A systematic review of outcome reporting and outcome measures in studies investigating uterine-sparing treatment for adenomyosis

Adenomyosis is a common benign disease of the uterus that can be found in 20–70% of patients, depending on the characteristics of study populations ( Upson and Missmer, 2020 ). Despite its reported negative impact on quality of life (QOL), fertility and obstetric outcomes ( Harada et al. , 2019 ; Horton et al. , 2019 ; Upson and Missmer, 2020 ), data on the efficacy of treatments for adenomyosis are lacking. Systematic reviews evaluating interventions for adenomyosis have been unable to perform quantitative data-synthesis of commonly reported outcomes, such as abnormal uterine bleeding, due to the variation in outcome reporting ( de Bruijn et al. , 2017 ; Abbas et al. , 2020 ). These reviews highlighted a significant variation in both the definition and measurement of outcomes, thereby preventing useful comparison of treatment outcomes. Variations in outcome reporting and measurements also contribute to the exaggeration of treatment effects and reporting bias by omitting unfavourable data ( Duffy et al. , 2017 ). For example, there is a controversy regarding the extent to which surgery could improve fertility outcomes in patients with adenomyosis. As reporting of fertility and obstetric outcomes is highly selective, the success of treatment is interpreted differently by the authors, with the risk of being overstated ( Abbott, 2017 ; Dueholm, 2017 ). Carefully selected outcomes and outcome measures can enhance research quality, increase the relevance of research results for the people treated for a condition, and reduce research waste. There is a growing consensus that the use of standardized or ‘core’ outcome sets in clinical trials would improve research into womens’ health. Such examples are published consensus on core outcomes in endometriosis research or fertility reporting ( Duffy et al. , 2020 , 2021 ). There is currently no consensus amongst key stakeholders regarding which outcomes should be measured in trials assessing interventions for adenomyosis-related symptoms. A collection of 84 editors of women’s health journals, including the Cochrane Gynaecology and Fertility Group, have formed a consortium to support core outcome sets (COSs): the Core Outcomes in Women’s Health (CROWN) ( Khan, 2016 ). The Core Outcome Set in Adenomyosis Research (COSAR) initiative aims to develop a COS for studies investigating therapeutic interventions for adenomyosis in conjunction with the CROWN-network. As part of this work, the aim of the present review was to develop an inventory and systematically evaluate the outcomes and outcome measures reported in clinical trials investigating the treatment of adenomyosis.

T.T. receives a grant (2020083) from the South Eastern Norwegian Health Authority, which had no influence on study design, data extraction, analysis or interpretation.

