Diuretic Tolerance to Repeated-dose Furosemide in Infants Born Very Preterm with Bronchopulmonary Dysplasia

Abstract

Objectives

To assess the presence and timing of furosemide diuretic tolerance in infants with bronchopulmonary dysplasia (BPD), and to determine if tolerance is modified by thiazide co-administration.

Study Design

We performed an historical cohort study among infants born very preterm with BPD exposed to repeated-dose furosemide for 72 hours, measuring net fluid balance (total intake minus total output) as a surrogate of diuresis in the 3 days before and after exposure. The primary comparison was the difference in fluid balance between the first and third 24 hours of furosemide exposure. We fit a general linear model for within-subject repeated measures of fluid balance over time, with thiazide co-administration as an interaction variable. Secondary analyses included an evaluation of weight trajectories over time.

Results

In 83 infants, median fluid balance ranged between + 43.6 and + 52.7 ml/kg/d in the 3 days prior to furosemide exposure. Fluid balance decreased to a median of + 29.1 ml/kg/d in the first 24 hours after furosemide, but then increased to +47.5 ml/kg/d by the third 24-hour interval, consistent with tolerance (P < 0.001. Thiazides did not modify the change in fluid balance during furosemide exposure for any time-period. Weight decreased significantly in the first 24 hours after furosemide and increased thereafter (P < 0.001).

Conclusions

The net fluid balance response to furosemide decreases rapidly during repeated-dose exposures in infants with BPD, consistent with diuretic tolerance. Clinicians should consider this finding in the context of an infant’s therapeutic goals. Further research efforts to identify safe and effective furosemide dosage strategies are needed.

Furosemide is the most commonly used pharmacotherapy among preterm infants with grade 2 or 3 bronchopulmonary dysplasia (BPD) admitted to United States children’s hospitals.(1) Prolonged repeated-dose exposures are common in this population, with half of infants exposed to a furosemide course of at least one week in duration, and 1 in 4 infants exposed to furosemide for over a month.(2)

Characterizing the effects of a prolonged medication exposure is particularly relevant when the response may change over time. For furosemide, tolerance to its diuretic effect has been described in adults, children and younger preterm infants.(3–6) Although no standard quantitative definition exists, tolerance is described as an initial diuretic response that diminishes in magnitude over time. In a small trial of infants with evolving BPD, a statistically significant reduction in the urine output response to furosemide was noted within 3 days of sustained use.(6) Exposures that continue despite diuretic tolerance would be expected to have diminished diuretic-mediated benefits. In turn, furosemide’s pharmacotherapeutic target (the Na-K-Cl cotransporter) is ubiquitously expressed in human tissues, and various harms occur through extra-renal mechanisms.(7,8) Therefore, a reduction in diuresis over time may not reduce the likelihood of other harms, raising concern for a worsening benefit to harm profile over time.

Tolerance to furosemide diuresis is mediated by homeostatic mechanisms to preserve sodium and water balance following the inhibition of sodium reabsorption in the loop of Henle and a subsequent reduction in total body water. Prominent among these is a compensatory increase in distal tubular sodium reabsorption. This provides the rationale for “sequential nephron blockade”, whereby the co-administration of thiazides are expected to inhibit compensatory reabsorption at the sodium-chloride transporter in the distal convoluted tubule, thereby diminishing the extent of tolerance.(3,6,9)

Despite the common use of prolonged repeated-dose furosemide in infants with established BPD, diuretic tolerance has not been reported in this population, nor has the influence of thiazide co-administration. The objectives of our study were to assess the presence and timing of furosemide diuretic tolerance in infants with established BPD, and to determine if tolerance is modified by thiazide co-administration. Using net fluid balance (total intake minus total output) as a surrogate measure of diuresis, we hypothesized that after an initial decrease from baseline in the first day of furosemide exposure, fluid balance would significantly increase by the third day of sustained exposure, consistent with tolerance. Further, we hypothesized that thiazide co-administration would diminish the extent of tolerance. Secondary analyses evaluated characteristics associated with the magnitude of tolerance and weight trajectories during the study period.

