90 min feeds (p = 0.998). There was also no significant interaction between mixing and duration on percent of baseline calories delivered (F(2, 48) = 1.697, p = 0.194, η_p² = 0.07).

Figure 3.

Comparison of macronutrient delivery percentages of non-mixed donor human milk, shown in grey bars, and mixed breastmilk, shown in black bars. Values expressed as percentage of baseline delivered. **: significant change (p < 0.01) observed in delivery efficiency of mixed milk as compared to non-mixed milk at same time frame

The percentage of carbohydrates in the samples were consistent across time and mixing conditions. There was no significant main effect of mixing (F(1, 48) = 0.701, p = 0.407, η_p² = 0.01) or duration (F(2, 48) = 0.439, p = 0.648, η_p² = 0.02) on percentage of carbohydrates delivered, and the interaction between mixing and duration was not significant (F(2, 48) = 0.376, p = 0.689, η_p² = 0.02).

Protein delivery was not affected by mixing but differed as a function of time. There was no significant difference between mixing conditions (H(1) = 0.003, p = 0.959, η² = −0.02), but there was a significant effect of duration (H(2) = 19.913, p < 0.001, η² = 0.35). Post-hoc tests showed that protein concentration was significantly lower at 90 min compared to 60 min (p < 0.001) and the percentage of protein was also lower at 90 min compared to 4 hours (p < 0.001). Oddly, there was no significant difference in the percentage of protein between 60 min and 4 hours (p = 0.965). Protein values were typically the same as baseline at 60 min and 4 hour feeds, but several were below baseline level at 90 min.

Discussion:

This study is the first to demonstrate the significant reduction in fat and calorie delivery of donor human milk associated with the bottle feeding-pump system. To mitigate this macronutrient loss, we evaluated manual mixing of donor human milk as a practical method to increase fat and calorie delivery. Our results demonstrated a significant effect of mixing on the percentage of fat and calories delivered, showing that manual mixing of donor human milk as it is delivered via the bottle-feeding pump system results in improved fat and calorie delivery as compared to the current standard of practice. The mixing of donor human milk accounts for 71% of the variance of delivered fat and 58% of the variance of delivered calories in the ANOVA models; overall, we found that only an average of 54.5% ±11.7 of fat remained in unmixed samples compared with 77.0% ±8.9 when samples were mixed. We postulate that the improved fat and calorie delivery is related to homogenization of human milk and decreased gravitational separation of fat from other aqueous components of milk, similar to that described in previous studies.^3,4

Various other strategies have been reported to improve fat delivery in human milk, including ultrasonic homogenization, addition of human-milk fortifier Prolacta^® and human milk-derived cream, and the addition of lecithin.^15,19,20 Jarjour et al developed a bedside apparatus utilizing mechanical mixing methods to decrease fat loss and demonstrated significant reduction in fat losses during one-hour feeds.²¹ Our results of improved fat delivery are comparable to these previous studies. However, while these studies required additional technology or additives to limit fat loss, we improved fat retention and calorie delivery by simply inverting the bottle of milk multiple times during the feed. In addition, the feeding system used in this study was meant to resemble what we currently use in clinical practice in our NICU. We also used a polyurethane feeding tube as is done in our current clinical practice, even though it has been shown that silicone tubing can also improve fat delivery.¹² We continue to use polyurethane tubing at this time given its decreased cost, larger internal diameter, and decreased flexibility.

The percentage of carbohydrates and protein delivered were not affected by mixing, which is consistent with previous published literature. An interesting finding was the effect of duration on protein delivery; post-hoc analysis revealed the percentage of protein delivered during 90-min feeds was significantly lower than the percentage of protein delivered during 60 min or 4 hour feeds. This may be due to poor precision of the protein measurements as our average protein concentration was 0.84 ± 0.1 g/100mL, yet the HMA has a true protein detection range of 0.6–2.4g/100mL and a resolution of 0.1g/100mL.²² It is possible that there is a real reduction in protein concentration over time, however it is difficult to confirm given that our measured baseline concentration is close to the minimum detectible value. While our protein level is low and close to the minimum detectable limit of the HMA, it is still within the reported range for donor human milk noted in the literature.²³ Additionally, other studies have shown that calculation of the total protein from crude protein measurement and use of a correction factor, as is the case for the Miris HMA, may underestimate the true protein content.²⁴

This study does have limitations. We tested unfortified donor milk as the Miris HMA does not have the ability to test formula-based fortifiers; ideally, we could simulate feeds with fortified milk as many premature neonates are fed fortified human milk and previous literature has shown that certain fortifiers increase fat delivery of human milk, thus altering the macronutrients delivered ¹⁵. The pooled donor human milk utilized in our study underwent Holder pasteurization, which has been shown to affect the fat content of milk.²⁵ We did not investigate unpasteurized milk and are therefore unable to generalize our results to feeds utilizing mother’s own milk.

The fat and calorie loss seen in our in vitro study likely reflects the macronutrient loss seen in a clinical setting as we utilized the same equipment clinically used in our NICU and the feeding simulations were completed in the same way our nurses administer feeds at the bedside. Thus, our experimental findings are translatable to the clinical setting. Future studies should attempt to determine whether improved fat and calorie delivery influence clinical outcomes in high-risk neonates, including low birth weight infants and infants with intestinal failure requiring prolonged continuous feeds. Manual mixing of donor human milk is an effective measure for delivering increased fat and calories to assess such outcomes.

Conclusions:

We demonstrated a significant loss of fat and calories when utilizing a bottle-feeding pump system. Our method of simply inverting the bottle of donor human milk multiple times at fixed 30-minute intervals during prolonged simulated feeds resulted in significant improvement in fat and calorie delivery.