Objective
Transabdominal ultrasound (TAUS) cervical length (CL) screening has been proposed as an alternative to universal transvaginal screening to identify women at an increased risk of preterm birth. We sought to identify whether and under what circumstances TAUS would be cost-effective.
Study Design
This is a decision analytic model designed to compare an initial TAUS CL screening approach with universal transvaginal screening in a hypothetical cohort of women with a singleton pregnancy. Cost, probability, and utility estimates were derived from the existing literature.
Results
Under baseline assumptions, universal transvaginal was the dominant strategy. In comparison to TAUS, universal transvaginal CL screening reduced preterm birth by 0.03%, reduced costs by $1.2 million and increased quality-adjusted life years by 70 per 100,000 women. Although robust to many changes in many estimates, the model was sensitive to the cost of a transvaginal ultrasound, the prevalence of a short cervix and the test characteristics (ie, sensitivity and specificity) of a TAUS screening examination for short CL.
Conclusion
Compared with an initial TAUS screen, universal transvaginal ultrasound was a more cost-effective strategy under most assumptions. Optimizing TAUS testing characteristics or applying a transabdominal screening strategy in lower risk populations may yield an initial TAUS to be cost-effective.
Recent research has shown that vaginal progesterone effectively reduces spontaneous preterm birth among women with a short cervical length (CL) in the midtrimester. However, the proportion of women with a cervical length meeting criteria for this intervention is relatively small. Identification of these women requires the use of a transvaginal ultrasound (TVUS), which adds time and expense to each examination, in addition to patient discomfort. In the United States alone, universal TVUS screening would cost over $200 million annually. Nevertheless, prior decision analyses have suggested that universal TVUS screening would be cost-effective given the high costs associated with preterm birth and the corresponding savings associated with its prevention. In light of this evidence, both American College of Obstetricians and Gynecologists and Society for Maternal-Fetal Medicine have published clinical guideline documents stating that universal transvaginal cervical length screening is reasonable but not mandated.
Yet, the cost-effective analyses that have been performed have compared TVUS to no screening at all or TVUS only in high risk cohorts, which are not the only alternatives. Transabdominal ultrasound (TAUS) could be used to assess CL, and only those women with a short cervix on TAUS would then receive TVUS. Because sonography is routinely used in the midtrimester for a fetal anatomic survey, a cervical length assessment during a TAUS would not add additional cost. This strategy would decrease the number of TVUS required and therefore reduce the costs related to cervical length screening. However, with an initial TAUS screen, some women with a short CL may not be identified, disallowing the maximum reduction in preterm births and their associated costs. Additionally, a proportion of women would require a TVUS assessment of their cervix, despite ultimately being found to have a normal cervical length. The balance between the benefits and risks of a TAUS screening strategy and the cost-effectiveness compared with universal TVUS remain unknown.
Using decision analytic modeling, we compared a universal TVUS approach to a strategy that incorporated an initial TAUS CL assessment followed by a TVUS only in women found to have a short cervix on TAUS. Our aim was to identify whether and under what circumstances this TAUS strategy would be cost-effective.
Materials and Methods
This was a cost-effectiveness analysis of 2 strategies for screening CL from a societal perspective. Our theoretic cohort included women with a singleton pregnancy undergoing CL screening between 18 and 24 weeks of gestation. One strategy involved universal TVUS with vaginal progesterone administered if the CL was less than or equal to 2 centimeters. The second strategy used TAUS, performed during a fetal anatomic survey, to initially assess the cervix. Only if the TAUS CL was under a specific threshold (baseline case ≤32mm derived from Friedman et al ) was a TVUS then performed. In this scenario, only TVUS length, if performed, was used to determine if vaginal progesterone was administered. Gestational age at delivery was dichotomized to either before or after 35 weeks. This gestational cutoff was chosen to capture the impact of both early and late prematurity using available probability estimates. Each neonate could subsequently either have a normal outcome, long-term morbidity, or mortality, with proportions of each outcome weighted by the probability of delivery at different gestational ages. An outline of the decision tree was created with Tree Plan, version 1.61 (Decision Support Services, San Francisco, CA) and is depicted in Figure 1 .
