Expectant management versus surgical evacuation in first trimester miscarriage: health-related quality of life in randomized and non-randomized patients

Margreet Wieringa-de Waard,1, Esther E. Hartman, Willem M. Ankum, Johannes B. Reitsma, Patrick J.E. Bindels and Gouke J. Bonsel

Academic Medical Center–University of Amsterdam, Department of General Practice/Family Medicine, Amsterdam, The Netherlands


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
BACKGROUND: Expectant management, although less effective, is an alternative treatment option for surgical evacuation in women with a miscarriage. We assessed health-related quality of life (HRQL) differences over time between expectant and surgical management in women with a miscarriage. METHODS: Women with a miscarriage were randomized to either expectant (n = 64) or surgical (n = 58) management, and 305 eligible women who refused randomization because of a preference for either treatment option were managed according to their choice following the same clinical protocol (126 expectant, 179 surgical). The main outcome measures were score differences of HRQL during 12 weeks. Repeated measures analysis was applied. RESULTS: Out of a total of 427 women, 198 were excluded in the questionnaire follow-up, leaving 229 women who participated. Mental health of women allocated to expectant management improved more and earlier (treatment effect) than of women allocated to surgical evacuation. Mental health scores were significantly better in women who chose, rather than women who were randomized, to curettage. The groups managed according to their own preference showed no differences in mental health scores. CONCLUSION: Women with a miscarriage who chose their own treatment had the best HRQL over time, supporting the role of free choice from a clinical point of view. Women without a treatment preference should be encouraged to start with expectant management for psychological reasons.

Key words: anxiety/grief/quality of life/randomized controlled trial/spontaneous abortion


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
A woman diagnosed with early fetal demise may decide to wait for a spontaneous miscarriage, or to undergo surgical uterine evacuation. In many countries surgical evacuation is the treatment of choice among gynaecologists. Several observational studies from general practice, however, have propagated expectant management as an alternative treatment option (Ambulatory Sentinel Practice Network, 1988Go; Wiebe and Janssen, 1998Go). In a randomized controlled trial, no differences were found between expectant management and surgical evacuation in terms of safety and complication rate (Nielsen and Hahlin, 1995Go). The failure of expectant management in this study was 20%.

Equivalence in terms of complication rate, robust as these findings may be, does not rule out differences in health-related quality of life (HRQL) in women undergoing either expectant management or surgical evacuation. If substantial, these differences may even play a dominant role in guidelines and individual decisions.

In a randomized controlled trial we measured general and specific quality of life in women with miscarriages, who were managed either expectantly or by surgical evacuation. Assuming that quality of life might well be influenced by patients' treatment preferences, these results were compared with those of women who, because of a treatment preference, were managed according to their own choice (Henshaw et al., 1993Go).


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Protocol
In a partially randomized controlled trial we measured the HRQL of women with a first-trimester miscarriage. The clinical outcome measures will be reported separately. The study was conducted between April 1998 and September 2000 in two Amsterdam hospitals, the Academic Medical Center and the Onze Lieve Vrouwe Gasthuis, and was approved by the ethics committees of both hospitals. Patients with an established diagnosis of early fetal demise or incomplete miscarriage at a gestational age of <16 completed weeks participated in the medical outcome analysis and the HRQL study. Exclusion criterion for enrolment in the HRQL study was the inability to understand the Dutch or English language.

Transvaginal ultrasonographic criteria for early fetal demise were: a mean gestational sac diameter >15 mm without a measurable embryonic pole, an embryo without cardiac activity, or a gestational sac diameter <15 mm, not showing any growth after a 7 day interval (Coulam et al., 1997Go; Deaton et al., 1997Go). An incomplete miscarriage was diagnosed in cases of ultrasonographical evidence of retained products of conception >15 mm in the anteroposterior (AP) diameter. All transvaginal scans were performed by trained physicians using a transvaginal 6.5 MHz sonographic probe (Hitachi Corporation, Tokyo, Japan).

Assignment
After written informed consent had been obtained, women were randomized by the attending physician to either expectant (n = 64) or surgical (n = 58) management, using central electronic randomization. Women who refused randomization because of a preference for either one of the management options received the treatment of their own choice (preference groups, n = 126 expectant versus n = 179 surgical).

