Dr S Petrou, National Perinatal Epidemiology Unit, University of Oxford (Old Road Campus), Old Road, Headington, Oxford OX3 7LF, UK. Email email@example.com
Objectives To compare the cost-effectiveness of alternative management methods of first-trimester miscarriage.
Design Economic evaluation conducted alongside a large randomised controlled trial (the MIST trial).
Setting Early pregnancy assessment units of seven participating hospitals in southern England.
Sample A total of 1200 women with a confirmed pregnancy of less than 13 weeks of gestation with a diagnosis of incomplete miscarriage or missed miscarriage.
Methods Random allocation to expectant management, medical management or surgical management. Collection of health service and broader resource use data, unit costs for each resource item and clinical outcomes.
Main outcome measures Costs (£, 2001–02 prices) to the health service, social services, women, carers and wider society during the first 8 weeks postrandomisation. Cost-effectiveness estimates, expressed in terms of incremental cost per gynaecological infection prevented; cost-effectiveness acceptability curves presented at alternative willingness-to-pay thresholds for preventing gynaecological infection.
Results There was no significant difference in the incidence of gynaecological infection between groups. The net societal cost per woman was estimated at £1086.20 in the expectant group, £1410.40 in the medical group and £1585.30 in the surgical group. Expectant management had a 97.8% probability of being the most cost-effective management method at a willingness-to-pay threshold of £10,000 for preventing one gynaecological infection, while medical management had a 2.2% probability of being the most cost-effective management method. Expectant management retained the highest probability of being the most cost-effective management method at all willingness-to-pay thresholds of less than £70,000 for preventing one gynaecological infection.
Conclusions Expectant and medical management of first-trimester miscarriage possess significant economic advantages over traditional surgical management.
Approximately one in seven confirmed early pregnancies end in miscarriage during the first trimester.1 The optimal strategy for managing the two types of first-trimester miscarriage, incomplete miscarriage (expulsion of products of conception has begun) and missed miscarriage or early fetal demise (visible fetus but absent heart activity), remains uncertain.2,3 The traditional approach of surgical evacuation developed during the first half of the 20th century as a result of high rates of gynaecological infection from the retained products of conception and the ensuing mortality from septicaemia.4 More recently, expectant management has been acknowledged by many GPs as a means of facilitating a well-regulated natural process in human reproduction, while medical management has translated the treatment regimens widely used for medical terminations of pregnancy to the management of spontaneous miscarriage.4
A number of randomised controlled trials comparing expectant management with surgical management,5–7 medical management with surgical management8–12 and expectant management with medical management13 have been conducted. However, to date, no randomised controlled trial has compared all three management methods. Moreover, the randomised controlled trials that have been conducted lack information on resource use, thus preventing assessments of cost-effectiveness from being made. The economic evaluation reported in this study provides, for the first time, both clinical information on the three alternative management methods and information on the use of resources associated with treatment and management of complications, thereby allowing the cost-effectiveness of expectant, medical and surgical management of first-trimester miscarriage to be assessed. The economic evaluation is based upon the MIScarriage Treatment (MIST) trial, the clinical results of which are reported elsewhere.14
A pragmatic randomised controlled trial was conducted to evaluate the benefits and risks associated with expectant and medical management, in comparison with surgical management, of first-trimester miscarriage. Women with a confirmed pregnancy of less than 13 weeks of gestation, who had been diagnosed through transvaginal scan with either an incomplete miscarriage or missed miscarriage/early fetal demise, were recruited from one of seven early pregnancy assessment units in southern England between May 1997 and December 2001. Women who consented to participate in the trial were randomised by telephone to expectant, medical or surgical management. Minimisation was used to ensure comparability between women allocated to the three management groups with respect to four prognostic variables: participating centre; parity (nulliparous or parous); miscarriage type (incomplete miscarriage or missed miscarriage/early fetal demise) and gestation (<56 days, 56–76 days, 77 or more days or unknown). Women allocated to the expectant group were allowed to go home with no intervention. The management of women allocated to the medical group depended on the type of miscarriage. Those with incomplete miscarriages were admitted to hospital and given a single vaginal dose of 800 μg of misoprostol, while those with missed