Cost-effectiveness of diagnostic testing strategies including cervical-length measurement and fibronectin testing in women with symptoms of preterm labor
ABSTRACT
Objective
To evaluate the cost-effectiveness of combining cervical-length (CL) measurement and fetal fibronectin (fFN) testing in women with symptoms of preterm labor between 24 and 34 weeks' gestation.
Methods
This was a model-based cost-effectiveness analysis evaluating seven test–treatment strategies based on CL measurement and/or fFN testing in women with symptoms of preterm labor from a societal perspective, in which neonatal outcomes and costs were weighted. Estimates of disease prevalence, test accuracy and costs were based on two recently performed nationwide cohort studies in The Netherlands.
Results
Strategies using fFN testing and CL measurement separately to predict preterm delivery are associated with higher costs and incidence of adverse neonatal outcomes compared with strategies that combine both tests. Additional fFN testing when CL is 15–30 mm was considered cost effective, leading to a cost saving of €3919 per woman when compared with a treat-all strategy, with a small deterioration in neonatal health outcomes, namely one additional perinatal death and 21 adverse outcomes per 10 000 women with signs of preterm labor (incremental cost-effectiveness ratios €39 million and €1.9 million, respectively). Implementing this strategy in The Netherlands, a country with about 180 000 deliveries annually, could lead to an annual cost saving of between €2.4 million and €7.6 million, with only a small deterioration in neonatal health outcomes.
Conclusion
In women with symptoms of preterm labor at 24–34 weeks' gestation, performing additional fFN testing when CL is between 15 and 30 mm is a viable and cost-saving strategy. Copyright © 2017 ISUOG. Published by John Wiley & Sons Ltd.
INTRODUCTION
In developed countries, most women presenting with symptoms of preterm labor before 34 weeks' gestation are transferred to a perinatal center and receive treatment such as tocolysis and corticosteroids1. Yet, the majority (85–95%) of these women do not deliver within 1 week of presentation2. As transfer and subsequent treatment lead to substantial costs, in addition to maternal anxiety with side-effects for both the mother and the unborn child, appropriate identification of women at low risk of immediate preterm delivery could reduce unnecessary interventions in these women, resulting in cost savings and reducing discomfort and side-effects3-6.
A recent meta-analysis has concluded that the combination of fetal fibronectin (fFN) testing and cervical-length (CL) measurement may reduce unnecessary costs compared with a single-test strategy7. We have previously evaluated the cost-effectiveness of several testing strategies, including those incorporating CL measurement and fFN assessment, for predicting preterm delivery within 7 days of presentation in women with symptoms of preterm labor at 24–34 weeks' gestation8. In that model-based analysis, estimates were retrieved from published studies, and we concluded that additional fFN testing when CL is between 10 and 30 mm was the preferred strategy.
The studies included in that cost-effectiveness analysis reported on small study groups in which both low- and high-risk populations were included. In addition, there was variation in the way fFN testing and CL measurement were combined in each study and in the definition of preterm birth. Consequently, no robust and consistent data were available, resulting in uncertainty in the results. To overcome these limitations, we set up the Assessment of Perinatal Outcome with Sustained Tocolysis in Early Labour (APOSTEL)-I study in The Netherlands, which assessed the accuracy of the combination of CL measurement and fFN testing in predicting delivery in women with symptoms of preterm labor9. Additional fFN testing in women with a CL between 15 and 30 mm was found to result in a reduction in the number of false-positive test results, without overlooking women who would deliver within 7 days of presentation, compared with either testing strategy in isolation10.
In this study, we present an evaluation of the costs and effectiveness of risk stratification based on combined CL measurement and fFN testing in women with symptoms of preterm labor at 24–34 weeks' gestation, based on the findings of our two nationwide cohort studies9, 11.
METHODS
Model structure
The structure of the decision model is shown in Figure 1. We included women with symptoms of preterm labor, intact membranes and a gestational age between 24 and 34 weeks. All women in the model had symptoms of preterm labor (contractions (more than three per 30 min), vaginal bleeding or abdominal or back pain), such that they would be tested for CL and fFN.
