Timing of Elective Delivery in Gastroschisis: A Decision and Cost Effectiveness Analysis

Abstract

Objective

To determine the most cost-effective delivery timing in pregnancies complicated by gastroschisis using a decision-analytic model.

Methods

We created a decision analytic model to compare planned delivery at 35, 36, 37, 38, & 39 weeks. Outcomes considered were stillbirth, death within 1 year of life, & respiratory distress syndrome (RDS). Probability estimates of events (stillbirth, complex gastroschisis, and RDS at each gestational age, and risk of death in simple and complex gastroschisis), utilities, & costs assigned to the outcomes were obtained from published literature. Cost analysis was from a societal perspective using a willingness to pay threshold of $100,000 per surviving infant. Outcomes and costs were considered through 1 year of life. Multi-way sensitivity analyses were performed to address uncertainties in baseline assumptions.

Results

In the base case analysis, delivery at 38 weeks is the most cost-effective strategy. Planned delivery at 35 weeks was associated with the fewest stillbirths and deaths within 1 year, due largely to a difference in ongoing risk of stillbirth. In Monte Carlo simulation when every variable was varied over its entire range, delivery at 38 weeks is cost-effective compared to 39 weeks in 76% of trials and delivery at 37 weeks is cost-effective in 69% of trials. Delivery at 38 weeks resulted in 3 additional cases of RDS for every 100 stillbirths or deaths within 1 year prevented.

Conclusions

In pregnancies complicated by gastroschisis, the most cost-effective timing of delivery is 38 weeks. Few additional cases of RDS are caused for every 1 stillbirth or death within 1 year prevented with delivery at 37–38 weeks.

Introduction

Gastroschisis is a paraumbilical, full-thickness abdominal wall defect that results in extrusion of bowel into the amniotic cavity, uncovered by any membrane.^1–4 Although other anomalies are typically absent, gastroschisis can be associated with a significant risk for stillbirth, as high as 12% in some series.^{5, 6} In addition to antenatal testing, many recommend delivery prior to 39 weeks in order to minimize the risk of stillbirth.^7–12 Another purported benefit of early delivery is to prevent the development of complex gastroschisis (defined as the presence of bowel atresia, perforation, necrosis or volvulus), which may develop with prolonged exposure of the fetal bowel to the amniotic fluid. Complex gastroschisis is associated with significantly increased risks of neonatal mortality, as well as longer durations of hospitalization.^{13, 14}

While early delivery may decrease the risk of stillbirth and complex gastroschisis, early term and late preterm births carry increased risks of neonatal morbidity and mortality, particularly respiratory morbidity.^{15, 16} The rarity of gastroschisis, combined with the infrequency of stillbirth and respiratory distress at term, makes a randomized control trial to evaluate the timing of delivery in gastroschisis impractical. Therefore, we sought to determine the optimum timing of elective delivery in pregnancies complicated by gastroschisis using a decision analytic and cost-effectiveness model.

Methods

We created a decision analytic model to estimate the most cost-effective timing of elective delivery for pregnancies complicated by gastroschisis (Figure 1). We compared planned elective delivery at 35, 36, 37, 38, and 39 weeks of gestation. Outcomes considered were stillbirth, composite mortality (defined as stillbirth and death within 1 year of life), complex versus simple gastroschisis, and respiratory distress syndrome (RDS).

Figure 1.

Decision Analytic Model

It was assumed that any patient entering the model would be 34 weeks gestation with a live fetus and no indication for delivery. The risk of stillbirth at 35 weeks gestation was therefore estimated as the point estimate of stillbirth in the 35^th week of pregnancy in pregnancies complicated by gastroschisis. Thereafter, the risk of stillbirth was estimated as the cumulative risk of stillbirth for each additional week of pregnancy. As this was an analysis of the timing of elective delivery of gastroschisis, we assumed that women in this model would not require delivery (either for spontaneous labor, preeclampsia, or non-reassuring fetal testing) until the planned timing.

We conducted a systematic literature review searching the PubMed database of English articles using the MeSH term and keyword term gastroschisis. Articles considered for review were randomized control trials, prospective cohorts, retrospective cohorts, and systematic reviews and meta-analyses that reported stillbirth and/or neonatal outcomes by either gestational age or by complex versus simple gastroschisis. As neonatal outcomes have changed significantly due to advances in neonatal care, articles published prior to 1998 were not considered. Due to variations in the definition of RDS and wide probability ranges in the literature, articles were only used for point estimates in this analysis if all gestational ages from 35–39 weeks were included in the analysis. If only a single probability point estimate was available, a range was defined by the 95% confidence interval, calculated using an exact 95% confidence interval of binomial proportions.