miscarriages were pretreated with a single oral dose of 200 mg of mifepristone and then admitted 24–48 hours later for a single vaginal dose of 800 μg of misoprostol. Women allocated to surgical management were admitted for surgical evacuation of the retained products of conception (ERPC) in line with the usual policy of each participating centre. Documented gynaecological infection constituted the primary clinical outcome of the trial. This was defined as two or more of the following: purulent vaginal discharge; pyrexia >38.0°C; tenderness over the uterus on abdominal examination and/or increase in white cell count above 15 × 109/ml. The documentation of gynaecological infection covered the first 14 days and the first 8 weeks following trial entry. All analyses and comparisons were performed on the basis of intention to treat. Ethical approval for the study was obtained from the relevant research ethics committees. Further details of the design and conduct of the trial are reported elsewhere.14
Type of economic evaluation, perspective and time horizon
An incremental cost-effectiveness analysis was performed in which we calculated the incremental costs and incremental effectiveness of expectant and medical management, in comparison with surgical management, and expectant management in comparison with medical management and expressed these as incremental cost-effectiveness ratios (ICERs). The economic evaluation was conducted from a societal perspective and covered the cost of hospital and community health and social services, the costs borne by women themselves and their informal carers, as well as the costs of lost production. The time horizon for the economic evaluation mirrored the time horizon for the randomised controlled trial, namely the period between randomisation and 8 weeks postrandomisation.
Resource use data
Data were collected about all significant resource inputs during the study period. For each woman in the study, the trial data collection forms recorded information on the duration of the hospital admission for medical or surgical management, as well as the number and duration of emergency hospital consultations and emergency hospital admissions during the study period. For each hospital contact, the trial data collection forms recorded the type and quantity of analgesia, antiemetics and antibiotics administered; the volume of intravenous infusion; the number of units of blood transfused; the type (general, regional, local) and duration of anaesthesia performed; the duration of surgery performed; as well as a profile of anaesthetic and surgical complications. Observational research in the form of staff tick charts, which medical and nursing staff at the clinical centres were asked to complete following each contact with the women, provided a record of all hospital staff inputs during the study period. These tick charts recorded the type (medical student, healthcare assistant, nurse, medical doctor, ultrasonographer, other) and grade of the health professional caring for each woman, the number of contacts with each woman and the duration of each contact.
Postal questionnaires completed by the women approximately 10–14 days and 8 weeks following trial entry recorded the number, type and duration of community health and social care contacts made by the women. These questionnaires also recorded the direct nonmedical resource inputs attributable to the healthcare process, for example, child care support of siblings and distances travelled to healthcare providers. In addition, the questionnaires recorded the time that women had to take off work as a consequence of their miscarriage and any additional items that they had to purchase, for example, additional sanitary towels.
Valuation of resource use
Unit costs for resources used by the women who participated in the trial were obtained from a variety of sources. All unit costs employed followed recent guidelines on costing health and social care services as part of economic evaluation.15,16 Unit costs of hospital-based resources were calculated by sending to each clinical centre a detailed questionnaire, requesting cost data for the main resource categories of drugs, disposables, consumables, surgical procedures, equipment, staff and overheads and then apportioning these to different categories of patient using a ‘top–down’ methodology.17 Each clinical centre was visited to ensure consistency in the apportionment and reporting of cost data. The unit costs of ultrasound scans18 and blood transfusions19 were derived from published sources. The unit costs of community health and social services were largely derived from national sources,20 and took account of time spent by professionals on indirect activities, such as travelling and paper work. However, some unit costs of community health and social services were calculated from first principles using established accounting methods.17 Work absences were valued using gender-specific median salaries provided by the New Earnings Survey.21 All other direct nonmedical resource inputs were valued using estimates of costs provided by the women themselves in the 10- to 14-day and 8-week questionnaires. Unit costs were combined with resource volumes to obtain a net cost per woman over the study period. All costs are expressed in pounds sterling and valued at 2001–02 prices.