Interventions and follow-up
Surgical uterine evacuation was performed within a week after inclusion under local or general anaesthesia in daytime surgery according to the women's preference and setting availability.

Both randomized and preference groups were assessed clinically and by ultrasonography during the bi-weekly appointments at the outpatient clinic until a complete evacuation of the uterus was established either by surgical evacuation or through spontaneous loss. Women who became impatient while being managed expectantly and requested surgical evacuation were scheduled to undergo curettage within a week, but were analysed in the original treatment group on an intention-to-treat basis.

Questionnaires were filled out at five time points: immediately after inclusion, and 2, 4, 6 and 12 weeks later. All questionnaires comprised the Medical Outcome Study 36-Item Short-Form Health Survey (SF-36), the State-trait Anxiety Inventory (STAI) and the Perinatal Grief Scale (PGS). The first questionnaires also comprised sociodemographic questions.

The questionnaires were completed by the patient at home and were returned in a pre-stamped envelope. Reminders were sent to non-responders once after each time point.

Outcome measures
We used three measures, all validated in Dutch, to analyse differences in HRQL between patients: the generic SF-36, the domain-specific STAI, and a short form of the disease-specific PGS (van der Ploeg et al., 1981Go; Toedter et al., 1988Go; Potvin et al., 1989Go; Ware and Sherbourne, 1992Go; Aaronson et al., 1998Go). We used the `acute' version of the SF-36, employing a 1 week time frame. The SF-36 covers eight sub-scales of health status: physical functioning, mental health, role limitations because of emotional problems, role limitations because of physical problems, social functioning, bodily pain, vitality and general health perceptions. The scores of the eight sub-scales, which range from zero (worst health) to 100 (best health), were conventionally aggregated into the standardized Physical Component Summary scale (PCS) and Mental Component Summary scale (MCS), both with mean 50, SD 10. The STAI contains two 20-item scales covering both current (state) and background (trait) anxiety; items are rated on a four-point scale with total scores ranging from 20–80, where higher scores represent higher levels of anxiety (Spielberger et al., 1970Go; van der Ploeg et al., 1981Go). The PGS measures grief associated with pregnancy loss. The short form used here contains 33 questions, to be answered on a five-point Likert scale ranging from `strongly agree' to `strongly disagree' with a neutral mid-point. The PGS consists of three sub-scales: active grief, difficulty coping and despair, which range from 11–55 with higher scores indicating more grief. Active grief is the normal grief reaction that may follow a pregnancy loss. Difficulty coping and despair represent increasingly severe forms of grieving. The PGS had previously been confirmed as reliable and valid in the evaluation of women who miscarried (Potvin et al., 1989Go; Cuisinier et al., 2000Go).

Reference scores for the SF-36 sub-scales were obtained from published data of a Dutch population sample (16–40 years of age) (Ware et al., 1993Go, 1994Go, 1998Go; Aaronson et al., 1998Go).

Sample size
For the power calculation we focused on the SF-36. Assuming a five-point difference in the summary PCS or MCS measures as being clinically relevant, we needed 33 patients in each randomized group to detect this difference (alpha = 0.05, beta = 0.2) (Ware et al., 1994Go).

Statistical analysis
We analysed the data according to the intention-to-treat principle, i.e. women were analysed in the group into which they were randomized, irrespective of whether or not they received that treatment. Women with missing data were included if at least two sets of questionnaires were available. We assumed that women with a short period until complete evacuation of the uterus (spontaneously or by surgical intervention) might be under-represented in the study and we therefore compared the time until complete evacuation between included and excluded women. Differences in patient characteristics and HRQL at inclusion between the randomized groups and the preference groups were analysed using descriptive statistics. We also analysed the relation between period of delay in conceiving (in months) and the mental health score after 12 weeks.

We employed repeated measurements design to compare the HRQL changes of the randomized groups and the preference groups in more detail. A mixed linear model was used to sort out the treatment effect and the treatment assignment effect, apart from time and other effects, as expressed in observed patterns in the various quality of life measurements. The mixed linear model methodology accounts for the correlation between different measurements over time within the same women, thus avoiding overestimation of specific effects such as due to treatment. For each participant we assumed in this analysis that the more measurements were closer in time, the more they were correlated. Our model included the following explanatory variables: the (initial) treatment option (at four different levels: curettage and expectant in both the randomized and preference group), time (as a categorical variable with four levels, excluding the baseline measurements), interaction between treatment and time, and the baseline measurement. P values < 0.05 were considered significant. Results were presented as differences in mean score with 95% confidence intervals (Verbeke and Molenberghs, 1997Go; Littell et al., 2000Go).

