Predictors of complete miscarriage after expectant management or misoprostol treatment of non-viable early pregnancy in women with vaginal bleeding

Purpose To identify predictors of complete miscarriage after expectant management or misoprostol treatment of non-viable early pregnancy in women with vaginal bleeding. Methods This was a planned secondary analysis of data from a published randomized controlled trial comparing expectant management with vaginal single dose of 800 µg misoprostol treatment of women with embryonic or anembryonic miscarriage. Predefined variables—serum-progesterone, serum-β-human chorionic gonadotropin, parity, previous vaginal deliveries, gestational age, clinical symptoms (bleeding and pain), mean diameter and shape of the gestational sac, crown-rump-length, type of miscarriage, and presence of blood flow in the intervillous space—were tested as predictors of treatment success (no gestational sac in the uterine cavity and maximum anterior–posterior intracavitary diameter was ≤ 15 mm as measured with transvaginal ultrasound on a sagittal view) in univariable and multivariable logistic regression. Results Variables from 174 women (83 expectant management versus 91 misoprostol) were analyzed for prediction of complete miscarriage at ≤ 17 days. In patients managed expectantly, the rate of complete miscarriage was 62.7% (32/51) in embryonic miscarriages versus 37.5% (12/32) in anembryonic miscarriages (P = 0.02). In multivariable logistic regression, the likelihood of success increased with increasing gestational age, increasing crown-rump-length and decreasing gestational sac diameter. Misoprostol treatment was successful in 80.0% (73/91). No variable predicted success of misoprostol treatment. Conclusions Complete miscarriage after expectant management is significantly more likely in embryonic miscarriage than in anembryonic miscarriage. Gestational age, crown-rump-length, and gestational sac diameter are independent predictors of success of expectant management. Predictors of treatment success may help counselling women with early miscarriage.

The aim of this work is to identify predictors of success of expectant management or misoprostol treatment in a welldefined group of patients, i.e., patients with embryonic or anembryonic miscarriage reporting vaginal bleeding.

Material and methods
This is a planned secondary analysis of prospectively collected data in a published randomized controlled trial comparing expectant management with vaginal single dose 800 µg misoprostol treatment of embryonic and anembryonic first-trimester miscarriage [6] (ClinicalTrials.gov ID: NCT01033903). The regional ethical review board, Lund University, Sweden, approved the trial (Dnr 83/2008), which was carried out in accordance with the code of ethics of the Declaration of Helsinki. The primary trial outcome was complete miscarriage ≤ 10 weeks. Secondary outcomes were complete miscarriage ≤ 17, ≤ 24 and ≤ 31 days. Complete miscarriage was defined as no gestational sac in the uterine cavity and maximum anterior-posterior diameter of the intracavitary contents ≤ 15 mm as measured with transvaginal ultrasound on a sagittal view. Trial details (including a CONSORT flow chart) have been published [6] and are briefly outlined below.
Women consulting the gynecological emergency clinic of Skåne University Hospital, Malmö, Sweden and reporting vaginal bleeding in early pregnancy were eligible for inclusion if hemodynamically stable. Women with heavy bleeding needing urgent surgical evacuation of the uterine cavity, as judged clinically, were not eligible. Inclusion criteria were: ≥ 18 years old, understanding written and spoken Swedish, hemoglobin concentration > 80 g/L, no contraindications to misoprostol treatment, fulfilling ultrasound criteria of anembryonic or embryonic miscarriage [11], and fetal crown-rump-length ≤ 33 mm. Because of new recommendations [33][34][35] our ultrasound criteria of non-viability were changed in 2014. Before 2014 the criteria were (1) intracavitary gestational sac with a diameter (mean of three orthogonal diameters) of > 16 mm with no embryonic pole [36] or (2) intracavitary gestational sac with an embryo with crown-rump-length ≥ 5 mm without cardiac pulsations [36], or (3) if the above criteria were not fulfilled, intracavitary gestational sac with or without an embryo showing no significant development at a repeat scan after 7 days [36]. After April 2014 our non-viability criteria were: (1) intracavitary gestational sac with a mean diameter ≥ 25 mm with no embryonic pole, or (2) intracavitary gestational sac with an embryo with crown-rump-length ≥ 7 mm without cardiac pulsations, or (3) if the above criteria were not fulfilled, no significant development at a repeat scan after 7 days [33,34,37,38]. Women with incomplete miscarriage were not eligible. Women giving written consent were randomized into two parallel groups in an open-label 1:1 ratio to misoprostol treatment or expectant management. All patients were managed as outpatients.