Representation of cost-effectiveness
The cost-effectiveness of expectant and medical management, in comparison with surgical management, and expectant management in comparison with medical management, was expressed in terms of incremental cost per gynaecological infection avoided. The probabilities that each of the management methods is optimal in terms of cost-effectiveness, at different values for decision makers’ willingness to pay for preventing one gynaecological infection, are represented by cost-effectiveness acceptability curves.22 Cost-effectiveness acceptability curves have been promoted as the appropriate approach for representing cost-effectiveness, even in the case of an observed lack of statistical significance in either the cost or effect differences between healthcare interventions.23 They therefore represent the correct analytical focus for the economic evaluation conducted alongside the MIST trial. For the purposes of our analysis, we have indicated the probabilities that each of the management methods is optimal in terms of cost-effectiveness at decision makers’ willingness-to-pay thresholds of between zero and £100,000 for each gynaecological infection avoided.
The statistical approach developed by Lin et al.24 was used to simulate costs for women for whom one of the economic questionnaires was not completed and whose information could therefore be described as censored. Discounting of costs and health effects was unnecessary because the time horizon for the economic evaluation was restricted to the first 8 weeks following trial entry.
All results are reported as mean values with SD and as mean differences in costs and effects with 95% confidence intervals (CI) where applicable. We tested for differences in resource use and costs between the comparator groups using the independent-samples t test procedure. The differences in resource use, costs and effects between the comparator groups were considered significant if two-tailed P values were 0.05 or less. As the data for costs were skewed, we used nonparametric bootstrap estimation to derive 95% CI for mean cost differences between the comparator groups.25 Each of these CIs was calculated using 1000 bias-corrected bootstrap replications. Nonparametric bootstrap simulation of the cost–effect pairs was also used to generate 1000 replications of each of the ICERs, which were represented graphically on four-quadrant cost-effectiveness planes.26 All analyses were performed with a microcomputer running Excel (Microsoft, Redmond, WA, USA) and Statistical Package for the Social Sciences software (SPSS, Chicago, IL, USA).
We undertook sensitivity analyses to explore the implications of uncertainty surrounding the base-case ICERs.27 The values of the following variables were varied as part of the sensitivity analyses: (i) community service utilisation (increased by 10, 20 and 30% in response to a tendency, on the part of the participants in health economic studies, to under-report numbers of community service contacts28); (ii) the per diem costs for each level of gynaecological care (reduced and increased by 20% in order to reflect variations in the relative price structures of resource inputs across hospital settings29); (iii) the mean level of medical and nursing staffing support provided to each woman while in hospital (reduced and increased by 25% in order to reflect the variations in the level of staffing inputs that might be observed in routine practice) and (iv) the economic value of each day of work absence (set at the 25th and 75th centiles of the female national income distribution in order to capture the impact of the sociodemographic profile of the wider population on the study results).
Resource use and clinical effectiveness data were collected for 1200 women who were randomly allocated to either expectant (n= 399), medical (n= 398) or surgical (n= 403) management. The clinical results of the trial are presented in detail elsewhere.14 In brief, there were no significant differences between the comparator groups in the incidence of gynaecological infection within the first 14 days or within the first 8 weeks postrandomisation.14Table 1 shows the utilisation of health and social care services, as well as the broader societal resource impacts, by trial group. Women allocated to the medical and surgical groups spent, on average, a significantly greater number of days in hospital during the initial admission and throughout the study period than women allocated to the expectant group, while women allocated to the medical group spent a significantly greater number of days in hospital during both the periods than women allocated to the surgical group (P < 0.05). Women allocated to both the expectant and medical groups made a significantly greater number of emergency consultations and admissions than women allocated to the surgical group (P < 0.05). The duration of general anaesthesia and surgery was significantly greater for women allocated to the surgical group than for women allocated to both the expectant and medical groups (P < 0.05). There were no significant differences in the duration of staff contacts between the trial groups, with the exception of the duration of specialist registrar contacts, which was significantly greater for women allocated to the surgical group than for women allocated to the expectant and medical groups, and for the duration of nurse grade D and nurse grade E contacts, which was significantly greater for women allocated to the surgical group than for women allocated to the medical and expectant groups, respectively (P < 0.05). The mean number of doses of co-dydramol, pethidine and intravenous/intramuscular prochlorperazine and metoclopramide was significantly greater for women allocated to the medical group than for women allocated to other two trial groups (P < 0.05). Additionally, the mean number of milligrams of other analgesia and the mean number of doses of oral metoclopramide and other antiemetics was significantly greater for women allocated to the medical group than for women allocated to the expectant group (P < 0.05). When the utilisation of broader societal resources was considered, the only difference between the groups was the utilisation of other community health and social services, which was significantly greater for women allocated to the expectant group than for women allocated to the surgical group (P < 0.05).
