Vaginal birth after caesarean versus elective repeat caesarean delivery after one previous caesarean section: a cost-effectiveness analysis in four European countries

A decision analytic model was developed to compare the costs and consequences of VBAC and ERCD. The decision tree (Fig. 1) was built based on previously published models and expert opinion (GF) [7, 9]. A hypothetical cohort of 100,000 women was used for each country (Belgium, Germany, Ireland and Italy). Women included in the model were low-risk (i.e. without pre-existing medical conditions or risk factors), with a singleton pregnancy, at term gestation who had one previous caesarean section that was performed using low uterine transverse incision. The assumption is that in this cohort of women, there are no contraindications to VBAC [26] at time of admission for giving birth. The time horizon of the model was from pregnancy recognition up to 6 weeks postpartum for short-term consequences and a lifetime horizon was applied for long-term consequences. The decision tree consisted of two main arms, VBAC and ERCD. Women in the VBAC arm may have a successful VBAC resulting in an unassisted vaginal birth or instrumental vaginal birth. Women in this arm may also experience an emergency caesarean section or uterine rupture. In the ERCD arm, women may have an ERCD, and a small proportion of women might go into labour spontaneously resulting in an unassisted vaginal birth or instrumental vaginal birth. Women planning an ERCD may also suffer uterine rupture but, as it occurs only rarely, we have not included it in the model. In all cases, women could have no morbidity, suffer a morbidity, or die. To account for maternal morbidity, the maternal outcomes included were: uterine rupture, endometritis, peripartum hysterectomy, blood transfusion, thrombotic events, operative injury and wound complications [11, 15‐21, 27‐30]. Neonatal health outcomes were not built directly into the decision tree but accounted for via the health state of the women (see below) [11]. A neonate could be born alive and have no morbidity, suffer a morbidity or die. In cases of neonatal morbidity, the following neonatal outcomes were evaluated: hypoxic ischemic encephalopathy (HIE), sepsis and respiratory conditions including transient tachypnoea of the newborn (TTN) and respiratory distress syndrome (RDS) [5, 6, 11, 17, 20, 31‐38]. Cerebral palsy was included in the model as a long-term consequence of HIE, assuming that 12% of the infants would finally be diagnosed with cerebral palsy [11, 39]. Country-specific data on the VBAC success rates were available for Germany (74%), Ireland (66%) and Italy (64%), and the overall European success rate was used for Belgium (73%) [9, 40‐42]. All other probabilities, including maternal and neonatal morbidities and mortality, were based on data obtained from the literature, and were identical for all countries (Additional file 1: Table S2) [5, 6, 9, 11, 15‐21, 27‐38, 43, 44]. Based on the probability data, the probability per arm was calculated by multiplying the probability of each branch to calculate the final probability of the endnodes e.g. for unassisted vaginal birth: probability of unassisted vaginal birth (VBAC success rate * proportion of unassisted vaginal birth * probability to be alive * probability to have no morbidities) (Fig. 1).

The analysis was performed from a societal perspective and included maternal and neonatal health outcomes and costs. The costs included direct medical costs for the ante-, peri- and postnatal periods and non-medical costs and indirect costs. An overview of all costs and their respective sources can be found in Table 1 and in Additional file 1: Tables S1 and S2. For each country, antenatal resource usage calculations were based on the national antenatal/neonatal guidelines and were validated by an expert in each country. One extra consultation with an anaesthesiologist was included when an ERCD was planned. Ambulatory costs, i.e. costs for outpatient care, were obtained from country specific ambulatory tariff lists and linked to resource use. Inpatient costs of the mother and newborn were calculated per health-state based on country-specific Diagnostic Related Groups (DRG) data. Two extra visits with an obstetrician and paediatrician were included in cases of short-term maternal and/or neonatal morbidities, and direct non-medical costs (i.e. travel costs) were also included. The average distance to ambulatory centres and hospitals was multiplied by the unit cost for transportation. The cost of HIE was estimated as half of the yearly cost of CP, for 2 years [11]. For long-term consequences such as cerebral palsy and maternal and neonatal mortality, lifetime costs were calculated including the lifetime societal cost for cerebral palsy for 50 years [13] and lifetime productivity loss for maternal and neonatal mortality. National discount rates for costs and utilities were used: Belgium (3%, 1.5%), Germany (5%, 5%), Ireland (5%, 5%) and Italy (3%, 3%) [45‐48]. The Harmonized Indices of Consumer Prices (HICP) - health were used to convert all costs to 2016 euros [49].

Quality-adjusted life years (QALYs) were used to measure outcomes for the different health states (Table 2 and Additional file 1: Table S2). The country-specific utilities for normal health were considered as the base-case utility [50, 51]. Ireland did not have country-specific values available, and UK utility values were used. Utility values during the antenatal period were assumed to be equal for both groups, VBAC and ERCD. Thereafter, utilities were calculated for each maternal and neonatal outcome separately by reducing the country-specific utilities for normal health with the corresponding disutility of the maternal and neonatal outcome. No country-specific disutilities could be obtained, consequently, the disutility values were based on data obtained from the literature [9, 11, 12, 52].