Treatment consisted of administration of tocolysis and steroids, combined with transfer of women to a perinatal center if they were in a general hospital. Preterm delivery was defined as delivery within 7 days of presentation, because this outcome is related to the need for immediate transfer to a perinatal center, hospitalization and treatment. We also distinguished between women who would deliver before 34 weeks and those who would deliver after 34 weeks, which allows for an assessment of differences in health-related neonatal outcomes and costs.
Seven test–treatment strategies were evaluated: (1) a treat-all, reference strategy that consisted of treating all women with symptoms of preterm labor with tocolysis and steroids, and transferring them to a tertiary center; (2) treat if CL < 25 mm; (3) treat if fFN test is positive; (4) treat if fFN test is positive and CL < 25 mm; (5) treat if either fFN test is positive or CL < 25 mm; (6) treat if CL < 15 mm or perform fFN test if CL is between 15 and 30 mm and treat if fFN test is positive; and (7) no treatment. The five management schemes (strategies 2–6) incorporated risk assessment with fFN testing and/or CL measurement. These strategies include performing a fFN test when CL is between 15 and 30 mm, which appears to be a viable option according to the APOSTEL-I study10.
In the model, we assumed that in all hospitals, facilities were available to perform CL measurement and fFN testing. Finally, we assumed that within each strategy, except for the reference and the no-treatment strategies, allocation of treatment would depend strictly on the test results.
Outcomes
The seven strategies were compared in terms of the following outcomes: proportion of patients treated, perinatal death, a composite of adverse neonatal outcomes and total costs. The composite of adverse neonatal outcomes consisted of perinatal death, chronic lung disease, neonatal sepsis, intraventricular hemorrhage greater than Grade II, periventricular leukomalacia greater than Grade I and necrotizing enterocolitis. Costs were calculated from a societal perspective. Only costs and neonatal outcome until neonatal discharge were taken into consideration because adequate information to specify model parameters in a decision model with a longer time horizon was lacking.
Model parameters
Model parameters are presented in Table 1 with the best available point estimate as the base–case, and a plausible range expressing parameter uncertainty. These parameters were calculated from the APOSTEL-I and APOSTEL-II studies9-11. The APOSTEL-I study assessed the accuracy of combining CL measurement and fFN testing in predicting delivery in 714 women with symptoms of preterm labor between December 2009 and August 2012 in The Netherlands. The APOSTEL-II study investigated the effect of prolonged tocolysis in 406 women with intact membranes and symptoms of preterm labor between 24 and 34 weeks' gestation11.
| Parameter | Base–case (range) | Distribution | Reference |
|---|---|---|---|
| Cost (2011 €) | |||
| Patient transfer after positive test | 423 (400–446) | Gamma | Dutch costing guideline13 |
| CL measurement | 52 (51–53) | Gamma | APOSTEL-I10 |
| fFN test | 75 (73–76) | Gamma | APOSTEL-I10 |
| Other diagnostics | 60 (57–63) | Gamma | APOSTEL-I10 |
| Treatment threatened labor*, † | 605 (483–727) | Gamma | APOSTEL-I10 |
| Treatment threatened labor‡ | 296 (248–344) | Gamma | APOSTEL-I10 |
| Admission if treated | 2894 (2110–3678) | Normal | Assumption |
| Antepartum admission care* | 2258 (1913–2603) | Gamma | APOSTEL-I & II10, 11 |