Probabilities
Probabilities of each event were derived from the published literature and are listed in Table 1 . When multiple point estimates were available, a weighted average using the study sample population size was determined. Ranges used in the sensitivity analysis included both the highest and the lowest available estimates. When only one estimate was available, the binomial proportions method was used to obtain the 95% confidence interval, which was used as the range. The probability of neonatal long-term morbidity and mortality was based on a weighted average of these adverse outcomes that incorporated the probability of delivery and of an adverse outcome at each gestational age. The probability that a woman with a cervix on TVUS ≤2 cm was ascertained with TAUS was quantified as the sensitivity of TAUS. Correspondingly, the probability that a short cervix would be missed with initial TAUS screening is quantified as 1 minus the sensitivity of a TAUS. Conversely, the probability that a TVUS would be indicated by the results of the initial TAUS but ultimately a normal cervical length would be found is quantified as 1 minus the specificity of a TAUS. Under baseline assumptions, 46% of women who attempt screening with a TAUS would ultimately require a TVUS with 3.48% ultimately having a short cervix by TVUS.
| Variable | Point estimate | Range | Reference |
|---|---|---|---|
| Probabilities | |||
| CL ≤2.0 cm | 3.48% | 0.90–3.77% | |
| Sensitivity of TAUS for CL ≤2.0 cm on TVUS | 90.60% | 60.00–99.99% | |
| Specificity of TAUS for CL ≤2.0 cm on TVUS | 55.50% | 30.00–80.00% | |
| Preterm birth if CL ≤2.0 cm on TVUS | 21.95% | 20.00–22.30% | |
| RR for preterm birth with vaginal progesterone use | 0.59 | 0.36–0.90 | |
| Preterm birth with CL >2 cm | 2.35% | 2.18–2.52% | |
| Neonatal death if delivery <35 wk | 6.12% | 3.76–8.33% | |
| Severe neonatal morbidity if delivery <35 wk | 4.80% | 4.55–5.05% | |
| Neonatal death if delivery ≥35 wk | 0.08% | 0.04–0.12% | |
| Severe neonatal morbidity if delivery ≥35 wk | 1.74% | 1.71–1.77% | |
| Costs | |||
| TAUS | 0.00 | 0.00–30.00 | Estimated |
| TVUS | 61.00 | 50.00–300.00 | |
| Vaginal progesterone | 261.15 | 100.00–400.00 | |
| Severe morbidity | 1,240,297.84 | 200,000–1,500,000 | |
| Neonatal care if delivery <35 wk with neonatal death | 130,851.98 | 2244.50–587,698.75 | |
| Neonatal care if delivery ≥35 wk and neonatal death | 94,877.43 | 663.86–401,936.29 | |
| Neonatal care if delivery <35 wk and survive | 87,009.39 | 30,068.06–204,637.19 | |
| Neonatal care if delivery ≥35 wk and survive | 2963.49 | 322.19–9877.17 | |
| Utility of severe neonatal morbidity | 0.55 | 0.50–0.60 |
Given the limited literature on the sensitivity and specificity of TAUS as a screen for a short CL on TVUS, baseline estimates were derived from the literature but then varied widely in sensitivity analysis.
Costs and utilities
Costs were derived from literature review and were inflated to 2012 US dollars using the medical care component of the Consumer Price Index. Future costs were discounted at an annual rate of 3%. Utilities were derived from the published literature. Quality-adjusted life years (QALYs) (except for those with perinatal mortality) were then calculated based on an average life span of 75 years for an infant with a normal outcome. The lifespan of those born with a major morbidity ranged from 25 years to 75 years. The baseline estimates and ranges of both costs and utilities used are listed in Table 1 .
We assumed no added cost of CL measurement during a routine anatomic survey as the cervix is already evaluated as part of the routine examination. One study in which an additional TAUS CL measurement was performed supports this assumption. Thus, at baseline the cost of TAUS was set as $0. Nevertheless, because of the uncertainty of this estimate, the marginal cost of the TAUS was estimated to range in sensitivity analysis up to $30, half of the baseline cost of a TVUS.
Costs for surviving neonates were calculated as the costs of medical care at each gestational age. The cost of a neonatal demise included the neonatal costs associated with the care of newborns who required initial care, but did not ultimately survive. The cost of long-term morbidity, regardless of underlying cause, was assumed to be similar to the direct medical costs for an individual with cerebral palsy. This assumption was made given the paucity of data on long-term care costs ascribed to other complications of prematurity but the similar long-standing medical support that is needed in cases of severe morbidity. Given this assumption and the significant variation in reported costs, this estimate was varied widely in sensitivity analysis, using ranges derived from the available literature.
Analysis
We calculated the probability of preterm birth less than 35 weeks and the subsequent total costs and QALYs for each screening strategy. An incremental cost-effectiveness ratio (ICER) was calculated by dividing the difference in costs by the difference in QALYs between the universal TVUS approach and the initial TAUS screening approach. An ICER of $50,000 was considered to define cost-effectiveness. We performed univariable sensitivity analyses, and for those variables that resulted in the greatest effect on the outcome, threshold analyses were performed. Finally, we performed a probabilistic sensitivity analysis, with variables assumed to have a normal distribution, using Monte Carlo simulation. This study was exempt from institutional review board approval.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