The primary outcome was the incremental cost-effectiveness ratio (ICER) per QALY gained using the expected costs and QALYs gained by using the following formula:

The cost-effectiveness threshold was country dependent. For Belgium, Germany and Italy, the gross domestic product (GDP) per capita was used, with an ICER of €36,633/QALY, €37,719/QALY and €27,219/QALY respectively [53]. The cost-effectiveness threshold for Ireland was historically set at €45,000/QALY [9]. Additionally, the incremental net monetary benefit (INMB) between VBAC versus ERCD was calculated applying various thresholds in order to evaluate the cost-effectiveness (Additional file 1: Figure S2).

To identify the uncertainty of parameters and the robustness of the results of the base-case model, sensitivity analyses were performed. Both one-way and probabilistic sensitivity analyses were conducted on the input parameters for all costs, probabilities and utilities used in the model. In the one-way sensitivity analysis, input parameters were varied by 30% of their initial value, and parameters with the highest impact are presented in Tornado diagrams. Simulations were conducted by using beta and Dirichlet distributions for probabilities and utilities, and gamma distributions for costs. A probabilistic sensitivity analysis was performed using Monte Carlo simulations with 5000 iterations.

The base-case analysis comparing VBAC with ERCD for this hypothetical cohort resulted in an average of 32,312 fewer caesarean sections per 100,000 women in each of the four countries (Additional file 1: Table S3). The ICER results of the deterministic analysis for a 6-week time horizon demonstrated that VBAC was a dominant strategy in all four countries, which means less costly and more effective than ERCD. This results in a negative ICER of -€441/QALY for Germany, −€2035/QALY for Belgium, −€3042/QALY for Italy and -€9891/QALY for Ireland (Table 3). When considering a lifetime time-horizon, VBAC was a dominant strategy in Ireland (−€96,753/QALY) and Italy (−€15,965/QALY). Compared to ERCD, VBAC remained a cost-effective strategy, which means slightly more costly and more effective, for Belgium (€3669/QALY) and Germany (€12,817/QALY). The overall maternal morbidity rate was higher in the VBAC arm compared to the ERCD arm (5.64% versus 3.25%) due to the morbidity related to performing emergency caesarean sections. The neonatal morbidity rate, 6.82% for VBAC and 5.14% for ERCD, and mortality rate, 0.02% for VBAC and 0.00% for ERCD, were higher in the VBAC arm due to the higher amount of complications in case of an emergency caesarean section or uterine rupture. The overall maternal mortality rate was comparable for both groups (0.01% and 0.00% in both groups).

The costs and QALYs saved differ between the four countries. Within a 6-week time horizon, the total costs saved (per 100,000 women) were €3,334,052 in Germany, €15,271,888 in Belgium, €19,465,446 in Italy and €66,162,379 in Ireland and the total number of QALYs were 6399 in Italy, 6689 in Ireland, 7503 in Belgium and 7561 in Germany more per 100,000 women. When considering lifetime costs and consequences, the total cost of the VBAC strategy was €1,386,210 in Belgium and €8,481,119 in Germany with 376 QALYs and 662 QALYs more in Belgium and Germany. The total costs saved were €53,955,166 in Ireland and €6,558,036 in Italy and the total number QALYs were 558 in Ireland and 411 in Italy.

The results of the one-way sensitivity analysis showed the robustness of the models when varying the input parameters. Overall, the model was robust, i.e. varying parameters by the sensitivity analysis did not result in large variations of the ICERs. Four of the most influential drivers on the ICERs were similar for all four countries: the costs of antenatal care in both ERCD and VBAC, cost of an ERCD and unassisted vaginal birth without complications. For Belgium, Ireland and Italy, the (dis)utility associated with emergency caesarean birth was found to be sensitive variables for the ICER result. The probability of an unassisted vaginal birth was sensitive for Belgium, Germany and Italy. For the two latter countries, the disutility of an ERCD was also a sensitive variable. For Germany, the model was also sensitive for the probability of an ERCD and for the disutility of unassisted vaginal birth in Ireland (Additional file 1: Figure S2). The probabilistic sensitivity analysis resulted in a mainly dominant ICER with a reduction of the costs and increased effects when comparing VBAC with ERCD. The results differed between countries, ranging from -€423/QALY in Germany to -€9886/QALY in Ireland for a 6 week time horizon (Table 3), and from €8609/QALY in Germany to -€60,905 in Ireland for a lifetime time horizon. The most favourable ICER was for Ireland which can be explained by the high incremental cost of an ERCD compared to a vaginal birth. The lowest ICER was for Germany, as a result of the high probability and cost of neonatal sepsis in case of an emergency caesarean section or uterine rupture. The Monte Carlo simulations (6-week time horizon) are graphically presented as cost-effectiveness planes (Fig. 2), and are mainly plotted in the south-east quadrant for all four countries. For Belgium and Germany 7.34% and 34.50% of all simulations were plotted in the north-east quadrant, with the latter showing VBAC to be slightly more expensive compared to ERCD. All simulations remained far below the national cost-effectiveness thresholds [53] and the likelihood of cost-effectiveness at the national thresholds were 100% for all four countries. For the lifetime model, VBAC was the preferred strategy for 98.66% (Belgium) and 100.00% (Germany, Ireland and Italy) of all Monte-Carlo simulations.