The literature search identified 1364 unique citations; eight additional studies were found through searching of reference lists ( Fig. 1 ). In total, 38 studies were included in the final selection. The characteristics of the final 38 articles are listed in Table I . PRISMA flow diagram for a systematic review of outcome reporting and outcome measures in studies investigating uterine-sparing treatment for adenomyosis . Study characteristics and types of intervention of the included studies. 200 (40/40/40/40/20/20) Group 1: 3.75 mg GnRH-a Group 2: 1.88 mg GnRH-a Group 3: LNG IUS Group 4: 3.75 mg GnRH-a + LNG IUS Group 5: 1.88 mg GnRH-a + LNG IUS Group 6: San-Jie-Zhen-Tong capsules GnRH-a, gonadotropin-releasing hormone analogue; HIFU, high-intensity focused ultrasound; LNG IUS, levonorgestrel-releasing intrauterine system; TVUS, transvaginal ultrasound. We included 38 trials, reporting on data from 5175 women ( Table I ) ( Fedele et al. , 1997 ; Maia et al. , 2003 ; Hadisaputra and Anggraeni, 2006 ; Bragheto et al. , 2007 ; Kim et al. , 2007 ; Cho et al. , 2008 ; Kang et al. , 2009 ; Sheng et al. , 2009 ; Kang et al. , 2010 ; Ozdegirmenci et al. , 2011 ; Zhou et al. , 2011 ; Kelekci et al. , 2012 ; Ekin et al. , 2013 ; Liu et al. , 2014 ; Zhang et al. , 2014 ; Huang et al. , 2015 ; Lee et al. , 2015 , 2019 ; Long et al. , 2015 ; Chong et al. , 2016 ; Liu et al. , 2016 ; Park et al. , 2016 ; Hai et al. , 2017 ; Huang et al. , 2017 ; Liu et al. , 2017 ; Osuga et al. , 2017 ; Yang et al. , 2017 , 2019 ; Alizzi, 2018 ; Guo et al. , 2018 ; Jun-Min et al. , 2018 ; Li et al. , 2018 , 2020 ; Huang et al. , 2020 ; Lin et al. , 2020 ; Sun et al. , 2020 ; Kwack et al. , 2021 ; Sun et al. , 2021 ). There were six (16%) randomized controlled trials (RCTs) ( Hadisaputra and Anggraeni, 2006 ; Kang et al. , 2010 ; Ozdegirmenci et al. , 2011 ; Zhang et al. , 2014 ; Osuga et al. , 2017 ; Lin et al. , 2020 ); 13 studies (34%) were prospective, non-randomized trials, of which nine had cohorts with a single arm ( Fedele et al. , 1997 ; Bragheto et al. , 2007 ; Kim et al. , 2007 ; Cho et al. , 2008 ; Sheng et al. , 2009 ; Zhou et al. , 2011 ; Ekin et al. , 2013 ; Alizzi, 2018 ; Yang et al. , 2019 ) and four studies had two or more arms ( Kelekci et al. , 2012 ; Huang et al. , 2015 ; Yang et al. , 2017 ; Li et al. , 2018 ). There were 17 (45%) retrospective cohort studies, 12 with a single arm ( Kang et al. , 2009 ; Liu et al. , 2014 , 2016 ; Lee et al. , 2015 , 2019 ; Long et al. , 2015 ; Chong et al. , 2016 ; Park et al. , 2016 ; Hai et al. , 2017 ; Huang et al. , 2017 ; Jun-Min et al. , 2018 ; Kwack et al. , 2021 ) and five with two or more arms ( Maia et al. , 2003 ; Liu et al. , 2017 ; Guo et al. , 2018 ; Huang et al. , 2020 ; Li et al. , 2020 ). Two studies with a single cohort did not specify if the cohort was retrospective or prospective ( Sun et al. , 2020 , 2021 ). Only five studies had a low risk of bias ( Ozdegirmenci et al. , 2011 ; Zhang et al. , 2014 ; Osuga et al. , 2017 ; Yang et al. , 2017 ; Lin et al. , 2020 ) with all the other studies having an unclear or high risk of bias in at least one domain ( Supplementary Data S2 ). Common concerns in terms of risk of bias were the retrospective nature of the studies, unclear representativeness of participants, the lack of control groups, and lack of randomization. The majority of the 38 studies (84%) were conducted in Asia, of which 22 (58%) were from China ( Fig. 2 ). World map with an overview over the countries of origin for the included studies. Ten different interventions, alone or in combination, were described in at least one arm ( Table I ). We identified 203 outcomes in 29 domains, including 41 complications or adverse outcomes ( Table II and Supplementary Data S3 ). Table III shows which studies measured the most frequent outcomes. Number of outcomes reported, classified by core area and outcome domain. Outcomes could be classified in several domains but are counted once in the total. All individual outcomes are reported in Supplementary Data S3 . Outcome reporting and outcome quality scores in adenomyosis trials. 