The patients were examined clinically and with transvaginal ultrasound in the lithotomy position on the day of randomization. The trial clinician estimated bleeding and pain at speculum and clinical examination and placed 800 μg misoprostol in the posterior vaginal fornix of the patients allocated to medical treatment. The first follow-up was after 10 days. Subsequent follow-up visits were scheduled every 7 days until the miscarriage was complete (definition, see above). The patient was then discharged with no planned follow-up visits. If complete evacuation was not achieved on day 31, dilatation and evacuation was recommended. However, the participants could ask for dilatation and evacuation at any time for any reason during the trial.
At inclusion, information on demographic background data was collected from the woman and documented in research forms, and blood was drawn for analysis of hemoglobin, β-human chorionic gonadotropin (β-hCG), progesterone and blood type. hCG + β was measured with a sandwich immunoassay on a Cobas® instrument; Intact hCG + the β-subunit assay (Roche Diagnostics, Mannheim, Germany). Progesterone was measured with a competitive immunoassay on a Cobas® instrument; Progesterone III assay (Roche Diagnostics, Mannheim, Germany). On every follow-up visit clinical examination and transvaginal ultrasound were performed. Results were prospectively entered into research forms.
The ultrasound machine used was a Sequoia 512 ultrasound system (Siemens Medical Solutions Inc., Ultrasound Division, Mountain View, CA, USA) with a 4-8 MHz transducer. The shape of the gestational sac was assessed, the sac diameter and the crown-rump-length (if an embryo/fetus was present) were measured. The miscarriage was classified as embryonic or anembryonic. Assessment of blood flow in the presumed intervillous space was first made by looking for  [39,40]. Presence of blood flow in the presumed intervillous space as assessed on grey scale and color Doppler and of arterial Doppler shift signals in the presumed intervillous space was noted.
Women not showing up on a scheduled visit were included in this secondary analysis only if on a later visit the miscarriage was incomplete; the miscarriage was then classified as incomplete also on the missed visit. Women with a complete miscarriage on the first visit after the missed one were not included in the analysis, because we do not know, if the miscarriage was complete also on the missed visit.
We considered the following variables to be possibly related to treatment outcome and explored their ability to predict complete miscarriage ≤ 10 days and ≤ 17 days: serum/plasma levels of progesterone (nmol/L) and β-hCG (IU/L) at inclusion, gestational age (days) according to the last menstrual period (LMP), previous vaginal delivery (yes/no), parity (yes/no), bleeding at inclusion (moderate or heavy versus mild or none as assessed by the trial physician at speculum examination), pain at inclusion (yes/no), shape of the gestational sac (round or oval versus else), mean gestational sac diameter (mm), crown-rump-length (mm), type of miscarriage (embryonic or anembryonic), presence of blood flow in the presumed intervillous space according to grey scale and color Doppler ultrasound (yes/no), and presence of arterial Doppler shift signals in the presumed intervillous space (yes/no).
Statistical analysis was performed using SPSS Statistics, version 21 (IBM Corp., Armonk, NY, USA). It was done separately for the misoprostol group and the expectantly managed group and separately for treatment success ≤ 10 days and ≤ 17 days. The relation between the predefined predictor variables and treatment success was tested for statistical significance using univariable logistic regression with the likelihood ratio test. Two-tailed P values < 0.05 were considered statistically significant. Multivariable logistic regression was used to elucidate which variables were independently associated with treatment success and for building models to predict complete miscarriage. Because crown-rump-length and miscarriage type are related (in anembryonic miscarriages the crown-rump-length is zero), only one of these variables was included in the same multivariable analysis. We used different approaches for model building. In one, we included all predetermined variables (with a minimum of five individuals in each cell of a fourfield table) in multivariable backward step-wise logistic regression analysis. In another, we started with including only variables with a P value < 0.20 in univariable analysis, and then we tested to add variables that we found clinically relevant.