Methods

Study Design, Data Source, and Sample

We performed a retrospective single center cohort study using the clinical data warehouse at Children’s Hospital of Philadelphia (CHOP). Identification of the study population is depicted in Figure 1 (available at www.jpeds.com). We included very preterm (gestational age < 32 weeks) infants admitted to the CHOP NICU between 2010–21 with grade 2 or 3 BPD as defined by the 2019 Jensen classification.(10) Qualifying furosemide administrations were of a consistent milligram per kilogram (mg/kg) dose, dose frequency, and route of administration for at least 72 hours (h); were administered at a frequency of every 12 or 24 h; lacked intermittent packed red blood cell or fluid bolus co-administrations; were preceded by at least 7 days (d) without furosemide exposure; and occurred between 36 – 60 weeks PMA. We chose a 72 h exposure duration as prior data in younger preterm infants suggests tolerance is observed within this time-frame.(6) We excluded infants with furosemide exposures in the preceding 7 d given the likely influence of prior exposures on diuretic tolerance. Lastly, we excluded infants with congenital anomalies that plausibly impact diuretic responsiveness, identifying these by reviewing International Classification of Diseases diagnostic codes. The data used in this study were approved for collection by the CHOP Institutional Review Board, #19-016420.

Figure 1.

Study cohort flow diagram

Study Variables

As our objective was to identify changes in net fluid balance over time during sustained furosemide exposure, the primary predictor variable was time, modeled as a categorical variable with 6 values: the 3 24-h intervals preceding exposure, and the 3 24-h intervals following furosemide exposures. We chose to model 24-h intervals for clinical relevance as medications are commonly prescribed for a series of 24-h intervals, e.g. “3 days of treatment.” The primary outcome variable was net fluid balance, as a surrogate measure of diuresis. Fluid balance was calculated as the net of all fluids administered minus all outputs for each time interval. Input was inclusive of medication volumes, intravenous fluids, parenteral nutrition, and enteral feeding volumes, while output was inclusive of urine, stool, emesis and drains, when present. To adjust for inexact total time between intervals, fluid balance was standardized as milliliters per kilogram per (24 h) day (ml/kg/d), accounting for fractional hours in the denominator. As an interaction variable, we identified the absence or presence of chlorothiazide or hydrochlorothiazide on the day the qualifying furosemide exposure began.

To identify drug and infant characteristics associated with the magnitude of tolerance, we quantified tolerance for each infant as the change in fluid balance between the first and third 24-h interval following furosemide exposure. We considered the following drug characteristics: route of administration, IV-equivalent dose, assuming a bioavailability of 50% for enteral formulations, such that a 2 mg/kg enteral dosing was reported as a 1 mg/kg IV equivalent (11,12); and dose frequency. We also considered the following infant characteristics: birth gestational age; sex; PMA at furosemide exposure; plasma albumin, blood urea nitrogen, sodium and chloride levels, ascertained at the closest time prior to the qualifying furosemide exposure, when available, and closest in time after exposure when not otherwise available; estimated glomerular filtration rate (13); and co-exposures to: thiazide diuretics, hydrocortisone, dexamethasone, and dopamine, all categorized as absent or present at the time of furosemide initiation. Lastly, to examine the relationship between time and body weight trajectories during the 6 d study period, we identified all body weights during the study period.

Statistical Analyses

Cohort characteristics were summarized with descriptive statistics. To model net fluid balance (outcome) over time, we fit a general linear model for repeated measures (GLM) with an exchangeable covariance structure. Akaike and Bayesian information criterion values were used for covariance structure selection. The model included time as a categorical variable (predictor), and thiazide co-administration as an interaction variable. As we prescribe thiazides for long-term therapy, we modeled this variable as time-invariant. This model allowed us to compare fluid balance between all time-intervals, with the change in fluid balance between the first and third 24-h interval after furosemide exposure being the primary comparison of interest, and to also assess whether thiazide co-administration modified the change in fluid balance over time. To identify drug and infant characteristics associated with the magnitude of tolerance, we used linear regression with robust variance estimates. Here, we modeled the change in fluid balance between the first and third 24-h after furosemide exposure in each infant as the outcome variable and considered all of the drug and infants characteristics described above as candidate predictor variables. We first examined the unadjusted association between predictors and the outcome in bivariable analyses, including those associated with the outcome at p < 0.10 with furosemide route, dose, and frequency (included a priori) as covariates in a multivariable model.