| Antepartum admission care† | 7505 (6304–8706) | Gamma | APOSTEL-I & II10, 11 |
| Antepartum admission care‡ | 4802 (4092–5512) | Gamma | APOSTEL-I & II10, 11 |
| Delivery* | 1499 (1276–1722) | Gamma | APOSTEL-I & II10, 11 |
| Delivery† | 1036 (972–1100) | Gamma | APOSTEL-I & II10, 11 |
| Delivery‡ | 956 (897–1015) | Gamma | APOSTEL-I & II10, 11 |
| Maternal admission postpartum | 1391 (1287–1495) | Gamma | APOSTEL-I & II10, 11 |
| No adverse outcome* | 41 092 (35 076–47 108) | Gamma | APOSTEL-I & II10, 11 |
| No adverse outcome† | 35 235 (30 328–40 142) | Gamma | APOSTEL-I & II10, 11 |
| No adverse outcome‡ | 3271 (2552–3990) | Gamma | APOSTEL-I & II10, 11 |
| Neonatal morbidity | 79 821 (69 563–90 079) | Gamma | APOSTEL-I & II10, 11 |
| Neonatal mortality | 41 154 (20 211–62 097) | Gamma | APOSTEL-I & II10, 11 |
| Extra neonatal care, no adverse outcome | 407 (363–451) | Gamma | APOSTEL-I & II10, 11 |
| Extra neonatal care, adverse outcome | 1336 (1099–1573) | Gamma | APOSTEL-I & II10, 11 |
| Extra neonatal care, mortality | 1506 (880–2132) | Gamma | APOSTEL-I & II10, 11 |
| Productivity loss, no adverse outcome | 94 (59–129) | Gamma | APOSTEL-I & II10, 11 |
| Productivity loss, adverse outcome | 949 (727–1171) | Gamma | APOSTEL-I & II10, 11 |
| Productivity loss, mortality | 359 (51–667) | Gamma | APOSTEL-I & II10, 11 |
| Travel, no adverse outcome | 151 (134–168) | Gamma | APOSTEL-I & II10, 11 |
| Travel, adverse outcome | 706 (615–797) | Gamma | APOSTEL-I & II10, 11 |
| Travel, mortality | 299 (147–451) | Gamma | APOSTEL-I & II10, 11 |
| Probability | |||
| PTD within 7 days of testing | 0.14 (0.12–0.17) | Beta | APOSTEL-I10 |
| PTD > 7 days of testing and < 34 weeks | 0.10 (0.08–0.13) | Beta | APOSTEL-I10 |
| CL < 25 mm* | 0.97 (0.91–0.99) | Beta | APOSTEL-I10 |
| CL < 25 mm† | 0.79 (0.67–0.87) | Beta | APOSTEL-I10 |
| CL < 25 mm‡ | 0.54 (0.50–0.58) | Beta | APOSTEL-I10 |
| CL < 15 mm or fFN positive if CL 15–30 mm* | 0.93 (0.86–0.96) | Beta | APOSTEL-I10 |
| CL < 15 mm or fFN positive if CL 15–30 mm† | 0.55 (0.43–0.66) | Beta | APOSTEL-I10 |
| CL < 15 mm or fFN positive if CL 15–30 mm‡ | 0.30 (0.26–0.34) | Beta | APOSTEL-I10 |
| CL 15–30 mm | 0.45 (0.41–0.49) | Beta | APOSTEL-I10 |
| fFN positive* | 0.90 (0.82–0.94) | Beta | APOSTEL-I10 |
| fFN positive† | 0.61 (0.49–0.72) | Beta | APOSTEL-I10 |
| fFN positive‡ | 0.37 (0.33–0.41) | Beta | APOSTEL-I10 |
| CL < 25 mm or fFN positive* | 0.99 (0.94–1.00) | Beta | APOSTEL-I10 |
| CL < 25 mm or fFN positive† | 0.81 (0.70–0.89) | Beta | APOSTEL-I10 |
| CL < 25 mm or fFN positive‡ | 0.57 (0.53–0.61) | Beta | APOSTEL-I10 |
| CL < 25 mm and fFN positive* | 0.86 (0.78–0.92) | Beta | APOSTEL-I10 |
| CL < 25 mm and fFN positive† | 0.49 (0.37–0.62) | Beta | APOSTEL-I10 |
| CL < 25 mm and fFN positive‡ | 0.26 (0.23–0.30) | Beta | APOSTEL-I10 |
| Perinatal death with antenatal corticosteroids* | 0.05 (0.02–0.10) | Beta | APOSTEL-II11 |
| Perinatal death with antenatal corticosteroids† | 0.04 (0.02–0.08) | Beta | APOSTEL-II11 |
| Perinatal death‡ | 0.01 (0.00–0.02) | Beta | APOSTEL-II11 |
| Severe ANO§ with antenatal corticosteroids* | 0.29 (0.22–0.38) | Beta | APOSTEL-II11 |
| Severe ANO§ with antenatal corticosteroids† | 0.19 (0.14–0.26) | Beta | APOSTEL-II11 |
| Severe ANO‡§ | 0.01 (0.00–0.02) | Beta | APOSTEL-II11 |
| Relative risk | |||
| Perinatal death if antenatal corticosteroids | 0.77 (0.67–0.89) | Log-normal | Roberts12 |
| Severe ANO*§ | 0.59 (0.41–0.88) | Log-normal | Roberts12 |
- * Preterm delivery (PTD) within 7 days after testing.
