Psychological interventions improve quality of life despite persistent pain in endometriosis: results of a 3-armed randomized controlled trial

Patients were randomly assigned to one of three conditions: (1) A specific mindfulness- and acceptance-based PI called “Mind Your ENDOmetriosis” (MY-ENDO), (2) A non-specific PI (Non-specific) that matched MY-ENDO in non-specific factors such as empathy, the therapeutic alliance, a cogent rationale, and expectations of improvement, but did not include the assumed specific ingredient, mindfulness or (3) A WL that involved treatment, as usual, to control for the natural fluctuations in pain [31]. Participants in the waitlist group were offered one of the two PIs after the end of the study period. All groups received medical treatment as usual. This design enabled a rigorous examination of the efficacy of MY-ENDO to clarify to which extent specific mindfulness- and acceptance ingredients are essential for the potential effects of this intervention.

The study was preregistered with The Danish Data Protection Agency (journal no. 2015-57-0002), approved by The Central Denmark Region Committees on Health Research Ethics (registration no. 1-10-72-138-15), and retrospectively registered at clinicaltrials.gov (NCT02761382). Data was collected from March 2016 to October 2018.

Participants were recruited from three specialized outpatient clinics for endometriosis in Denmark and from the Danish Endometriosis Patients Association. All patients considered for inclusion underwent screening to assess in- and exclusion criteria. Inclusion criteria were: (a) 18–47 years old, (b) surgery or MRI-confirmed endometriosis diagnosis, (c) moderate to severe CPP (i.e., an average of ≥ 5 measured on an 11-point Numerical Rating Scale (NRS) from 0 = no pain to 10 = worst pain imaginable), (d) relevant clinical and surgical treatment according to the European Society of Human Reproduction and Embryology (ESHRE) guidelines for endometriosis [32] had been tried, (e) willingness to spend 30–45 min on homework 5–7 days a week for 10 weeks. Exclusion criteria were (a) other serious physical pain diseases (e.g., fibromyalgia, Crohn’s disease, Colitis Ulcerosa), (b) severe psychiatric diagnosis, (c) pregnancy or planned pregnancy during the study period, and (d) an estimated lack of mental or physical surplus to enter into a psychological treatment or linguistic or cultural barriers.

A letter was sent to interested patients with study details and a pain diary to be filled out before the screening session. At the screening, patients were informed about study requirements and screened for in- and exclusion criteria. They provided written informed consent before enrolment in the study and randomization. Patients were informed that they would be randomized to one of two different psychological interventions or a waitlist control group. This should keep participants blinded to the psychological method and intervention content in the comparison group. To keep the research group blinded to intervention assignment throughout data collection, a research assistant, not part of the research group, provided patients with an anonymous id-number used for data collection. The numbers were randomized in blocks of six by another research assistant using a computer-generated randomization list.

Questionnaires were sent to participants by postal mail, filled out, and returned. Baseline measurements were obtained during the 2 weeks period prior to treatment start, and post-intervention measurements were obtained during the 2 weeks post-treatment period. At home, patients also completed a 12-week pain diary starting 1 week pre-intervention until 1-week post-intervention. To investigate potential changes in pain processing and sensibility a female doctor carried out a gynecological experimental pain assessment during the 2-week period prior to treatment start and again during the 2 weeks post-treatment. However, the experimental pain assessment was optional and not required for participation (See Fig. 1. Study timeline).

Two private practicing clinical psychologists, both licensed by the Danish board of psychologists and closely matched on essential parameters (training, apprenticeship, and competence), were recruited to deliver the interventions in a “crossed-therapist” design, with both therapists providing both treatments within the study.

Along with a sociodemographic questionnaire, patients received the following questionnaires:

Sample size was based on power analysis of a small randomized 3-armed pilot study (unpublished) for the primary outcome of pelvic pain between the groups: (1) PI vs. WL and (2) MY-ENDO vs. Non-specific and for the secondary outcome of QoL between the groups PI vs. WL. Pelvic pain was measured on NRS. The NRS scale score is standardized on a range from 0–10, defined by a mean of M = 6.0 and the standard deviation (SD) = 1.5. With the reasonable assumptions: Mean n1 = 6.0, Mean n2 = 5.0, SD = 1.5, power (1 − β) = 0.80, α = 0.05, two-sample, two-sided test, the number of participants needed would be 53 vs. 27 participants for differences between the groups PI vs. WL, and 36 vs. 36 participants for differences between the groups MY-ENDO vs. Non-specific. QoL was measured on EHP-30. The EHP-30 scale scores are standardized on a range from 0 to 100, defined by a mean of M = 50.0 and the standard deviation SD = 12.0. With the reasonable assumptions: Mean n1 = 50.0, Mean n2 = 40.0, SD = 12.0, power = 0.80, α = 0.05, two-sample, two-sided test, the number of participants needed would be 34 vs. 17 participants for differences between the groups PI vs. WL. Based on the power analyses it was planned to include 3 × 27 participants in the study.