We completed two post-hoc analyses. First, after confirming a statistically significant decrease in fluid balance in the first 24 h after furosemide initiation (“initial response”) and a statistically significant increase in fluid balance between the first and third 24 h of ongoing furosemide exposure (“tolerance”), we assessed whether the magnitude of tolerance was correlated with the magnitude of the initial response. Second, after noting that a statistically significant increase in fluid balance following the initial response was present by the second 24-h post-furosemide interval, we examined the relationship between time and body weight trajectories by applying a linear spline model with random intercept and random slope, specifying knots at time 0 and 24 h to create three weight-trajectory segments: baseline prior to furosemide exposure (−72 to 0 h), initial furosemide response (0 to +24 h), following onset of tolerance (+24 to +72 h). For each time segment, we report the weight trajectory over time (model slope). P-values < 0.05 were considered statistically significant. Analyses were performed with STATA 16 (College Park, TX).

Results

Cohort Characteristics

Between 2010 and 2021, we identified 83 very preterm infants that met inclusion criteria (Figure 1, available at www.jpeds.com). Cohort characteristics are displayed in Table I. Infants had a median birth gestational age of 25 weeks, 65% had grade 3 BPD, and furosemide exposure occurred at a median PMA of 43 weeks. Furosemide exposures were most commonly gastric (49%), an IV-equivalent dose of 1 mg/kg (2 mg/kg enteral) (77%) and administered every 24 hours (71%). Thiazide co-administration was present in 37 (45%) infants.

Table 1.

Cohort characteristics

Characteristics	(N = 83)
Birth gestational age, median [IQR], weeks	25.3 [24.3 – 27.9]
Female sex, No (%)	35 (42)
Race, No (%)
Black	37 (45)
White	21 (25)
Other	25 (30)
Hispanic ethnicity, No (%)	10 (12)
Bronchopulmonary dysplasia severity
Grade 2	29 (35)
Grade 3	54 (65)
Postmenstrual age at furosemide exposure, median [IQR], weeks	43.1 [38.9 – 49.3]
Chronologic age at furosemide exposure, median [IQR], days	119 [93 – 166]
Furosemide route of administration, No (%)
Intravenous	19 (23)
Gastric	41 (49)
Post-pyloric	23 (28)
Furosemide dose, intravenous equivalent a, mg/kg, No (%)
0.5	19 (23)
1.0	64 (77)
Furosemide dose frequency, No (%)
Every 12 hours	24 (29)
Every 24 hours	59 (71)
Medication co-administrations, at furosemide exposure
Thiazides, No (%)	37 (45)
Dopamine, No (%)	3 (4)
Hydrocortisone, No (%)	14 (17)
Dexamethasone, No (%)	4 (5)
Baseline laboratory values, at furosemide exposure
Albumin, median [IQR], g/dL	3.3 [2.8 – 3.6]
Blood urea nitrogen, median [IQR], mg/dL b	12 [8 – 16]
Sodium, median [IQR], mmol/L	135 [133 – 137]
Chloride, median [IQR], mmol/L	100 [97 – 102]
Creatinine, median [IQR], mg/dL	0.2 [0.20 – 0.22]
Estimated glomerular filtration rate, median [IQR], ml/min/1.73 m2 c	76.5 [62.0 – 90.6]

^aIncludes both intravenous and enteral exposures, reported as intravenous equivalents using a 1:2 intravenous to enteral ratio to convert for presumed 50% bioavailability

^bn = 78, represents greatest degree of missingness

^cEstimated with Schwartz formula modified for preterm infants = (length in centimeters * k) / serum creatinine; k = 0.34