31 (82) 30 (79) 27 (71) 9 (24) 9 (24) 8 (21) 9 (24) 2 (5) 8 (21) 24 (63) 7 (18) The mean quality score for the outcomes was 3.5 ± 0.51, with scores of 2, 3 and 4 being equally frequent and accounting for 66% of the studies ( Table III ). The association between the age of the publication and its quality score was not statistically significantly ( P = 0.08). Only seven (18%) studies had a clearly defined primary outcome ( Fedele et al. , 1997 ; Maia et al. , 2003 ; Sheng et al. , 2009 ; Chong et al. , 2016 ; Liu et al. , 2017 ; Osuga et al. , 2017 ; Sun et al. , 2021 ) and 11 (29%) studies stated multiple or all reported outcomes to be the primary ( Hadisaputra and Anggraeni, 2006 ; Bragheto et al. , 2007 ; Kang et al. , 2010 ; Ozdegirmenci et al. , 2011 ; Zhou et al. , 2011 ; Kelekci et al. , 2012 ; Ekin et al. , 2013 ; Liu et al. , 2014 ; Long et al. , 2015 ; Li et al. , 2018 ; Huang et al. , 2020 ). Twenty studies (53%) did not define a primary outcome at all ( Kim et al. , 2007 ; Cho et al. , 2008 ; Kang et al. , 2009 , 2010 ; Zhang et al. , 2014 ; Huang et al. , 2015 ; Lee et al. , 2015 , 2019 ; Liu et al. , 2016 ; Park et al. , 2016 ; Hai et al. , 2017 ; Huang et al. , 2017 ; Yang et al. , 2017 , 2019 ; Alizzi, 2018 ; Guo et al. , 2018 ; Jun-Min et al. , 2018 ; Li et al. , 2018 ; Lin et al. , 2020 ; Sun et al. , 2020 ; Kwack et al. , 2021 ). Most of these studies described the aims in non-specific terms, such as ‘clinical efficiency’, but without defining the terms ‘clinical’ or ‘efficiency’. Most studies did not provide a justification for the chosen outcomes. Only three (8%) of the studies provided a sample size calculation based on an outcome ( Liu et al. , 2014 ; Osuga et al. , 2017 ; Ozdegirmenci et al. , 2011 ), and none of the other 35 studies provided a post hoc estimation of statistical power. The majority of the studies provided an outcome measure for the main outcomes. The most common time points for measuring outcomes were at 3, 6 and 12 months after the intervention, as described in 14 studies ( Fedele et al. , 1997 ; Kang et al. , 2009 ; Sheng et al. , 2009 ; Huang et al. , 2015 ; Lee et al. , 2015 , 2019 ; Long et al. , 2015 ; Liu et al. , 2016 ; Li et al. , 2018 ; Yang et al. , 2019 ; Lin et al. , 2020 ; Sun et al. , 2020 ). However, only 19 studies reported on all their outcomes at all the predetermined time points according to the described methods. Several studies provided only a visual or summarized outcome reporting for at least one of the outcomes, without values, SD or 95% CI ( Maia et al. , 2003 ; Bragheto et al. , 2007 ; Ozdegirmenci et al. , 2011 ; Ekin et al. , 2013 ; Alizzi, 2018 ). The various outcome measures and interpretation of the most common outcomes, namely dysmenorrhoea, menstrual volume/menorrhagia and QOL, are presented in Supplementary Data S4 . There were 14 different measuring tools and interpretations for dysmenorrhoea, and 17 for menstrual blood loss ( Supplementary Data S4 ). Only a minority of the studies reported how outcome measurement was performed, for example if the patients were instructed to use the Pictorial Blood Loss Assessment Chart (PBLAC), or if questionnaires were filled out by the patient or by the doctor, for example by telephone interview. Uterine volume was reported as an outcome in 27 (71%) studies ( Table III ). In most cases, the volume was measured using transvaginal ultrasound. Only three studies reported if the measurement included the cervix and how uterine length was measured. None of the papers provided a clinical justification for this outcome. Eight (21%) studies followed a classification when registering and reporting adverse events ( Zhou et al. , 2011 ; Liu et al. , 2016 ; Hai et al. , 2017 ; Li et al. , 2018 ; Lin et al. , 2020 ), while the others did not report how complications or side effects were registered or reported.

