Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) that evaluated the effects of the LNG-IUS on the endometrium in women with atypical endometrial hyperplasia were included.
Types of participants
women with contraindications to the LNG-IUS (acute genital tract inflammatory disease, genital bleeding of unknown aetiology, pregnancy or suspicion of pregnancy, hypersensitivity to any component of this product, congenital or acquired uterine anomaly, known or suspected breast cancer, known or suspected uterine and cervical neoplasia or unresolved or abnormal Pap smear, acute liver disease or liver tumour, etc);
women with concurrent endometrial cancer;
women with a history of any other malignancy.
Types of interventions
LNG-IUS insertion (Nova-T®, Levonova®/Mirena®, Femilis®, Fibroplant®, Mirena®).
Types of outcome measures
1. Rate of regression (complete or partial): a 'complete regression' is defined as a return of the atypical endometrium to normal, often with associated secretory glandular changes and atrophy; a 'partial regression' is defined as a change from atypical to non-atypical hyperplasia. The regression rate includes the complete and partial regression rate.
2. Adverse effects associated with hormonal problems: weight gain, acne, nausea, headache, dizziness, vomiting, depression, breast tenderness.
3. Rate of recurrence
4. Proportion of women undergoing hysterectomy (histologically indicated or non-histologically indicated)
5. Other adverse effects: bleeding (spotting, amenorrhoea, prolonged or heavy bleeding), pelvic inflammatory disease (PID), ovarian cyst, uterine myoma, device-related problems (expulsion, perforation)
6. Withdrawal from treatment because of adverse events
7. Satisfaction with treatment
8. Quality of life (QoL), measured using a scale that has been validated through reporting of norms in a peer-reviewed publication
Search methods for identification of studies
See: Cochrane Menstrual Disorders and Subfertility Group (MDSG) methods used in reviews. A comprehensive search strategy was formulated in order to identify all RCTs regardless of language. The following electronic databases were searched:
Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2012, Issue 8);
Menstrual Disorders and Subfertility Specialised Register;
MEDLINE (1948 to 2012 November);
EMBASE (1986 to 2012 November);
Chinese National Knowledge Infrastructure (CNKI) (1977 to 2012 November);
CBM (China Biomedicine Database) (1995 to 2012 November);
PsycINFO (1988 to 2012 November).
The search strategies, based on terms related to the review topic, are presented for each database in Appendices 1 to 7. The PubMed 'related articles' feature was used to check for articles related to eligible studies. The bibliographies of all relevant papers selected through this strategy were searched. Where possible, personal communications with corresponding authors and clinical experts were established to enquire about other published or unpublished relevant studies.
Searching other resources
Five major clinical trials registries (listed below) were searched for ongoing and registered trials using the following words: levonorgestrel, intrauterine device, randomized, endometrial simple hyperplasia, endometrial complex hyperplasia, endometrial atypia. We searched:
1. National Research Register;
2. Meta-register of Controlled Trials;
3. Chinese Clinical Trial Registry;
4. World Health Organization (WHO) International Clinical Trials Registry Platform;
5. Clinical Trials gov.
Grey literature was searched on the European OpenSingle (http://opensigle.inist.fr/).
The abstracts of scientific meetings were also searched on the ISI Web of Knowledge (http://isiwebofknowledge.com/).
Data collection and analysis
Selection of studies
All titles and abstracts retrieved by electronic searching were downloaded to the reference management database (EndNote), duplicates removed and the remaining references independently examined by two review authors (LL, LB). Those studies which clearly did not meet the inclusion criteria were excluded and copies of the full texts of potentially relevant references were obtained. The eligibility of retrieved papers was assessed independently by two review authors (LL, LB). Disagreements were resolved by discussion. Reasons for exclusion were documented.
Data extraction and management
Data on the characteristics of participants and interventions, study quality and endpoints were independently extracted by LL and LB using a piloted data extraction form designed according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Differences between review authors were resolved by discussion or by appeal to a third review author (ZY or LJ).