The Statistical Package of the Social Sciences (SPSS, version 10.07) was used for all standard analyses. All mixed models were done in SAS 6.12 (syntax and data available from the authors).


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Participant flow and follow-up
Out of 1101 women, 652 were excluded after ultrasound diagnosis (viable pregnancy, complete miscarriage) leaving 449 women with an ultrasound diagnosis of early fetal demise or incomplete miscarriage. In 22 women an immediate curettage was necessary, leaving 427 eligible for randomization (Figure 1Go). Of these, 122 gave consent to randomization and 305 women with a strong treatment preference were managed according to their own choice. Of the women allocated to expectant management or preferring this management, 50% requested (and received) a curettage after a period of waiting in vain for a spontaneous loss (data not shown).



View larger version (32K):
[in this window]
[in a new window]
 
Figure 1. Flow diagram. HRQL = health-related quality of life; RCT = randomized controlled trial.

 
Of all 427 women participating in the medical outcome analyses, 198 (or 46%) were excluded in the questionnaire follow-up of the HRQL study because of insufficient Dutch or English language skills (n = 90), because they refused to participate in this research (n = 30), or because they returned only one questionnaire (n = 78). The remaining 229 women (54%) completed two or more questionnaires and were included in the present study. Participants originated more often from western European countries compared with excluded women (72% versus 46%). They also showed a higher degree of education. No differences in age or prior experience with any of the management options were observed between participants and excluded women; also the interval between enrolment in the study and curettage or spontaneous loss of pregnancy did not differ (data not shown).

Analysis
No significant differences in sociodemographic or clinical characteristics were present between the randomized groups, nor between the preference groups. Also, no differences in prior experience with one of the management options, education, native country or anxiety (STAI) were observed.

At study inclusion we did not find significant differences in scores between the two randomized groups (data not shown). However, women preferring expectant management showed significantly better scores on the mental health sub-scale (mean score 66) than women preferring curettage (mean score 57). Similarly, at inclusion women preferring expectant management showed significantly less active grief (mean score 30) than women preferring curettage (mean score 35).

The course of the (unadjusted) mean values for the SF-36 mental and physical summary scores, STAI and PGS total scores in the four groups are illustrated in Figure 2Go.



View larger version (29K):
[in this window]
[in a new window]
 
Figure 2. Mean health-related quality of life (HRQL) scores of randomized and preference groups. Direction of arrows indicates better scores. rct = randomized controlled trial; pref = preferred; norm = normal.

 
We found no correlation between the time taken to achieve pregnancy and mental health scores at 12 weeks.

Changes in HRQL over time in the four groups were analysed in a repeated measurements mixed model.

Randomized groups
There were no significant differences between the two treatments in physical health (PCS) scores (treatment effect P = 0.7). Women randomized to expectant management, however, had significantly better mental health (MCS) scores (treatment effect P = 0.004). The overall differential treatment effect (difference between the expectant management group versus the curettage group, all time points together) was 7.4 points on the MCS in favour of expectant management. The difference at 12 weeks was still 6.3. There were no significant differences in changes over time between the two treatment arms in PCS and MCS (time by treatment P = 0.6).

We found no significant differences in improvement between the two arms in state–anxiety scores of the STAI. The negative score indicates a lower average score (less anxiety) for the expectant management group, but this was not significant (treatment effect P = 0.09).

Patients randomized to expectant management also showed lower scores on the perinatal grief scale total and its sub-scales (indicating less grief) than patients randomized to curettage, but again the difference did not reach statistical significance (treatment effect P = 0.3).

Preference groups
No statistically significant differences in any of the HRQL measures was observed between women preferring expectant management and those preferring surgical evacuation.

Randomized versus preference groups
In expectantly managed patients, no differences in HRQL measures were found between those treated according to randomized treatment allocation compared with those managed according to their preference. Between women randomized to curettage and women who preferred curettage a significant overall five-point difference (P = 0.03) was found in MCS scores. Women treated according to randomized allocation performed worse than those who received the same treatment at their own request.