- † PTD > 7 days after testing but < 34 weeks' gestation.
- ‡ PTD > 34 weeks.
- § Composite of perinatal death, chronic lung disease, neonatal sepsis, intraventricular hemorrhage > Grade II, periventricular leukomalacia > Grade I and necrotizing enterocolitis.
- ANO, adverse neonatal outcome.
All probabilities were conditional, i.e. the probability of a test being positive under the condition that a woman will deliver within 7 days (also known as the sensitivity). In the strategies combining fFN testing and CL measurement, if it was not possible to perform either test, or in case of an invalid test result, it was assumed that the treatment decision was based on the results of other tests. Health outcomes in non-treated women were estimated based on relative risk ratios available from a meta-analysis on adverse neonatal outcome when women were treated with tocolysis and corticosteroids12.
The quantities of the various resources used, as observed in the APOSTEL-I and APOSTEL-II studies, were multiplied by the respective unit costs, resulting in overall costs per procedure or admission9-11. Direct medical costs due to CL measurement and fFN testing, maternal and neonatal admission, treatment and medical transport were considered, as were direct non-medical costs (owing to personal transport) and indirect costs (productivity losses of the parents).
Different levels of intensity of care were taken into account, i.e. admission to medium-, high-, or intensive-care wards. We took into account all in-utero transfers between hospitals due to a shortage of beds in neonatal intensive care. Additional neonatal care included radiodiagnostic procedures, use of expensive medication such as surfactant therapy and number of days of continuous positive airway pressure and intubation, because neonates require more intensive care, materials and drugs in this period.
Unit-cost estimates for maternal and neonatal hospitalizations were derived from the financial departments of one academic and one general hospital in The Netherlands (top-down cost estimation). Costs of medical transport were derived from Dutch guideline prices, costs of medication from the Dutch pharmacotherapeutic register and fFN costs, which included equipment and maintenance costs, from the Hologic price list (Table S1)13-15. Costs were expressed in 2011 €, and were inflated where appropriate using consumer price indexing16. Annual discounting was unnecessary because of the 1-year time horizon.
Cost ranges used for sensitivity analyses were based on the 95% CIs observed in the APOSTEL-I and APOSTEL-II studies. During multivariate sensitivity analysis, beta distributions were assumed for the probability parameters (prevalence, test accuracy, health outcomes) and gamma distributions for the cost parameters17.
Statistical analysis
First, dominated strategies, which are less effective and more costly than other strategies, were identified. We used one-way sensitivity analysis, in which we varied the estimates for the variables over the plausible ranges, to check if dominance was consistent. If dominance changed, it was considered to be non-consistent and reported17.
Then, each dominant strategy was compared with the next less effective strategy in terms of cost savings and neonatal health loss, starting with the most costly reference strategy (the treat-all strategy). The data generated were used to select the optimal strategy, in which we weighted the magnitude of cost savings against the amount of neonatal health loss. This is expressed as an incremental cost-effectiveness ratio (ICER), which reflects the monetary gain by accepting the loss of one unit of health outcome.