Individual data for each patient were inserted into the regression models to calculate the probability of complete miscarriage for each patient and to plot receiver-operating characteristics curves. The area under the receiver-operating characteristic curve (AUC) and its 95% confidence interval (CI) were calculated. If the lower limit of the CI was > 0.5 the model was considered to have discriminatory ability. The larger the AUC the better the discriminative ability.

Results
Between September 2008 and December 2015 189 women were recruited into the trial. Ninety-five women were allocated to expectant management and 94 to misoprostol treatment. Twenty-one women (expectant group: 11; misoprostol group: 10) were recruited after the new ultrasound criteria of non-viability were adopted. After exclusions, our planned secondary analysis included 177 women (expectant group: 85; misoprostol group: 92). The reasons for exclusion were: Data are presented as n (%), mean ± SD or median (range) CI confidence interval a Likelihood ratio test b After mid May 2013 beta-human chorionic gonadotropin and progesterone were analyzed in plasma not in serum, reference intervals are not affected c Odds ratio is calculated for units of 1000 d Bleeding as judged by the trial physician at speculum examination-all women reported vaginal bleeding e Mean of three orthogonal diameters withdrawal of consent (expectant group, n = 2), not fulfilling inclusion criteria (expectant group, n = 3), undergoing dilatation and evacuation before first follow-up at 10 days (expectant group, n = 5; misoprostol group, n = 2). For the analysis regarding prediction of complete miscarriage ≤ 17 days another three patients were excluded due to dilatation and evacuation on patient's request (expectant group n = 1) or missed follow-up visit (expectant group n = 1, misoprostol group n = 1), the numbers analyzed for this outcome being 83 (expectant group) and 91 (misoprostol group). Expectant management was successful ≤ 10 days in 45.8% (39/85) of the patients and ≤ 17 days in 53.0% (44/83). Variables associated with success of expectant management in univariable and multivariable analyses are shown in Tables 1, 2, 3, 4 and 5. Treatment success was more common in embryonic than anembryonic miscarriages: complete miscarriage ≤ 10 days 53.8% (28/52) versus 33.0% (11/33) (P = 0.06), complete miscarriage ≤ 17 days 62.7% (32/51) versus 37.5% (12/32) (P = 0.02). In multivariable analyses, the following variables were independently associated with treatment success: gestational age according to LMP (the higher the more likely successful treatment), mean gestational sac diameter (the smaller the more likely successful treatment) and crown-rump-length (the larger the more likely successful treatment). When we replaced crownrump-length with type of miscarriage in the multivariable model, the odds of treatment success were approximately six times higher in embryonic than anembryonic miscarriages (Table 3). When s-β-hCG and s-progesterone were tested as predictors in multivariable analysis, either s-progesterone or s-β-hCG (not both) and crown-rump-length or miscarriage type were independently associated with complete miscarriage (Tables 4 and 5). The AUCs of the models ranged from 0.71 to 0.77.
In embryonic miscarriages, the higher the gestational age according to LMP, the larger the crown-rump-length and the smaller the mean gestational sac diameter the higher the likelihood of complete miscarriage ≤ 10 and ≤ 17 days (Table 6); and the lower the s-progesterone or s-β-hCG and the longer the crown-rump-length the higher the success rate ( Table 7). The AUCs of the models ranged from 0.80 to 0.84. No variable was statistically significantly associated with complete miscarriage of anembryonic miscarriages when using expectant management (details available from the authors on request).
Misoprostol treatment was successful ≤ 10 days in 67% (62/92) of patients and ≤ 17 days in 80% (73/91). No variable predicted success of misoprostol treatment either in embryonic or anembryonic miscarriages. (Details available from the authors on request).

Comments
In patients managed expectantly the likelihood of spontaneous complete miscarriage ≤ 10 or ≤ 17 days was twice as high in embryonic as in anembryonic miscarriages. The likelihood of complete miscarriage increased with increasing gestational age according to LMP, increasing crownrump-length and decreasing gestational sac diameter; and the larger the crown-rump-length and the lower the s-β-hCG or s-progesterone level the higher the likelihood of treatment success. No variable predicted treatment outcome in the misoprostol group.