Additional analysis
In randomized women no differences were found between the HRQL scores of women allocated to the treatment for which they expressed a slight preference shortly before randomization and those not allocated to their preferred treatment. We also found no differences between the proportion of high responders (i.e. women who returned four or five questionnaires) and low responders (returning two or three questionnaires) in the randomized and in the preference groups.


    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
In this study of women with a miscarriage we found better scores of the SF-36 mental component summary scale of women managed expectantly versus those treated surgically according to randomized assignment. This difference was present during the entire 3 month study period and still existed after 12 weeks. In contrast, no differences in HRQL measures were seen between the two treatment options in women who were managed according to their own preference. In the comparison between randomized and non-randomized patients, women preferring curettage showed significantly better mental health scores than women undergoing the same treatment after randomization. Our results confirm the impact on experienced HRQL of patients' preferences and support the shared decisions model in the management of miscarriages (Ankum et al., 2001Go).

We would conclude that a decreased motivation among women randomized to a moderately non-preferred treatment could not explain the differences between the two randomized groups; HRQL scores did not differ between women who were randomized to their preferred treatment and women who were not. Another explanation for the differences found in mental health scores may be the absence of a treatment option for women randomized to curettage as opposed to women randomized to expectant management, who still have the option to change their mind and have a curettage at a later stage. This escape is far from theoretical as 50% of the women allocated to expectant management took advantage of this option after a waiting period. Although expectant management is less effective than surgical evacuation, a waiting period may prevent many curettages. In our study, 38% of the expectantly managed women miscarried within a week (data not shown). Another study (Nielsen and Hahlin, 1995Go) reported a lower failure of expectant management. However, in this study most women already had an incomplete miscarriage. A study comparing misoprostol and expectant management (Ngai et al., 2001Go) reported a success rate in the misoprostol group versus the expectant group of 83.3 versus 46.3%.

The potential for generalization of a randomized trial may be limited if many patients show an outspoken treatment preference, with consequent refusal to participate in such a study. Usually they receive the treatment of their choice after having been excluded from the trial (Brewin and Bradley, 1989Go). Our study, including both assignment groups with identical clinical protocol, confirms that providing a free choice of treatment increases the perceived quality of life that comes with it (Henshaw et al., 1993Go).

We did not expect the substantially lower mental health scores at inclusion among women preferring surgical evacuation in comparison with all other women. Maybe this reflects the mental attitude and stability required in accepting the uncertainties of expectant management.

The differential response to both treatments has received little attention so far. To our knowledge only one study has compared anxiety and depressive reactions in a randomized trial comparing expectant management and surgical evacuation (Nielsen et al., 1996Go). That study reported no differences in anxiety or depressive reactions between the treatment groups after a 2 week period. Similarly, we found no differences in anxiety after 2 weeks or at 12 weeks, but we did observe an overall better mental health during the entire 12 week observation period in women allocated to expectant management. This justifies the recommendation of expectant management to those women without strong treatment preferences.


    Acknowledgements
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
All authors participated in the writing of the paper. M.W. collected the data. M.W., E.E.H. and J.B.R. analysed the data. W.M.A. and P.J.E.B. were involved in the design of the study and supervised the study; G.J.B. was involved in the design of the study and supervised data analysis. The study was supported by grants from the Dutch Health Research and Development Council (ZON) and the Dutch Ministry of Health, Welfare and Sports.


    Notes
 
1 To whom correspondence should be addressed at: Academic Medical Center–University of Amsterdam, Department of General Practice/Family Medicine, Division of Public Health, PO Box 22700, 1100 DE Amsterdam, The Netherlands. E-mail: m.wieringa{at}amc.uva.nl Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Aaronson, N.K., Muller, M., Cohen, P.D.A., Essink-Bot, M., Fekkes, M., Sanderman, R., Sprangers, M.A.G., te Velde, A. and Verripst, E. (1998) Translation, validation, and norming of the Dutch language version of the SF-36 health survey in community and chronic disease populations. J. Clin. Epidemiol., 51, 1055–1068.[ISI][Medline]

Ambulatory Sentinel Practice Network (1988) Spontaneous abortion in primary care. J. Am. Board. Fam. Pract., 1, 15–23.[Medline]