Cost estimates were derived from the literature (Table 1). Estimates of cost included the cost of antenatal testing (per week), hospital stay (per day) in neonatal intensive care unit, and respiratory distress syndrome. The cost of hospitalization included all costs related to hospitalization including surgery and total parenteral nutrition. To account for regional variation in costs, estimates were varied widely around the point estimate. Due to wide variation in the mode of delivery (vaginal versus elective cesarean), and lack of data regarding the risk of cesarean by gestational age in pregnancies complicated by gastroschisis, we did not include the cost of delivery in this model. Because all costs in this analysis are encountered within 1 year of life, no discounting was used. The analysis was performed from a societal perspective, using a willingness to pay threshold of $100,000 per survivor.

Table 1.

Probabilities and Costs

	Base Case	Range	Reference
Probability of Stillbirth
35 wks	0.004	0.0039–0.0047	South et al 201321 Schaffer et al 201322
36 wks	0.011	0.010–0.012
37 wks	0.014	0.013–0.015
38 wks	0.022	0.021–0.023
39 wks	0.029	0.028–0.03
Probability of Complex Gastroschisis
35 wks	0.12	0.04–0.27	Charlesworth et al 200723 Huang et al 200224 Logghe et al 200520 Maramreddy et al 200925
36 wks	0.12	0.04–0.27
37 wks	0.07	0.028 – 0.1	Baud et al 20137 Charlesworth et al23 Logghe et al 200520 Maramreddy et al 200925 Puligandla et al 200426
38 wks	0.07	0.028 – 0.184
39 wks	0.07	0.028 – 0.184
Probability of Neonatal Death
Simple Gastroschisis	0.031	0.016–0.08	Bradnock et al 201113 Emil et al 201214 Gelas et al 200827 Vacharajani et al 200728
Complex Gastroschisis	0.053	0.03–0.58
Length of Stay
Simple Gastroschisis	40	20–365	Bradnock et al 201113 Emil et al 201214 Driver et al 200129 Gelas et al 200827 Gorra et al 201230 Payne et al 200931 Vachharajani et al 200728
Complex Gastroschisis	104	36–327
Probability of RDS
35 wks	0.062	0.052–0.099	Bailit et al 201032 Gouyon et al 201033 Hibbard et al 201034 McIntire et al 200835 Cain et al36
36 wks	0.034	0.029–0.099
37 wks	0.01	0.008–0.059
38 wks	0.003	0.002–0.059
39 wks	0.0025	0.002–0.063
Costs
Cost of Hospitalization per day	3190	2314–14957	Healthcare Cost and Utilization Project, KID Database37 Lao et al 201038
Cost of Neonatal Death	34616	24403–44828	Healthcare Cost and Utilization Project, KID Database37
Cost of RDS	71022	66494–75546	Healthcare Cost and Utilization Project, KID Database37
Cost of Antenatal Testing	444	71–888	Goeree et al 199539

To address uncertainty regarding several of the baseline assumptions and probability estimates, sensitivity analyses were performed varying estimates of probability, utility and cost across their plausible ranges, alone and in combination. Monte Carlo simulation was performed to simultaneously vary all values across their plausible ranges at random over 10,000 iterations to estimate the frequency that the conclusion of the model is concordant with the base case analysis.

All computations were performed using TreeAge Pro Software, 2009, Williamstown, MA. As no human subjects were involved with this study, institutional review board approval was not obtained.

Results

Of 756 articles identified (literature search performed 11/2013), 50 articles were identified by title/abstract review to be relevant to the analysis. An additional 5 articles were identified by review of reference lists. These 55 articles were reviewed by 2 independent reviewers (LH, KG) to obtain the base case probabilities. The probability ranges, for use in the sensitivity analysis, were defined as the extreme low and high values of the probability available in the literature (Table 1). In order to determine the probability of respiratory distress at a given gestational age, we conducted a second systematic literature review searching the PubMed database for English articles using the MeSH term and keyword terms “respiratory distress syndrome, newborn,” “gestational age,” “late preterm,” and “neonatal respiratory distress.”

The search for neonatal outcomes identified 701 articles, of which 20 were selected for further review based on title and abstract. One additional article was identified after review of reference lists.

Table 2 demonstrates the results of the base case analysis. In the base case analysis, planned elective delivery at 38 weeks was the most cost-effective strategy, with an incremental cost effectiveness ratio (ICER) of $88,348 per surviving infant at one year of life. For elective delivery at 37 weeks, the base case analysis demonstrates only 50 additional cases of RDS to prevent 100 composite mortality cases, although the ICER was slightly greater than the a priori willingness to pay threshold of $100,000.

Table 2.