T.T. receives fees from General Electrics for lectures on ultrasound independently of this project.

This review is registered with the International Prospective Register of Systematic Reviews (PROSPERO; CRD42020177466) and reports in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement ( Liberati et al. , 2009 ). COSAR is registered with the Core Outcome Measures in Effectiveness Trials (COMET) initiative (registration number 1649). No approval from the institutional review board or ethics committee was sought owing to the nature of this work. The literature search was performed with support from a trained medical librarian. The electronic databases Medline, Embase and Cochrane Register of Controlled Trials were searched using the terms ‘adenomyosis’ and ‘treatment’, as well as a range of treatment-specific key words that were identified through a pilot search. The electronic search strategy is presented in Supplementary Data S1 . The time range was from 1950 until February 2021. In addition, the reference lists of included articles and identified reviews on the topic were scanned, and manually searched for further studies. Studies reporting on any uterus-sparing intervention to treat adenomyosis, of any study design, both prospective and retrospective, were eligible for inclusion. Inclusion criteria were a study population comprising ≥20 women, a clear description of the modality and diagnostic criteria used to diagnose adenomyosis, a clear description of the intervention, follow-up time ≥6 months, loss to follow-up <20% and English language. Exclusion criteria included data presented in short communications, reviews, letters to the editors and congress abstracts. Studies with experimental design (e.g. performed on tissue samples or looking exclusively at molecular markers) or fundamental design flaws (unclear intervention) were also excluded. Two researchers (T.T. and M.O.) independently screened the retrieved titles and abstracts using the Rayyan application ( Ouzzani et al. , 2016 ). Potentially eligible studies were retrieved in full text for the assessment of their eligibility. Their methodological quality was independently assessed by two researchers (T.T. and M.O.) using the Evidence project risk of bias tool ( Kennedy et al. , 2019 ). The signalling questions were answered with ‘yes’, ‘no’, ‘not reported’ or ‘not applicable’. At each step, conflicting decisions were resolved through discussion. The quality of outcome reporting was assessed by T.T. and M.O. using the Management of Otitis Media with Effusion in Cleft Palate (MOMENT) criteria ( Harman et al. , 2013 ), as described previously ( Hirsch et al. , 2016 ; Pergialiotis et al. , 2018 ). One point was given for each of the following six domains: whether a primary outcome was clearly stated; whether the primary outcome was clearly defined for reproducible measures; whether the secondary outcomes were clearly stated; whether the secondary outcomes were clearly defined for reproducible measures; whether the authors explain the choice of outcome; whether the methods that were used were appropriate to enhance the quality of measures. We awarded a point for stating a primary outcome even if multiple primary outcomes were described. We did not award a point for a clear definition of the primary outcome if no primary outcome was stated. When most of the secondary outcomes were clearly defined for reproducible measures, we awarded a point even if not all secondary outcomes were clearly described. We did not award a point under the last domain if the study was retrospective and if it was not described how the outcomes were documented. The data extraction was performed by T.T. and M.O. First author, year of publication, country of origin, study design, number of participants with adenomyosis and type of intervention were noted. We also recorded whether a sample size calculation was carried out. Outcomes were documented as primary or secondary outcomes, according to how they were classified in the Materials and method (M&M) section. Generic statements found in the title or abstract, such as ‘efficacy’ or ‘clinical effect’, that were not further specified in the manuscript, were not regarded as (primary) outcomes. Outcomes that were described in the result section or the discussion, but not defined in the M&M section as outcomes, were still included as secondary outcomes in the synthesis. As this was a recurring problem, the authors found that it would not reflect the reporting of outcomes if only the outcomes mentioned in the M&M section were included. The outcome measure and, if given, the definition was recorded, as well as the time points of outcome measuring and reporting. The outcomes were classified to core areas and domains according to a taxonomy recommended by COMET ( Dodd et al. , 2018 ). Composite outcomes, such as QOL, were reported with each item classified to the respective domain. Results from this review are presented as percentages. Means and SD are calculated for normally distributed data. Distribution of data within the samples was assessed by analysing skewness and kurtosis. Associations between date of publication and quality of outcomes were analysed using linear regression. Probability values were rounded to two decimal places, with the exception of P < 0.001. Data analysis was performed using Microsoft Excel software (Version 2102, Microsoft Corporation, Redmont, USA).

Conception of work, study design: T.T., M.O., J.N., M.H., D.J.; Literature search, quality assessment, data extraction: T.T., M.O.; Interpretation of results, drafting of manuscript and approval of final version: T.T., M.O., J.N., M.H., D.J.