The strength of our study is the well-defined study population. We included only embryonic and anembryonic miscarriages and only women reporting vaginal bleeding. Our results are generalizable to such women. It is a limitation that we changed our definitions of non-viable pregnancy at the end of the recruitment period because of new research results and guidelines [33-35, 37, 38, 41-43]. However, it is unlikely that this had any major impact on our results. Our definition of successful treatment, also used by others [7,8,19,27,44], may also be criticized [45][46][47][48]. Absence of a gestational sac and/or cessation of vaginal bleeding may be better definitions [49]. It is a weakness that our prediction models have not been validated in a new study sample. In our trial, we did not find any association between presence of blood flow or pulsatile flow in the intervillous space and treatment success. However, we did not make any attempts Data are presented as n (%), mean ± SD, median (range) CI confidence interval a Likelihood ratio test b After mid May 2013 beta-human chorionic gonadotropin and progesterone were analyzed in plasma and not in serum, reference intervals are not affected c Odds ratio is calculated for units of 1000 d Bleeding as judged by the trial physician at speculum examination-all women reported vaginal bleeding e We consider this result unreliable, because only three patients in the failure group had heavy or moderate bleeding, the unreliable result is also reflected in the odds ratio f Mean of three orthogonal diameters to quantify blood flow, for example using three-dimensional ultrasound and calculating vascular indices. This may be seen as a limitation.
The results of other studies exploring possible predictors of complete miscarriage with expectant management are very heterogenous (Table 8) [13-22, 25, 30]. This is probably explained by differences in study populations (types of miscarriage, symptomatology), definitions of complete miscarriage and treatment success, and variables tested as predictors. However, more than one study reported that the lower the s-β-hCG and s-progesterone values the higher the likelihood of success of expectant management [16,18,19,21] and that success rate is higher in incomplete miscarriages than in embryonic or anembryonic miscarriages [15][16][17]. Only two studies are reasonably similar to ours with regard to inclusion criteria and definition of treatment success [18,21]. Schwärzler et al. found that pulsatile flow in the presumed intervillous space was a predictor of successful treatment [21]. We could not confirm this, perhaps because examination of blood flow in the intervillous space in early pregnancy requires skill, carefulness and time. Therefore, it is unlikely to be useful in busy emergency departments or early pregnancy units. Memtsa et al. reported that the older the patient and the lower the s-progesterone level the higher the likelihood of complete miscarriage < 7 days [18]. We did not test patient age as a predictor, because we found it unlikely to be related to the time to complete evacuation of the uterine cavity.
Our results seem plausible from a pathophysiological perspective. Higher concentrations of s-β-hCG and s-progesterone probably reflect better functioning corpus luteum and trophoblast and are compatible with earlier stages of miscarriage explaining longer time to complete miscarriage. The relation between gestational age according to LMP, crownrump-length and gestational sac diameter may reflect the time since the embryo died. The apoptotic process may be faster for the gestational sac than for the embryo. This could be an explanation for a smaller gestational sac being a predictor of treatment success. Long time between embryonic death and start of bleeding may also indicate resistance to expulsion. In some anembryonic miscarriages, the absence of an embryo might be explained by the embryo having been resorbed after having been dead for a long time.
The likely reason why we did not find any predictor of successful misoprostol treatment is that misoprostol is very effective in most patients. Results of other studies investigating predictors of successful misoprostol treatment are extremely variable (Table 8) [23, 24, 26-29, 31, 32] and no study is directly comparable to ours. The variable results are      probably explained by differences in patient selection, definition of treatment success and predictor variables tested. Some women prefer expectant management to medical intervention [17,49] and knowledge about prognostic factors is important for being able to provide them with realistic expectations of treatment success. Our prediction models are helpful in this respect but need to be prospectively validated. A simple blood test predicting successful outcome without treatment would be clinically valuable and therefore a goal of future research. Serum-progesterone or serum-hCG results may not always be available at the time of patient counselling. Therefore, they may be less clinically useful as predictors than clinical and ultrasound information available bedside. (+) the higher the value of the variable tested the higher the success rate of miscarriage management, alternatively if the variable was present the likelihood of success increased; (−) the higher the value of the variable tested the lower the success rate of miscarriage management, alternatively if the variable was present the likelihood of success decreased