Probabilistic Monte Carlo simulation was performed to determine the joint effect of uncertainty for all model parameters on the estimates of costs and effectiveness. Based on the Monte Carlo simulations we calculated 95% plausible intervals of health and cost differences between the strategies17.
The impact of implementing the optimal strategy in The Netherlands was assessed in a budget impact analysis. First we compared the preferred strategy with current practice in The Netherlands, where 180 000 deliveries and 25 000 women with symptoms of preterm labor are recorded annually, of whom approximately 5000 present before 34 weeks' gestation. Although practice variation exists, the proportion of women in whom CL is measured in order to make treatment decisions is estimated at between 70 and 100% in The Netherlands. The decision tree was created and analyzed with TreeAge Pro 2009 Suite software (TreeAge Software, Inc., Williamstown, MA, USA). The study protocol was approved by the ethical committee of the Amsterdam Medical Center (MEC 08/363).
RESULTS
In the base–case analysis, treating all patients without testing (Strategy 1) led to adverse neonatal outcome in 9% of cases, while no treatment at all (Strategy 7) led to adverse neonatal outcome in 12% of cases. Strategy 1 and Strategy 7 were also the most and least costly strategies, with average costs of €30 187 and €24 952 per pregnancy, respectively. Strategies using either fFN testing or CL measurement as sole tests for risk stratification (Strategies 2 and 3) were dominated because Strategy 6 (fFN test only if CL is 15–30 mm) was more effective at lower cost. The remaining strategies were combinations of fFN testing and CL measurement (Figure 2). Costs in the more expensive strategies were mainly driven by unnecessary admissions and treatment. One-way sensitivity analysis, in which we varied each of the variables within their predefined ranges, confirmed that the strategies including both CL and fFN measurement remained dominant over fFN testing alone and CL measurement alone.
, treat all (Strategy 1); 
, treat if cervical length (CL) < 25 mm (Strategy 2); 
, treat if fFN test positive (Strategy 3); 
, treat if CL < 25 mm and fFN test positive (Strategy 4); 
, treat if CL < 25 mm or fFN test positive (Strategy 5); 
, treat if CL < 15 mm or CL 15–30 mm and fFN test positive (Strategy 6); 
, no treatment (Strategy 7).Comparison of the second most expensive strategy (Strategy 5, test fFN and CL and treat if either test is positive) with the most expensive one (Strategy 1, treat all) showed an average cost reduction per patient presenting with signs of preterm labor of €2374 (95% CI, €1248–3575). This cost reduction was not accompanied by an increase in perinatal mortality (0.03 (95% CI, –0.14 to 0.34) per 1000 women), but it did lead to an increased incidence of adverse neonatal outcome (0.29 (95% CI, 0.01–1.49) per 1000 women), resulting in an ICER of €8.3 million saved for accepting every one additional case of adverse neonatal outcome (Table 2). Comparison of Strategy 6, which was the second least expensive strategy, with Strategy 5 revealed an additional cost reduction per patient of €1545 (95% CI, €765–2398). This cost reduction was accompanied by an increase in perinatal mortality (0.12 (95% CI, –0.04 to 0.54) per 1000 women) and a higher incidence of adverse neonatal outcome (1.8 (95% CI, 0.31–5.0) per 1000 women), resulting in ICERs of €15 million and €873 396, respectively. Comparing the least expensive alternative (Strategy 7, no treatment) with Strategy 6, showed an additional cost reduction per patient of €1316 (95% CI, –€17 to €2271), with an increase in perinatal mortality (1.8 (95% CI, 0.51–4.8) per 1000 women) and a higher incidence of adverse neonatal outcome (26.5 (95% CI, 18.4–36.3) per 1000 women), resulting in ICERs of €730 912 and €49 689, respectively (Table 2).