In this review, we identified substantial heterogeneity in outcome reporting in studies evaluating interventions for the treatment of adenomyosis-associated symptoms. Only six studies that met the inclusion criteria were RCTs. A small proportion of studies provided a clear primary outcome, or a sample size/power calculation and incomplete outcome reporting was common. The most frequently reported outcome was dysmenorrhoea, which was reported in 31 (82%) of studies. There were 14 different outcome measures used to assess dysmenorrhoea, with different visual analogue scales being the most frequently used. Most researchers attempted to use validated measuring tools for the main outcomes. Outcomes identified through this systematic review of published studies reflect outcomes that healthcare professionals or researchers have chosen to select, collect and report. These outcomes are largely focused on menstrual symptoms and uterine volume with only a minority of studies reporting outcomes relating to fertility or QOL. Dyspareunia, chronic pelvic pain and other pain-related outcomes were measured in very few studies. A review of outcome reporting in endometriosis, a condition which has significant overlap of both patient population and associated symptoms with adenomyosis, reported eight different pain-related outcomes ( Hirsch et al. , 2016 ). This difference may be explained by the wider variation in pain symptoms that women with endometriosis may experience. Patient-centredness is defined by ‘health care which takes into account the preferences and aspirations of individual service users’ and is one of the dimensions of quality of care ( World Health Organization, 2006 ). Focus on outcomes such as satisfaction with the treatment or health-related QOL in clinical studies reflect patient-centredness. These outcomes are important for patients to make informed decisions about different treatment options. However, those type of outcomes were reported infrequently. Patient-centredness is also reflected in outcomes being important to women, and we assume that dysmenorrhoea and heavy menstrual bleeding are amongst those. These were frequently reported. A challenge is, however, that many of those outcomes are by nature patient-reported and can be difficult to measure and replicate ( Magnay et al. , 2020 ). In addition, there is no disease-specific QOL measurement for adenomyosis, which makes the QOL results reported by other tools less reliable for this group of women. In contrast, imaging outcomes, such as the uterine size, were reported in the majority of the studies. It remains unclear whether this surrogate marker of disease severity is associated with clinical symptoms in women with adenomyosis. This suggests that uterine volume may be an outcome of convenience rather than clinical significance. Similarly, serum levels of CA-125 were commonly reported without a clear clinical justification. The use of these outcomes suggests a lack of patient involvement and input in adenomyosis research. Reporting fertility and pregnancy outcomes is highly relevant for adenomyosis trials, as many women with adenomyosis find it difficult to fall pregnant. Unfortunately, those outcomes are only reported sporadically. Seemingly random reporting on pregnancies or live birth, as well as leaving it unclear how many women in a study sample tried to get pregnant, possibly augments the effect of certain interventions on fertility outcomes. The lack of well-designed randomized trials in adenomyosis exacerbates the difficulty in determining which treatments are more effective and better to use. Outcome reporting variation seen in this study prohibits the combination, comparison, and synthesis of research data into meta-analysis. This limits the ability of research to inform clinical care guidelines and progress the specialty. This variation in outcome reporting may reflect selective outcome reporting and outcome reporting bias. This has been identified to be a major limitation in Cochrane systematic reviews. Following adjustment for outcome reporting bias, 19% of all their reviews would no longer have statistically significant treatment effects while 26% of their reviews would have over-estimated the treatment effects by 20% ( Chalmers and Glasziou, 2009 ). This represents a large area of potentially avoidable research waste. Three key areas of avoidable research waste are related to outcome reporting. These include: important outcomes are not assessed; research studies fail to consider outcomes in the context of previously published research; and over half of all outcomes collected are never reported in the final publication ( Chalmers and Glasziou, 2009 ). The strengths of this study include its originality, and the robust search strategy and design. The review process was performed by two independent researchers, to prevent bias. This is the first systematic review to describe outcome reporting variation in adenomyosis studies. This review is not without limitations. We included studies written in English only. Four studies published in Chinese had to be left out, but no further papers were excluded for language reasons. We included studies of differing methodological design, limiting the ability to compare and contrast the study quality. Most studies were retrospective and had a high risk of bias, which could have influenced the quality and type of reported outcomes. We considered limiting the inclusion criteria to high quality RCTs or prospective observational studies, however this would have limited the number of outcomes and not accurately reflected current outcome reporting. This review highlights the importance of the recent initiatives to enhance research methodology including the CONSORT statement, the AllTrials initiative and the CROWN initiative. These initiatives aim to ensure that all prospectively registered RCTs are published regardless of their findings, eliminating publication bias from studies that are withheld where there is negative or no effect demonstrated ( Song et al. , 2010 ). The development and use of a collection of widely agreed and well-defined outcomes, termed a COS, would help to address selective outcome reporting bias and facilitate the production of comparable data for improved evidence-based patient care. This progressive approach to standardize research methodology is supported by national and international stakeholders. The World Health Organization, the National Institutes of Health and the Cochrane Collaboration are committed to supporting, developing and implementing COSs. There is a clear and evident need for the development of a COS together with recommendations for uniform outcome measures in adenomyosis research and it is important that people with adenomyosis participate in this process. This systematic review is the first step in the development of a minimum data set to be selected, collected, and reported in all future clinical trials on adenomyosis. It will be developed by the COSAR initiative with reference to methods described by the COMET initiative ( Williamson et al. , 2017 ). The development of a COS for therapeutic interventional studies in adenomyosis research will enhance the quality of adenomyosis research facilitating a more patient-centred approach to care.

The data underlying this article are available in the article and in its online supplementary material.

Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file.