Assessment of Diuretic Tolerance and Modification by Thiazide Co-administration

Median fluid balance (ml/kg/d) over time is depicted in Figure 2. From a median range of +43.6 to +52.7 ml/kg/d, fluid balance decreased to +29.1 ml/kg/d in the first 24 hours following furosemide initiation, and then increased back to +38.5 and +47.5 ml/kg/d in the second and third 24-h intervals of sustained furosemide exposure. There was a significant difference in fluid balance between the first and third 24-h intervals of furosemide exposure, with a mean difference (95% CI) of +18.1 (+12.1, +24.1) ml/kg/d; p < 0.001. Compared with the first 24-h interval after furosemide exposure, fluid balance was significantly higher during all other pre- and post-furosemide intervals, including the second 24-h interval after furosemide exposure; p < 0.001. Thiazide co-administration did not modify the pattern of fluid balance changes over time within groups, with modest mean differences between thiazide exposed and unexposed infants and non-significant interaction p-values during all intervals, including the second (+2.6 ml/kg/d, p = 0.68) and third (+0.1 ml/kg/d, p = 0.99) 24-h intervals after furosemide initiation.

Figure 2.

Tolerance to furosemide diuresis in very preterm infants with BPD exposed to repeated-dose furosemide. Figure depicts fluid balance in the three 24-hour intervals pre- and post-exposure to furosemide. Box-and-whisker plots identify the median value (horizontal line within box), 25^th and 75^th percentile (lower and upper edges of box, respectively), lower and upper adjacent values (whiskers extending from box) and outlying values (individual dots). Fluid balance was significantly higher (p < 0.001) at all time intervals relative to first 24 hours after furosemide exposure (*) in a general linear model for repeated measures.

Hydrocortisone co-administration, dopamine co-administration and younger PMA at exposure were associated with greater tolerance at p < 0.10 in bivariable analyses and were therefore included as covariates with furosemide route, dose, and frequency (included a priori) in multivariable models. Only dopamine co-administration remained independently associated with greater tolerance following multivariable adjustment (Table II). In a post-hoc analysis, we found a significant correlation between the magnitude of tolerance and magnitude of initial diuretic response; Pearson’s correlation coefficient: −0.51, p <0.001 (Figure 4).

Table 2.

Characteristics associated with the magnitude of diuretic tolerance in very preterm infants with BPD exposed to repeated-dose furosemide

Characteristic	Diuretic tolerance, measured as the change in fluid balance between the first and third 24 hours of furosemide exposure, ml/kg/d
	Adjusted mean difference (95% CI) a	p-value
Furosemide route of administration, gastric as reference	-	-
Intravenous	−13.4 (−31.2, 4.4)	0.14
Post-pyloric	−7.2 (−18.6, 4.2)	0.21
Furosemide dose, intravenous equivalent, 0.5 as reference, mg/kg	-	-
1.0	−3.1 (−23.3, 17.2)	0.76
Furosemide dose frequency, every 12 hours as reference	-	-
Every 24 hours	−12.9 (−34.6, 8.8)	0.24
Hydrocortisone co-administration	12.4 (−5.7, 30.5)	0.18
Dopamine co-administration	18.5 (1.4, 35.6)	0.04
Postmenstrual age at exposure, per week	−1.10 (−2.22, 0.02)	0.05

^aAdjusted mean difference for change in fluid balance estimated through multivariable linear regression with robust variance estimates. For categorical variables, negative values reflect less tolerance compared to reference; positive values reflect more tolerance. Furosemide route of administration, dose and dose frequency included as covariates a priori. Hydrocortisone and dopamine co-exposure included due to association with diuretic tolerance at p < 0.10 in bivariable analysis.

Figure 4.

The magnitude of tolerance to furosemide diuresis (x-axis) is correlated with the magnitude of the initial diuretic response (y-axis) in very preterm infants with BPD exposed to repeated-dose furosemide. More negative values indicate a greater initial response, and greater positive values indicate a greater degree of tolerance. Figure depicts scatter plot with trend-line. Pearson’s correlation coefficient: −0.51, p < 0.001.