Ankum, W.M., Wieringa-de Waard, M. and Bindels, P.J.E. (2001) Management of spontaneous miscarriage in the first trimester: an example of putting informed shared decision making into practice. Br. Med. J., 322, 1343–1346.[Free Full Text]

Brewin, C.R. and Bradley, C. (1989) Patient preferences and randomised clinical trials. Br. Med. J., 299, 313–315.[ISI][Medline]

Coulam, C.B., Goodman, C. and Dorfmann, A. (1997) Comparison of ultrasonographic findings in spontaneous abortions with normal and abnormal karyotypes. Hum. Reprod., 12, 823–826.[Abstract]

Cuisinier, M.C., Graauw de, C.P. and Kuijpers, J.C. (2000) Meting van rouw na een zwangerschapsverlies. Eerste bevindingen met de Nederlandse versie van de Perinatal Grief Scale. Tijdschrift voor Psychiatrie, 34, 651–656.

Deaton, J.L., Honore, G.M., Huffman, C.S. and Bauguess, P. (1997) Early transvaginal ultrasound following an accurately dated pregnancy: the importance of finding a yolk sac or fetal heart motion. Hum. Reprod., 12, 2820–2823.[Abstract]

Henshaw, R.C., Naji, S.A., Russell, I.T. and Templeton, A.A. (1993) Comparison of medical abortion with surgical vacuum aspiration: women's preferences and acceptability of treatment. Br. Med. J., 307, 714–717.[ISI][Medline]

Littell, R., Pendergast, J. and Natarajan, R. (2000) Modelling covariance structure in the analysis of repeated measures data. Stat. Med., 19, 1793–1819.[ISI][Medline]

Ngai, S.W., Chan, Y.M., Tang, O.S. and Ho, P.C. (2001) Vaginal misoprostol as medical treatment for first trimester spontaneous miscarriage. Hum. Reprod., 7, 1493–1496.

Nielsen, S. and Hahlin, M. (1995) Expectant management of first-trimester spontaneous abortion. Lancet, 345, 84–86.[ISI][Medline]

Nielsen, S., Hahlin, M., Moller, A. and Granberg, S. (1996) Bereavement, grieving and psychological morbidity after first trimester spontaneous abortion: comparing expectant management with surgical evacuation. Hum. Reprod., 11, 1767–1770.[Abstract]

Ploeg van der, H.M., Defares, P.B. and Spielberger, C.D. (1981) Zelfbeoordelingsvragenlijst. Handleiding. Swets en Zeitlinger, Lisse.

Potvin, L., Larsen, J. and Toedter, L.J. (1989) Measuring grief: a short version of the perinatal grief scale. J. Psychopathol. Behavior. Assess., 11, 29–45.

Spielberger, C.D., Gorsuch, R.L. and Lushene, R.E. (1970) STAI manual for the state–trait anxiety inventory. Consulting Psychologists Press Inc., Palo Alto, CA, USA.

Toedter, L.J., Lasker, J.N. and Alhadeff, J.M. (1988) The perinatal grief scale: development and initial validation. Am. J. Orthopsychiat., 58, 435–449.[ISI][Medline]

Verbeke, G. and Molenberghs, G. (1997) Linear mixed models in practice: an SAS-oriented approach. Springer, New York.

Ware, J.E.J. and Sherbourne, C.D. (1992) The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med. Care., 30, 473–483.[ISI][Medline]

Ware, J.E., Snow, K.K., Kosinski, M. et al. (1993) SF-36 health survey manual and interpretation guide. New England Medical Center, The Health Institute, Boston.

Ware, J.E., Kosinski, M. and Keller, S.D. (1994) SF-36 physical and mental summary scales: a user's manual. New England Medical Center, The Health Institute, Boston.

Ware, J.E., Gandek, B., Kosinski, M., Aaronson, N.K., Apolone, G., Brazier, J., Bullinger, M., Kaasa, S., Leplege, A., Prieto, L. et al. (1998) The equivalence of SF-36 summary health scores estimated using standard and country-specific algorithms in 10 countries: results from the IQOLA project. International quality of life assessment. J. Clin. Epidemiol., 51, 1167–1170.[ISI][Medline]

Wiebe, E. and Janssen, P. (1998) Management of spontaneous abortion in family practices and hospitals. Fam. Med., 30, 293–296.[Medline]

Submitted on October 1, 2001; accepted on January 25, 2002.