Base Case Analysis

Strategy	Stillbirths per 10,000	Composite Mortality* per 10,000	Cases of RDS per 10,000	Cases of Complex Gastroschisis per 10,000	Cases of RDS per 100 Stillbirths Prevented	Cases of RDS per 100 Composite Deaths Prevented	Incremental Cost Effectiveness Ratio (2013$/Survivor)
Planned Delivery at 35 weeks	40	375	597	1195	229	248	$696, 230
Planned Delivery at 36 weeks	110	443	325	1187	168	185	$827,724
Planned Delivery at 37 weeks	140	461	95	690	48	50	$121,760
Planned Delivery at 38 weeks	220	538	28	685	3	3	$87,305
Planned Delivery at 39 weeks	290	606	23	680	Baseline	Baseline	Baseline

^*Composite mortality is the sum of stillbirths and deaths within one year of life

Planned elective delivery at 35 weeks resulted in the fewest stillbirths and in the lowest composite mortality but was associated with the highest incidence of RDS and of complex gastroschisis. As gestational age at planned elective delivery increased, so did the incidence of composite mortality, largely due to an increased risk of stillbirth. However, the number of cases of RDS and complex gastroschisis decreased as planned gestational age at elective delivery increased.

We then calculated the number of cases of RDS per stillbirth and composite mortality cases prevented, using planned elective delivery at 39 weeks as the reference. Compared to planned elective delivery at 39 weeks, planned elective delivery at 38 weeks resulted in an additional 3 cases of RDS for every 100 stillbirths prevented and every 100 composite mortality cases prevented. Elective delivery at 37 weeks was associated with 48 cases of RDS for every 100 stillbirths prevented and 50 cases of RDS for 100 composite mortality cases prevented.

A Monte Carlo simulation was performed to address uncertainties in the model (Table 3). In 71% of trials, a planned elective delivery at 38 weeks was superior (i.e. associated with decreased cost and increased survival) to a planned elective delivery at 39 weeks. Planned elective delivery at 38 weeks was associated with an increased survival and an increased cost less than the willingness-to-pay threshold of $100,000/survivor in another 5% of trials. Planned elective delivery at 38 weeks was inferior (i.e. increased cost and decreased survival) in only 7% of trials.

Table 3.

Results of Monte Carlo Simulation shown as the percentage of trials in which each strategy was superior or inferior to a strategy of planned Delivery at 39 weeks.

	Planned Delivery at 38 Weeks	Planned IOL at 37 Weeks	Planned Delivery at 36 Weeks	Planned Delivery at 35 Weeks
Superior (Decreased Cost, Increased Survival)	71.0%	54.0%	32.0%	17.0%
Increased Survival, Increased Cost <$100,000	5.0%	15.0%	16.0%	21.0%
Increased Survival, Increased Cost >$100,000	16.0%	26.0%	47.0%	58.0%
Inferior (Increased Cost, Decreased Survival)	7%	3.0%	5.0%	4.0%

Planned elective delivery at 37 weeks was also superior to planned elective delivery at 39 weeks in 54% of trials, and associated with increased survival and a cost below the willingness-to-pay threshold in an additional 15%. Planned elective delivery at 35 and 36 weeks was favorable in 40–55% of trials, but largely associated with costs above the willingness to pay threshold. Planned elective delivery at 35–37 weeks was inferior to 39 weeks in fewer than 5% of trials.

Discussion

In pregnancies complicated by gastroschisis, the most cost-effective timing of elective delivery is 38 weeks. Only 3 additional cases of RDS are caused for every 100 composite mortality cases prevented with delivery at 38 weeks. Elective delivery at 37 weeks may also be preferred to 39 weeks, as it is associated with an acceptable increase in the cases of RDS for every 100 composite mortality cases prevented (50 cases). Although the increased cost of delivery at 37 was above the willingness to pay threshold in the base case analysis, elective delivery at 37 weeks was preferred over 39 weeks in 69% of simulations in the Monte Carlo analysis.

Although one purported benefit of early delivery in gastroschisis is decreased incidence of complex gastroschisis, theoretically by limiting bowel exposure to amniotic fluid, this was not borne out in our systematic review of the literature. It is worth noting that in our systematic review and our decision analytic model, late preterm deliveries were associated with an increased number of cases of complex gastroschisis compared to term deliveries. Because we based our estimates on retrospective studies, we are unable to distinguish whether infants were delivered due to abnormal ultrasound findings associated with complex gastroschisis, or, alternatively, complications of gastroschisis arose after preterm delivery. It is possible that preterm delivery may decrease the risk of in utero bowel complications but increase the risk of postnatal bowel complications. It is also possible that these retrospective studies are confounded by the indication for delivery; at some institutions it is common practice to deliver preterm if bowel dilation is noted or complex gastroschisis is suspected. This would falsely increase the estimated incidence of complex gastroschisis with late preterm delivery. Additionally, neonatal outcomes are typically reported by pediatric surgeons and neonatologists; these reports do not typically included prenatal ultrasound findings and therefore we could not exclude subjects with findings of prenatal bowel dilation. However, a recent retrospective study analyzed both the impact of gestational age at delivery and the finding of prenatal bowel dilation on ultrasound.¹⁷ This study found a strong association between the complex gastroschisis and gestational age, but not between complex gastroschisis and prenatal bowel dilation.