| Neonatal outcome per 1000 women (n) | |||||||
|---|---|---|---|---|---|---|---|
| Strategy | Women treated (%) | Cost per woman (€) | Cost saving per woman (95% CI) (2011 €) | Perinatal mortality | Additional perinatal mortality (95% CI) | Poor outcome* | Additional poor outcome (95% CI) |
| 1: Treat all | 100 | 30 187 | N/A | 16.9 | N/A | 91.8 | N/A |
| 5: Treat if fFN positive or CL < 25 mm | 65 | 27 813 | 2374 (1248 to 3575)† | 16.9 | 0.03 (–0.14 to 0.34)¶ | 92.1 | 0.29 (0.01–1.49)¶ |
| 2: Treat if CL < 25 mm | 62 | 26 748 | Dominated | 17.0 | Dominated | 94.0 | Dominated |
| 3: Treat if fFN positive | 47 | 26 582 | Dominated | 17.0 | Dominated | 94.8 | Dominated |
| 6: Treat if CL < 15 mm or if CL 15–30 mm and fFN positive | 41 | 26 268 | 1545 (765 to 2398)‡ | 17.0 | 0.12 (–0.04 to 0.54)** | 93.9 | 1.8 (0.31–5.0)** |
| 4: Treat if fFN positive and CL < 25 mm | 37 | 26 295 | Dominated | 17.1 | Dominated | 95.8 | Dominated |
| 7: No treatment | 0 | 24 952 | 1316 (–17 to 2271)§ | 18.8 | 1.8 (0.51 to 4.8)†† | 120.3 | 26.5 (18.4–36.3)†† |
- Each strategy compared with its more expensive alternative to calculate cost savings and additional neonatal adverse outcomes.
- * Composite of perinatal death, chronic lung disease, neonatal sepsis, intraventricular hemorrhage > Grade II, periventricular leukomalacia > Grade I and necrotizing enterocolitis.
- † Cost saving for Strategy 5 compared with treat-all strategy (Strategy 1).
- ‡ Extra cost saving compared with Strategy 5.
- § Extra cost saving compared with Strategy 6.
- ¶ Additional compared with Strategy 1.
- ** Additional compared with Strategy 5.
- †† Additional compared with Strategy 6.
From this analysis, Strategy 6 was considered to be a viable strategy because it would result in major cost savings with a minor deterioration in neonatal health. We also compared the cost of additional fFN testing when CL measurement was between 15 and 30 mm (Strategy 6) with the treat-all strategy to estimate the overall cost savings, which amounted to €3919 per woman, with a small deterioration in neonatal health outcomes, namely one additional perinatal death and 21 cases of adverse outcome per 10 000 women with signs of preterm labor (ICERs of €39 million and €1.9 million, respectively).
Budget impact analyses comparing the optimal strategy, in terms of cost-effectiveness (Strategy 6) with the current care for women presenting with symptoms of preterm labor in The Netherlands, showed that potential savings amount to €2.4 million annually if we assumed that in current practice CL measurement is performed in all women and fFN testing is added only as appropriate. If CL measurement were to be implemented in 70% of these women, the estimated cost reduction would be €7.6 million (Table 3). The differences found in neonatal health outcomes between current practice in The Netherlands and Strategy 6 were small, involving 0.2 additional cases of neonatal mortality and three additional cases of neonatal morbidity per year per 1000 women with Strategy 6.