Baseline group differences were compared by the χ²-test or the Kruskal–Wallis test (due to non-normally distributed data). For continuous data, means and standard deviations were given. Normally distributed variables were compared using t-tests, and non-normally distributed variables were compared using non-parametric tests (i.e., Mann–Whitney). Study dropouts were defined as participants discontinuing the intervention or failing to return the questionnaires/diary. Mixed linear models (MLMs) were used to compare groups over time and to examine changes in outcomes over time within groups on all outcomes. MLMs tolerate missing values without compromising statistical power and take into account the nested nature of data. The MLM models were conducted using restricted maximum likelihood method (REML) and performed on an intention-to-treat basis. Data were hierarchically arranged with time as level one nested within individual as level two. Fixed effects were specified for intercept, time, group, and time × group interaction. All models included a random intercept, and a fixed slope was chosen due to the comparison of groups with small sample sizes. In order to compare end-point effects between measures, a linear function of time was estimated from baseline to post-treatment measure. All primary analyses were conducted blinded. Since analyses of the primary outcome yielded unexplained results, which were in contrast to the hypotheses, statistically significant (P < 0.05) baseline differences were entered as covariates in explorative post-hoc analyses of the primary outcome [47]. In addition, because of holiday periods, some patients completed an additional diary week, however, this holiday week was subtracted in the statistical analysis, and because some patients had missing values during the 12 weeks and some stopped completing the pain diary already after 11 weeks, sensitivity analysis was performed testing the robustness of the results. The sensitivity analyses were conducted using MLMs comparing the groups over time using last observation carried forward for missing values and comparing the groups over time after 11 weeks (week 12 was subtracted for all participants). Statistically significant results were defined as P < 0.05 (two-sided significance level). Effect sizes were expressed as Cohen´s d, with effect sizes of 0.2, 0.5, and 0.8 considered small, medium, and large, respectively. IBM SPSS statistics v.26 was used for all analyses.

A total of 58 patients (N_MY-ENDO = 20, N_Non-specific = 19, N_WL = 19) were included in the study. The CONSORT study flow diagram is shown in Fig. 2. Four patients dropped out before baseline measurement was obtained. Another 12 patients dropped out before study completion. Reasons for dropout are unknown. There were no statistically significant differences in dropout rate between the groups (P = 0.856). Nor were there any statistically significant differences regarding baseline characteristics for dropouts (N = 12) compared to completers (N = 42) except for previous use of alternative treatments (P = 0.020) as none of the dropouts had any previous experience with alternative treatments for endometriosis compared to 33.3% of completers.

Sociodemographic, clinical, and pain-related data are shown in Table 2. A significant baseline difference was found between the three study arms for the use of pain medication (P = 0.050), and the use of pain medication was therefore adjusted for in post-hoc analyses of the primary outcome [47].

Statistically significant time × group effects were found for pelvic pain intensity (P = 0.009, d = 0.23) and unpleasantness (P = 0.009, d = 0.22) (Table 3), due to pain reduction in the waitlist group. Comparable results were found in sensitivity analyses. However, when adjusting for use of pain medication in time × group interactions of the primary outcomes pelvic pain unpleasantness (P = 0.071, d = 0.59) no longer reached statistical significance (Table 3).

We also found statistically significant time × group interactions for the QoL-subscales ‘control and powerlessness’ (P = 0.019, d = 0.78), ‘emotional wellbeing’ (P = 0.003, d = 1.01) and ‘social support’ (P = 0.042, d = 0.66), and for the endometriosis-related symptoms ‘dyschezia’ (P = 0.047, d = 0.43), ‘constipation’ (P = 0.045, d = 0.47) and ‘nausea’ (P = 0.016, d = 0.47) (Table 3).