Weight trajectories over the study period

The relationship between time and body weight trajectories during the study period is depicted in Figure 5. The mean (95% CI) rate of weight gain in the 72 h preceding furosemide exposure was 41 (25, 57) grams/d. In the first 24 hours after exposure, there was a mean (95% CI) rate of weight loss of −46 (−83, −9) grams/d, followed by a mean weight gain of 12 (−12, 37) grams/d in the remaining 48 post-exposure hours. Weight gain rate (slope) changed significantly at both knots of the linear spline (furosemide initiation at time 0, and with tolerance beyond 24 h of study), and was significantly lower beyond 24 h of study vs before furosemide initiation; p < 0.05 for each comparison.

Figure 5.

Weight decreases transiently during repeated-dose furosemide exposures in very preterm infants with BPD. Figure depicts weight trajectories in the 72 hours leading up to furosemide initiation (first segment), the first 24 hours of exposure (second segment), and between 24 and 72 hours of sustained exposure (third segment). Figure depicts estimates and 95% confidence intervals generated by a linear spline mixed model with random intercept and random slope. Spline knots were specified at 0 hours (furosemide initiation) and 24 hours (onset of tolerance) post-hoc. P < 0.05 for change weight trajectory (slope) at both knots, and for the difference in slope between the first and third segments.

Discussion

Consistent with our hypothesis, net fluid balance decreased following furosemide initiation, but began to increase towards baseline values by the second exposure day. The magnitude of the initial diuretic response was approximately halved on the second day of exposure and halved again on the third day (Figure 2). A post-hoc assessment of weight trajectories over time supported our primary finding, with the decrease in weight after furosemide initiation limited to a single day (Figure 5). Our findings are consistent with reports of rapid tolerance in other populations. These include adults with congestive heart failure(3), broadly inclusive pediatric cohorts(4), and younger preterm infants with evolving “old” BPD(6). In one arm of a small three-arm trial enrolling preterm infants with evolving BPD (gestational age 31 weeks, postnatal age 30 days, on supplemental oxygen at 21 days) Segar et al. exposed 6 infants to 1 mg/kg of IV furosemide every 24 hours for 5 consecutive days. The initial increase in urinary flow rates decreased with subsequent doses, approaching baseline values and reaching statistical significance after the third dose.(6) The slight inconsistencies in the rapidity of tolerance between our studies may reflect differences in furosemide dosage, patient population, or sample size.

Contrary to our secondary hypothesis, furosemide diuretic tolerance was not modified by the co-administration of thiazide diuretics. The pattern of fluid balance over time was remarkably similar for both groups, without evidence of an interaction (Figure 3). These findings contrast with those of Segar et al., where the addition of metolazone (a thiazide-like diuretic) to the third furosemide dose prevented the development of tolerance in a second arm of their trial.(6) Our contrasting findings may also reflect differences in patient population or drug exposure (ie, metolazone vs chlorothiazide and hydrochlorothiazide), though studies in adults with congestive heart failure have reported findings similar to Segar et al. when adding hydrochlorothiazide.(9) Alternatively, we speculate that differences in the sequence of diuretic co-exposures may contribute to the observed differences. While Segar et al. administered metolazone to infants developing tolerance to repeated-dose furosemide, our study infants were already receiving thiazides at the time of furosemide initiation. Various mechanisms can contribute to fluid balance homeostasis(14), and alternative pathways may have been previously activated in our infants, obviating a reliance on increased sodium reabsorption in the distal convoluted tubule (where thiazides primarily act) to achieve tolerance. Our findings are consistent with a 2017 study by Kim et al.(4) In this cohort study of 61 pediatric patients exposed to furosemide for at least 3 consecutive days, no interaction was noted between fluid balance trends over time and concurrent diuretics.

Figure 3.

Tolerance to furosemide diuresis in very preterm infants with BPD exposed to repeated-dose furosemide is not modified by thiazide co-administration. Figure depicts fluid balance in the three 24-hour intervals pre- and post-exposure to furosemide, without and with existing thiazide co-administration. Box-and-whisker plots identify the median value (horizontal line within box), 25^th and 75^th percentile (lower and upper edges of box, respectively), lower and upper adjacent values (whiskers extending from box) and outlying values (individual dots). P > 0.05 for time and thiazide interaction at each interval in a general linear model for repeated measures.