| fFN test when CL between 15 and 30 mm | ||||||
|---|---|---|---|---|---|---|
| Current practice: women in whom CL measurement with decision threshold of 25 mm is used (%)* | Cost saving per woman with signs of preterm labor (2011 €) | Additional neonatal mortality per 1000 women | Additional poor neonatal outcome per 1000 women† | Annual cost savings in Netherlands (2011 €)‡ | Additional neonatal mortality per year in Netherlands‡ | Additional poor neonatal outcome per year in Netherlands†‡ |
| 100 | 480 (88 to 993) | –0.01 (–0.2 to 0.3) | –0.1 (–2.4 to 2.2) | 2.4 million | –0.002 (–1.0 to 1.4) | –0.5 (–12.0 to 11.0) |
| 90 | 824 (283 to 1473) | 0.01 (–0.2 to 0.3) | 0.2 (–2.1 to 2.4) | 4.1 million | 0.07 (–0.9 to 1.6) | 1.0 (–10.5 to 12.0) |
| 80 | 1168 (479 to 1952) | 0.02 (–0.2 to 0.4) | 0.4 (–1.9 to 2.6) | 5.8 million | 0.14 (–0.8 to 1.8) | 2.0 (–9.5 to 13.0) |
| 70 | 1512 (674 to 2432) | 0.04 (–0.1 to 0.4) | 0.6 (–1.6 to 2.8) | 7.6 million | 0.21 (–0.8 to 2.0) | 3.0 (–8.0 to 14.0) |
| 60 | 1855 (869 to 2911) | 0.05 (–0.1 to 0.5) | 0.8 (–1.3 to 3.0) | 9.3 million | 0.29 (–0.7 to 2.2) | 4.0 (–6.5 to 15.0) |
| 50 | 2199 (1065 to 3391) | 0.07 (–0.1 to 0.5) | 1.0 (–1.1 to 3.2) | 11.0 million | 0.36 (–0.6 to 2.4) | 5.0 (–5.5 to 16.0) |
| 40 | 2543 (1260 to 3871) | 0.08 (–0.1 to 0.5) | 1.2 (–0.8 to 3.4) | 12.7 million | 0.43 (–0.6 to 2.6) | 6.0 (–4.0 to 17.0) |
| 30 | 2887 (1455 to 4350) | 0.10 (–0.1 to 0.6) | 1.4 (–0.5 to 3.6) | 14.4 million | 0.50 (–0.5 to 2.8.) | 7.0 (–2.5 to 18.0) |
- Values in parentheses are 95% CI.
- * Results vary as subject of the percentage of women in whom CL measurement with a decision threshold of 25 mm is used in current practice.
- † Poor neonatal outcome defined as a composite of perinatal death, chronic lung disease, neonatal sepsis, intraventricular hemorrhage > Grade II, periventricular leukomalacia > Grade I and necrotizing enterocolitis.
- ‡ Based on 180 000 deliveries a year and 25 000 women with symptoms of preterm labor annually, of whom approximately 5000 present before 34 weeks' gestation.
DISCUSSION
This study evaluated the costs and health outcomes of seven test–treatment strategies in women with symptoms of preterm labor between 24 and 34 weeks' gestation. We found that strategies based on fFN testing or CL measurement as separate tests are associated with higher costs and incidence of adverse neonatal outcomes compared with strategies that combine both tests. A strategy based on CL measurement, treating women with a CL of < 15 mm, recommending additional fFN testing in women with a CL between 15 and 30 mm and also treating women with a positive fFN test, could reduce unnecessary transfers and treatment, leading to considerable cost savings, without worsening of neonatal outcomes compared with current practice. The potential cost savings in the antenatal period, as a result of using this strategy, were estimated at between €2.4 and €7.6 million a year in The Netherlands, a country with about 180 000 deliveries annually.
All model-based analyses simplify considerably a more complex reality, and ours is no exception. Side-effects of treatment that may lead to unnecessary adverse outcomes for the mother and child if preterm delivery does not occur, such as side-effects from (repeated) corticosteroids injections, tocolysis and hospital admissions, were not taken into account4-6. Such effects could offset the small beneficial effect of treating all patients with corticosteroids. Furthermore, this model did not account for a strategy in which women are triaged with serial digital examinations and only treated if a threshold cervical change or dilation occurs; neither did it consider strategies in which repeat CL measurements or fFN tests are performed during an episode of threatened preterm labor. Data to factor in hospital admissions in the setting of negative test results were unavailable. Although we are convinced that women with negative test results could be discharged with appropriate instructions, it is possible that clinicians might decide to admit these women. Such decisions may decrease the potential cost savings and eventually compromise the cost-effectiveness of the fFN test.