Table 1. Resource use during the 8-week postrandomisation period
Resource use variable
Expectant, n= 398
Medical, n= 398
Surgical, n= 402
IM, intramuscular; IV, intravenous.
Values are given as mean (SD).
Initial admission (duration, days)
Emergency consultations (number)
Emergency admissions (number)
Emergency admissions (duration, days)
All hospital admissions (duration, days)
Duration of general anaesthesia (minutes)
Duration of surgery (minutes)
Healthcare assistant contacts (minutes)
Nurse grade D contacts (minutes)
Nurse grade E contacts (minutes)
Nurse grade F contacts (minutes)
Nurse grade G contacts (minutes)
Medical student contacts (minutes)
Senior house officer contacts (minutes)
Specialist registrar contacts (minutes)
Consultant contacts (minutes)
Ultrasonographer contacts (minutes)
Other hospital staff contacts (minutes)
Other analgesia (mg)
Oral metoclopramide (doses)
Oral ondansetron (doses)
IV/IM prochlorperazine (doses)
IV/IM metoclopramide (doses)
IV/IM ondansetron (doses)
Other antiemetics (doses)
IV infusion (crystalloid or colloid) (ml)
Blood given (units)
Broader societal impacts
GP contacts (number)
Midwife contacts (number)
Practice nurse contacts (number)
Health visitor contacts (number)
Social worker contacts (number)
Other community contacts (number)
Time off work (days)
Table 2 shows the unit costs for each resource item. These were combined with resource volumes to generate mean costs per woman according to category of cost and trial group (Table 3). Medical and surgical management increased the cost of the initial hospitalisation by an average of £642.7 (bootstrap mean difference of £642.8, 95% CI: £594.4–£691.5; P < 0.001) and £994.1 (bootstrap mean difference of £994.2, 95% CI: £957.1–£1026.8; P < 0.001), respectively, compared with expectant management. In addition, surgical management increased the cost of the initial hospitalisation by an average of £351.4 (bootstrap mean difference of £351.3, 95% CI: £289.2–£411.1; P < 0.001) compared with medical management. However, this was counterbalanced by an increased cost of emergency consultations and admissions for women allocated to the expectant and medical groups compared with women allocated to the surgical group. The cost of other societal impacts were, on average, £4.1 (bootstrap mean difference of £4.1, 95% CI: £1.9–£6.1; P < 0.001) and £3.0 (bootstrap mean difference of £3.0, 95% CI: £0.3–£5.8; P= 0.038) higher among women allocated to the medical and surgical groups, respectively, compared with women allocated to the expectant group. There were no significant differences in the average costs of community care, travel, lost productivity, child care and additional items bought by the women between the comparator groups.
Table 2. Unit costs of resource items (UK pounds sterling, 2001–02 prices)
P values were calculated using one-way analysis of variance.
Nonparametric bootstrap estimation was performed using 1000 replications, bias corrected.
642.8 (594.4, 691.5)
−37.8 (−64.8 to 11.6)
−33.1 (−59.6 to −8.4)
4.8 (−24.7 to 36.5)
−112.1 (−165.3 to −59.0)
−485.3 (−573.7 to −401.9)
−373.2 (−468.4 to −280.2)
Total hospital care
1.4 (−3.1 to 5.6)
0.1 (−4.5 to 4.4)
−1.3 (−5.5 to 3.0)
−1.1 (−3.8 to 2.2)
−7.3 (−17.3 to 1.1)
−6.1 (−16.0 to 1.9)
−25.1 (−73.8 to 22.1)
27.4 (−18.7 to 74.2)
−0.9 (−2.6 to 0.5)
0.1 (−1.4 to 1.4)
1.0 (−1.0 to 3.0)
Additional items bought
−1.2 (−2.8 to 0.6)
−2.1 (−4.4 to −0.1)
−0.9 (−3.0 to 1.0)
Other societal impacts
−1.1 (−3.0 to 1.2)
The net societal cost per woman was estimated at £1086.2 in the expectant group, £1410.4 in the medical group and £1585.3 in the surgical group. This represents a societal cost saving per woman of £324.2 (bootstrap mean difference of £321.4, 95% CI: £197.0–£435.4; P < 0.001) and £499.1 (bootstrap mean difference of £496.7, 95% CI: £385.0–£603.4; P < 0.001) when expectant management is compared with medical and surgical management, respectively, and £174.9 (bootstrap mean difference £175.3, 95% CI: £79.1–£268.9, P < 0.001) when medical management is compared with surgical management. Analysis of the economic data by type of miscarriage revealed that the societal cost savings attributable to expectant management were highest among women with incomplete miscarriages (Table 4), reflecting the balance of clinical benefits and risks identified by the randomised controlled trial.14
Table 4. Mean costs and mean cost differences by type of miscarriage (UK pounds sterling, 2001–02 prices)
P values were calculated using one-way analysis of variance.