We were unable to note strong associations between drug and infant characteristics and the magnitude of diuretic tolerance. Clinical measures of overall fluid status, such as plasma sodium and blood urea nitrogen levels, were not significant associations. Only dopamine co-administration remained independently associated with the magnitude of diuretic tolerance, with infants exposed to dopamine at furosemide initiation demonstrating a significantly greater increase in fluid balance between the first and third exposure intervals (Table II). As only three cohort infants were exposed to dopamine, these secondary findings should be interpreted cautiously. We plan to more closely evaluate characteristics associated with initial diuretic responsiveness in a larger cohort of infants. A post-hoc analysis suggested a strong correlation between the magnitude of the initial diuretic response and the extent of tolerance (Figure 4, available at www.jpeds.com). The idea that the magnitude of the homeostatic response correlates to the magnitude of the initial diuretic effect is plausible but requires further study.

Our study has several limitations. We used fluid balance as a surrogate measure of diuresis, relying on real-world clinical data captured in the electronic health record. Fluid balance is routinely reported in neonatal medicine and is a pragmatic surrogate of diuresis, as output is influenced by intake, and quantitatively distinguishing urine from stool output isn’t routinely feasible in clinical practice. Direct measures of diuresis and natriuresis may have more accurately characterized changes in furosemide effect over time, but these data were unavailable, and direct measures of diuresis would have necessitated adjustment for within-infant differences in intake over the study period. Previously published literature has similarly relied on fluid balance.(4) Our data are susceptible to information bias, such as inaccurate measurements or calculations of fluid intake and output, as well as transcription errors into the electronic health record. Carefully conducted prospective studies would help avoid these. Lastly, with 83 infants, our cohort may have lacked statistical power to detect some clinically relevant associations. However, our study is the largest pediatric cohort evaluating furosemide tolerance to date, despite describing a specific population and exposure. We focused exclusively on infants with grade 2 and 3 BPD and applied strict criteria to define qualifying furosemide exposures. Multicenter studies may allow greater statistical precision.

Our findings are relevant to clinical practice and have implications for further research. BPD is the most common cause of late death among preterm infants, and is a key driver of postnatal healthcare expenditures in United States children’s hospitals.(15,16)(17,18) Identification of optimal furosemide use is needed to prepare for trials assessing its efficacy and safety. Despite its common use, neither are well established. In the meantime, clinicians should note that the diuretic effects of furosemide decrease rapidly during repeated-dose exposures. This finding should be considered within the context of each infant’s specific therapeutic goals and may be particularly relevant when the intent is to reduce edema through sustained diuresis. Because furosemide is associated with extra-renal harms, it is plausible that repeated-dose exposures worsen its benefit to harm profile. However, we do not interpret our findings as evidence that prolonged furosemide exposures are broadly ineffective. Modest reductions in fluid balance over time could lead to cumulative diuretic-mediated benefits, while mitigating harms from excessive volume depletion. Although we noted weight loss on a single day after furosemide initiation, the rate of weight gain over the next two days was lower than at baseline. Further, furosemide has potentially beneficial extra-renal effects, including direct action on the pulmonary vasculature and bronchial smooth muscle.(7,19) Research characterizing furosemide pharmacology in established BPD, and evaluating the comparative effects of alternative dosing strategies on both renal and pulmonary outcomes are needed. Lastly, our findings highlight uncertainties about the mechanisms underlying fluid balance homeostasis. Dedicated research comparing how the order of diuretic exposures during sequential nephron blockade impacts tolerance, and better characterizing how drug and clinical characteristics – such as those reflecting overall fluid status, influence both initial diuretic response and subsequent tolerance may provide mechanistic insights to inform future therapies in this population and others.

Abbreviations

BPD

bronchopulmonary dysplasia

CHOP

Children’s Hospital of Philadelphia

CI

confidence interval

GLM

general linear model

IV

intravenous

NICU

neonatal intensive care unit

PMA

postmenstrual age