We acknowledge that only short-term health outcomes and costs were evaluated in our analysis, while neonatal morbidity often implies future health problems, such as neurological impairment and pulmonary problems18. The long-term economic impact of neonatal morbidity can be substantial, both from a healthcare system perspective and from societal and human capital perspectives, and could counteract the short-term cost savings. However, the differences found in neonatal health outcomes between the treat-all strategy and our preferred strategy were small. As seen in the literature, most healthcare costs due to morbidity occur in the period immediately after delivery, and these costs were limited in our analysis19. Our composite neonatal health outcome includes less severe morbidity as well.
To guide clinical decision-making, willingness-to-accept thresholds for poor neonatal outcome and perinatal mortality would be informative. However, these thresholds do not exist. A solution would be to translate these outcomes into (future) quality-adjusted life years (QALYs), as willingness-to-accept thresholds are available for QALYs. Unfortunately, introducing QALYs is accompanied by some difficulties. Firstly, data on future QALYs and costs after (extreme) premature birth are lacking as far as we know; thus, introducing QALYs in the model based on assumptions would introduce a great amount of uncertainty. Secondly, future QALYs and costs incurred by the mother should be included as well, which makes the model too complex. For these reasons we presented only ICERs for poor neonatal outcome and perinatal mortality.
Previous randomized trials on the clinical utility of fFN in women with symptoms of preterm labor have failed to demonstrate benefits. It is important to realize that the use of fFN and CL measurement in symptomatic women will identify women who do not need treatment with tocolytics. As such, these tests will not prevent preterm birth and thus will not improve neonatal outcome directly20. A long-term benefit could be achieved if women at low risk for preterm birth were denied antenatal corticosteroids, as this might be correlated with poorer school performance21. Thus, the most important reason to triage women with threatened preterm labor is to prevent unnecessary referral, treatment and costs.
Compared with our previous cost-effectiveness analysis8, the calculated costs per patient were higher in this study because of the incidence of preterm delivery within 7 days, which was 12% instead of 8%, different cost estimates and the use of a societal perspective instead of a hospital perspective. The potential cost savings of implementing our preferred strategy, fFN testing when CL is between 15 and 30 mm, according to the present analysis, are somewhat more conservative than those of the previous economic evaluation, because initial maternal admission and treatment were less costly10.
Nowadays, a quantitative fFN test, which shows advantages over the qualitative fFN, is available in Europe. Women with a short cervix and a fFN concentration of < 10 ng/mL appeared to have a low risk of delivering preterm, while women with a long cervix (> 30 mm) and a fFN concentration > 500 ng/mL appeared to be at risk22. Introducing this test could possibly result in a more accurate prediction of preterm delivery in the short term and an improvement in cost savings.
The model described in this study is based on the Dutch healthcare system, which has an important role for midwives during pregnancy and (uncomplicated) delivery. It might be considered that midwives preselect women with symptoms of preterm labor before they are referred to the gynecologist, resulting in a higher a-priori risk for preterm delivery than in other countries, which leads to better positive predictive values of the test strategies. However, in previous studies we found that the a-priori risk was the same in women presenting themselves with and without any intervention from a midwife10. The optimal strategy, in terms of cost-effectiveness, would identify accurately low-risk women, regardless of their referral path. Cost estimates depend on the organization of a healthcare system. Before conclusions can be generalized to other countries, these findings need to be confirmed in prospective studies, among health systems with varying referral policies.
Combining fFN testing with CL measurement could lead to considerable cost savings in the short term in The Netherlands, without compromising neonatal health outcomes in women with symptoms of preterm labor between 24 and 34 weeks' gestation, if a protocol were used in which treatment decisions were strictly guided by test results.
ACKNOWLEDGMENT
This investigator-initiated study was funded by The Netherlands Organization for Health Research and Development (ZonMw, grant number 80-82310-98-09056).