Statistically significant time × group effects were found for pelvic pain intensity (P = 0.037, d = 0.22) and pelvic pain unpleasantness (P = 0.049, d = 0.20). The sensitivity analyses testing the robustness of the results yielded comparable results (Table 3). However, when adjusting for use of pain medication in time × group interactions of the primary outcomes neither pelvic pain intensity (P = 0.144, d = 0.59) nor pelvic pain unpleasantness (P = 0.289, d = 0.43) reached statistical significance (Table 3).

Neither did we find any statistically significant time × group interactions for the secondary outcomes except for diarrhea (P = 0.035, d = 0.25) (Table 3).

With regards to the time spent on home practice, we did not find a significant difference between MY-ENDO (Mean = 22.41 min/day, SD = 20.93) and Non-specific (Mean = 22.26 min/day, SD = 15.47) in the average amount of time (min/day) spent on homework during the 10-week treatment period (U = 47.000, N_MY-ENDO = 10, N_Non-specific = 11, P = 0.605).

Statistically significant pre-post changes for all three groups (MY-ENDO, Non-specific, and WL) are found in Table 4.

When dividing participants into two groups (taking vs. not taking pain medication) independent of randomization allocation, statistically significant time × group effects were found for pelvic pain intensity (F = 11.3, P = 0.001, d = 0.29) and pelvic pain unpleasantness (F = 13.9, P < 0.001, d = 0.32). In the group taking pain medication, a significant reduction in pelvic pain intensity (P = 0.001, d = 0.32) and pelvic pain unpleasantness (P < 0.001, d = 0.44) was found, whereas in the group not taking pain medication, a significant increase in pelvic pain intensity (P = 0.039, d = 0.43) and pelvic pain unpleasantness (P = 0.048, d = 0.39) was found (Fig. 3).

Statistical analysis revealed no significant therapist x time interactions except for work ability (P = 0.021, d = 1.36) and nausea (P = 0.040, d = 0.24) and no statistically significant therapist x treatment interactions except for nausea (P = 0.029, d = 0.93).

There were no unprompted reports of any adverse events or side effects of the interventions. The results of MY-ENDO compared to WL and Non-specific compared to WL are found in Table 5.

Notably, this study has some strengths that are rarely seen in RCTs examining the effects of psychological interventions. The inclusion of a carefully matched non-specific control condition and a no-treatment control in a 3-armed RCT together with the attempt to reduce significant threats to internal validity (e.g., balancing therapist training, adherence, and competence; controlling for therapist effects; ensuring an equal treatment structure across conditions; and balancing non-specific factors) entails that the findings of this study may be highly robust.

According to the power analysis we needed 81 participants to be able to detect significant differences in the primary outcome. Despite a smaller sample size, significant differences were found between the groups for the primary outcome pelvic pain intensity and pelvic pain unpleasantness. However, these differences appeared to be driven by significant improvements in the waitlist group and may be explained by differences in the use of pain medication. This explanation was further substantiated by the results of the post-hoc analysis and the results of the experimental pain test and it cannot be ruled out that the use of pain medication might have influenced the pain results.

While the small sample size would contribute considerably to the risk of type 2 error, one should also bear in mind that a large number of statistical tests were performed in the study, and hence there is considerable risk of type 1 error. Still, some interesting significant pre-post changes were found in our data including increased workability and lower fatigue in the MY-ENDO group. However, these effects did not show statistically significant differences between the groups, and larger sample sizes are needed in future studies to answer questions about specificity.

Recruiting participants was difficult due to patients suffering from physical disabilities and lack of energy making the 3-h in-person commitment a barrier to participation. Other barriers were geographical distance and working schedules.

Future studies could try new ways to overcome these limitations for example by stratifying participants by use of pain medication and offering digitally delivered therapy to this patient group.

Standard treatment for endometriosis is primarily focused on hormonal treatment, pain medication, and surgery. However, medical treatment can induce serious side effects leading to discontinuation of treatment and recurrence of symptoms, and surgery that resolves endometriosis may not necessarily resolve pain since the extent of pain may be unrelated to the extent of disease [17].

With this rigorous three-armed RCT we have demonstrated that PIs specifically targeting endometriosis can lead to significant and large improvements in QoL and improvements in dyschezia and constipation despite an ongoing experience of severe CPP. Therefore, PIs aimed at symptom management and the improvement of QoL could be an appropriate supplement to an interdisciplinary endometriosis treatment.