Nonparametric bootstrap estimation was performed using 1000 replications, bias corrected.
The ICERs attributable to changes in the management of first-trimester miscarriage are shown in Table 5. Both expectant and medical management led to nonsignificant reductions in the mean number of gynaecological infections and significant reductions in societal costs when compared with surgical management, resulting in incremental costs per gynaecological infection avoided of −£204,968.9 and −£26,258.4, respectively. In contrast, expectant management led to a nonsignificant increase in the mean number of gynaecological infections and a significant reduction in societal costs when compared with medical management, resulting in an incremental cost per gynaecological infection avoided of £63,096.8. The bootstrapped samples of cost-effectiveness were plotted on the cost-effectiveness planes (Figures 1–3) and show the uncertainty surrounding the mean estimates of cost-effectiveness reported by the respective ICERs. For all three comparisons between the trial groups, all the bootstrapped samples fall in the southeast and southwest quadrants of the cost-effectiveness planes. A bootstrapped sample falling in southeast quadrant represents improved outcomes and lower cost as a consequence of the new form of management, while a bootstrapped sample falling in southwest quadrant represents worse outcomes and lower costs.
Negative ICERs indicate that the new policy is both more effective and cost saving, while positive ratios indicate that new policy is less effective and cost saving.
Incremental cost per gynaecological infection avoided
Community service utilisation
10% greater than reported
20% greater than reported
30% greater than reported
Per diem costs for inpatient care
20% less than accounting methods
20% greater than accounting methods
Hospital staff costs
25% reduction in staff inputs
25% increase in staff inputs
25th centile of female income distribution
75th centile of female income distribution
The cost-effectiveness acceptability curves shown in Figure 4 represent the probability of each management method being optimal in terms of cost-effectiveness at different values for decision makers’ willingness to pay for preventing one gynaecological infection. Expectant management has a 97.8% probability of being the most cost-effective management method at a willingness-to-pay threshold of £10,000 for preventing one gynaecological infection, while medical management has a 2.2% probability of being the most cost-effective management method. Expectant management retains the highest probability of being the most cost-effective management method at all willingness-to-pay thresholds of less than £70,000 for preventing one gynaecological infection.
The sensitivity analyses did not find the ICERs sensitive to the number of community service contacts reported by women, the mean level of medical and nursing staff support provided to each woman while she was in hospital or the economic value of each day of work absence (Table 5). However, assuming that the per diem costs for each level of gynaecological care were 20% less and 20% greater than those generated by our accounting methods, had the effect of reducing and increasing the ICER by £10,751.5 when expectant management was compared with medical management. Similarly, assuming that the per diem costs for each level of gynaecological care were 20% less and 20% greater than those generated by our accounting methods, had the effect of reducing and increasing the ICER by £38,938.4 when expectant management was compared with surgical management and by £5,811.8 when medical management was compared with surgical management. Simultaneous variation of the values of the key economic parameters did not significantly affect the baseline study results.
This study has demonstrated that expectant and medical management of first-trimester miscarriage possess significant economic advantages over traditional surgical management. However, the probability that surgical management is the most cost-effective management method was found to increase as the notional willingness-to-pay threshold for preventing adverse health outcomes increased, although expectant management retained the highest probability of being the most cost-effective management method at all willingness-to-pay thresholds of less than £70,000 for preventing one gynaecological infection.