Assessment of risk of bias in included studies
We planned to assess the methodological quality using the Cochrane Collaboration's tool and the criteria specified in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We planned to assess sequence generation; allocation concealment; blinding of outcome assessors; completeness of outcome data; selective outcome reporting; and other potential sources of bias. We planned that two authors would assess these domains, with any disagreements resolved by consensus or by discussion with a third author. All judgments would be fully described. The conclusions would be presented in the 'Risk of bias' table and incorporated into the interpretation of review findings by means of sensitivity analyses.
Measures of treatment effect
We planned to use the following measures for the effect of treatment.
For dichotomous data (e.g. regression rate), the numbers of events in the control and intervention groups of each study woul be used to calculate Peto odds ratios.
For continuous outcomes (e.g. QoL scores), mean differences between treatment arms would be calculated if all studies reported exactly the same outcomes. If similar outcomes were reported on different scales (e.g. change in weight) the standardised mean differences would be calculated.
The 95% confidence intervals would be presented for all outcomes.
Unit of analysis issues
We anticipated that no unit of analysis issues would arise.
Dealing with missing data
We planned to contact trial authors to ask for any missing data. If suitable data were available, intention-to-treat analysis would be conducted. Where these were unobtainable, we planned that imputation of individual values would be undertaken for the primary outcomes only. Endometrial regression would be assumed not to have occurred in participants with an unreported outcome.
If studies reported sufficient detail to calculate mean differences but gave no information on the associated standard deviation (SD), we planned to assume the outcome had an SD equal to the highest SD from other studies within the same analysis.
For other outcomes, we planned that only the available data would be analysed. Any imputation that was undertaken would be subjected to sensitivity analysis.
Assessment of heterogeneity
The authors planned to consider whether the clinical and methodological characteristics of the included studies were sufficiently similar for meta-analysis to provide a meaningful summary. Statistical heterogeneity would be assessed by a measure of the I2 statistic. An I2 statistic greater than 50% would be taken to indicate substantial (high level) heterogeneity; 25% to 50% as moderate level; less than 25% as low level (Higgins 2002; Higgins 2003). If substantial heterogeneity was detected, possible explanations would have been explored in sensitivity analyses.
Assessment of reporting biases
In view of the difficulty in detecting and correcting for publication bias and other reporting biases, we aimed to minimize their potential impact by ensuring a comprehensive search for eligible studies and by being alert for duplication of data. If there were 10 or more studies in an analysis, we planned to use a funnel plot to explore the possibility of small study effects (a tendency for estimates of the intervention effect to be more beneficial in smaller studies).
If the RCTs were considered not appropriate for pooled analysis, only a descriptive analysis would have been conducted. Otherwise, we planned that the data from primary studies would be combined using fixed-effect models in the following comparisons.
1. LNG-IUS versus oral progestin, stratified by duration and dose:
(i) low dose and short duration;
(ii) high dose and short duration;
(iii) low dose and long duration;
(iv) high dose and long duration.
2. LNG-IUS (low dose) versus LNG-IUS (high dose), stratified by duration:
(i) short duration;
(ii) long duration.
3. LNG-IUS (short duration) versus LNG-IUS (long duration), stratified by dose:
(i) low dose;
(ii) high dose.
An increase in the odds of a particular outcome, which may be beneficial (for example endometrial regression) or detrimental (for example adverse effects), would be displayed graphically in the meta-analyses to the right of the centre-line and a decrease in the odds of an outcome to the left of the centre-line.
Subgroup analysis and investigation of heterogeneity
We planned to conduct subgroup analyses grouping the trials by the following variables.
1. Type of LNG-IUS used:
frameless LNG-IUS, releasing 14 μg LNG/day;
framed LNG–IUS, releasing 20 μg LNG/day.
2.Type of atypical endometrial hyperplasia:
Factors such as age, the duration of LNG-IUS intervention, length of follow-up, different diagnostic criteria of atypical endometrial hyperplasia, and adjusted or unadjusted analysis were to be considered in interpretation of any heterogeneity.
We planned to use sensitivity analysis to explore the influence of the following factors on effect size:
1. if eligibility was restricted to studies without high risk of bias;
2. if alternative imputation strategies were adopted;
3. if a random-effects model was adopted.