The question of when to electively deliver a pregnancy complicated by fetal gastroschisis has not been adequately answered in prior studies. Several retrospective studies compare “early” versus “late” delivery, with varying definitions of early (35–37 weeks) and late (>36–38 weeks).^{7, 8, 11, 18, 19} The decision to deliver in these studies are typically based on individual provider practice patterns or time periods associated with changes in delivery policy at a single institution, introducing many confounding factors other than gestational age at delivery. Additionally, as gastroschisis is a rare exposure and stillbirth is a rare outcome, these studies have not been adequately powered to address the question of which gestational age results in the highest survival rates. Some studies do not even report the incidence of neonatal death in each group.^{8, 11, 19}

Logghe et al performed a randomized control trial of elective delivery at 36 weeks versus expectant management, with the primary outcome of time to full enteral feeding and duration of hospital stay.²⁰ Compared to expectant management, the 21 infants randomized to early delivery did not have a shorter time to full enteral feeding or a shorter hospital stay. Further, 2 infants in the early delivery group died from short gut complications.

Due to the rarity of both the exposure (gastroschisis) and the outcomes of interest (stillbirth near term, RDS), an adequately powered randomized control trial to examine these clinically meaningful outcomes is impractical. Using a randomized control trial to demonstrate a reduction in composite mortality cases from the model’s incidence of 6.1% at 39 weeks to 4.6% at 37 weeks, 3664 patients per group would have to be enrolled. Therefore, we attempted to answer this fundamental question utilizing a decision and cost effectiveness analysis.

This study design has inherent limitations. Although our model and probability estimates are based on an exhaustive literature search, we are limited by the body of literature published on gastroschisis. We attempted to compensate for this by varying the probabilities around a range in the sensitivity analyses, commensurate with the level of evidence. As many studies of gastroschisis are small, the ranges used in the sensitivity analyses tend to be wide. Estimates for the risk of stillbirth came from a metaanalysis of previously published studies and from birth certificate data; consequently, some gastroschisis cases may not have been detected prenatally. These cases would not have undergone antenatal testing; therefore, estimates of stillbirth risk by week may be overestimated. Additionally, many studies of neonatal outcomes in gastroschisis divide patients simply by term (≥37 weeks) versus preterm (35–37 weeks). As such, our point estimates were the same or very similar for 35–36 weeks and 37–39 weeks. This would likely serve to decrease any differences that we could detect between 37, 38, and 39 weeks, thus biasing our findings toward delivery at 39 weeks. Additionally, this precludes us from including an analysis of expectant management until 40 and 41 weeks of gestation.

In addition, the definition of respiratory distress varies widely in the literature. In order to limit the impact of this variation, we only utilized studies which reported the incidence of RDS at each gestational age of interest (35–39 weeks). Thus, while variations in RDS definition may have widened the ranges used for the sensitivity analyses, the error is present equally at each gestational age. In other words, the incidence of RDS at each gestational age is not unduly influenced by the definition used in a single study.

Finally, our model considers the cost of antenatal testing per each additional week of gestation as this is considered standard of care in pregnancies complicated by gastroschisis. However, the impact of antenatal testing on the ongoing risk of stillbirth in pregnancies complicated by gastroschisis is unclear. The point estimates for the risk of stillbirth were derived from a large population-based study and from a meta-analysis, neither of which are able to account for the impact of antenatal testing on the risk of stillbirth. As a result, the risk of stillbirth at each gestational age may be over-estimated in a population undergoing antenatal testing, but costs may be underestimated due to changes in mode of delivery (i.e. cesarean for non-reassuring fetal testing) and earlier gestational ages at delivery.

Despite these limitations, we feel that useful conclusions can be drawn from this analysis. First, delivery at 39 weeks is associated with increased risks of stillbirth but decreasing risks of death within one year of life, complex gastroschisis, and RDS. In our model, early term deliveries (37–38 weeks) were associated with decreased risks of stillbirth and death within 1 year of life with minimal increases in the number of cases of RDS. Therefore, in pregnancies complicated by gastroschisis, early term delivery may be an acceptable approach, with delivery at 38 weeks being the most cost-effective strategy.