The economic evaluation summarised in this study was conducted according to nationally agreed design and reporting guidelines.16 It was based on a trial that was randomised and controlled, pragmatic in design and provided a vehicle for collecting a broad set of resource use and clinical effectiveness data. Moreover, the study’s cost accounting was rigorous and included all significant resource items calculated from a societal perspective. A comprehensive analytical strategy was pursued to handle uncertainty surrounding the baseline ICERs. This included the use of cost-effectiveness acceptability curves to represent the likelihood of the management methods being cost-effective at alternative willingness-to-pay thresholds that decision makers would consider acceptable, as well as the use of sensitivity analyses to handle uncertainty surrounding individual parameter values.
There are a number of caveats to the study findings, which should be borne in mind by readers. First, the outcomes data incorporated into our cost-effectiveness estimates are not based on a synthesis of all available outcomes data through a meta-analysis of all randomised controlled trials in this area. A number of methodological issues restricted a pooling of outcomes data, including a lack of a universal definition of ultrasound scan findings that indicate incomplete miscarriage at study entry, different drug regimens administered, the varying limits of endometrial thickness used to define the presence of retained products and different outcome measures applied. Nevertheless, the sample size of the MIST Trial was sufficiently large to detect significant differences in societal costs between the comparator groups. Therefore, we would argue that the economic evaluation was sufficiently sized to arrive at conclusions that are both meaningful and relevant to decision makers. Second, the time horizon for the economic evaluation, which extended to 8 weeks postrandomisation, may have overestimated the long-term cost-effectiveness of the expectant and medical management methods, compared with surgical management, if the additional emergency consultations and admissions experienced by women in the expectant and medical groups translate into increased health service utilisation over the longer term. We have made the assumption that the consequences of first-trimester miscarriage, in terms of health status and health service utilisation, are largely felt during the first 8 weeks, but this remains to be validated by long-term follow up of the trial participants. Third, the effectiveness of the management methods has not been measured in terms of a preference-based measure of health outcome, such as the quality-adjusted life year (QALY),30 which would have been more useful for comparative purposes.31 It may be possible to map the trial outcomes onto a multi-attribute utility measure, such as the EQ-5D32 or the SF-6D,33 and then to extrapolate QALYs attributable to each management method. However, the trial revealed no significant differences between the three groups with respect to any of the eight subscales of the UK Short Form-36 health-related quality of life measure,14,34 suggesting that an exercise into eliciting or simulating women’s preferences for health outcomes will not add greatly to the results of the economic evaluation. Fourth, the study did not adopt a broad economic approach to the valuation of outcomes of the alternative management methods. It is possible that women valued a number of attributes of care they received, such as the location of care, independently of their impact upon the trial outcomes. If this were the case, then a broader economic approach to the measurement of outcomes, such as contingent valuation35 or stated preference discrete choice modelling,36 might have provided more subtle information to decision makers.
If it is accepted that the policies of expectant management and medical management of first-trimester miscarriage are cost-effective in comparison with surgical management, then decision makers must consider whether the study results can be generalised to their own settings and must decide how to organise gynaecological services. The opportunity costs of the freed beds and theatre time that would follow increased implementation of expectant or medical management will depend upon their alternative uses. In the long term, it is conceivable that the freed gynaecological beds and theatres will be filled by new gynaecological patients, the care of whom might increase overall hospital expenditure. Alternatively, the freed beds and theatres might be closed, releasing staff and support services for other areas of gynaecological care. Similarly, decision makers must consider the implications that a shift in practice might have in terms of increased pressure placed on emergency services.
In conclusion, our study provides rigorous evidence of the economic advantages of expectant and medical management of first-trimester miscarriage in comparison with surgical management. It is incumbent upon decision makers to decide whether the study results are applicable to their own settings and, if so, the appropriate configuration of gynaecological services that should be provided.
We would like to thank colleagues at the National Perinatal Epidemiology Unit, University of Oxford, who have commented on successive drafts and have, as always, given helpful advice. We would also like to thank the research nurses at each of the clinical centres who assisted with the collection of staff resource inputs, as well as all the women who participated in the trial. The study was supported by a South and West NHS Executive Research and Development grant. The National Perinatal Epidemiology Unit is core funded by the Department of Health, England. The views expressed by the authors do not necessarily reflect those of the funding bodies.