Saturday, November 1, 2014 11:22

Benign Gynaecological Disease

Tagged with:
Posted by on Wednesday, September 8, 2010, 2:45
This news item was posted in B, Gynaecological category and has 2 Comments so far.

Surgical approach to hysterectomy for benign gynaecological disease

Abstract

Benign Gynaecological Disease
Benign Gynaecological Disease

Background
The three approaches to hysterectomy for benign disease are abdominal hysterectomy (AH), vaginal hysterectomy (VH), and laparoscopic hysterectomy (LH). Laparoscopic hysterectomy has three further subdivisions depending on the part of the procedure performed laparoscopically.

Objectives
To assess the most beneficial and least harmful surgical approach to hysterectomy for women with benign gynaecological conditions.

Search strategy
We searched the Cochrane Menstrual Disorders and Subfertility Group Specialised Register of controlled trials (15 August 2008), CENTRAL (The Cochrane Library 2008, Issue 3), MEDLINE (1950 to August 2008), EMBASE (1980 to August 2008), Biological Abstracts (1969 to August 2008), the National Research Register, and relevant citation lists.

Selection criteria
Only randomised controlled trials comparing one surgical approach to hysterectomy with another were included.

Data collection and analysis
Independent selection of trials and data extraction were employed following Cochrane guidelines.

Main results
There were 34 included studies with 4495 women. The benefits of VH versus AH were speedier return to normal activities (mean difference (MD) 9.5 days), fewer febrile episodes or unspecified infections (odds ratio (OR) 0.42), and shorter duration of hospital stay (MD 1.1 days). The benefits of LH versus AH were speedier return to normal activities (MD 13.6 days), lower intraoperative blood loss (MD 45 cc), a smaller drop in haemoglobin (MD 0.55 g/dl), shorter hospital stay (MD 2.0 days), and fewer wound or abdominal wall infections (OR 0.31) at the cost of more urinary tract (bladder or ureter) injuries (OR 2.41) and longer operation time (MD 20.3 minutes). The benefits of LAVH versus TLH were fewer febrile episodes or unspecified infection (OR 3.77) and shorter operation time (MD 25.3 minutes). There was no evidence of benefits of LH versus VH and the operation time (MD 39.3 minutes) as well as substantial bleeding (OR 2.76) were increased in LH. For some important outcomes, the analyses were underpowered to detect important differences or they were simply not reported in trials. Data were absent for many important long-term outcome measures.

Authors’ conclusions
Because of equal or significantly better outcomes on all parameters, VH should be performed in preference to AH where possible. Where VH is not possible, LH may avoid the need for AH however the length of the surgery increases as the extent of the surgery performed laparoscopically increases. The surgical approach to hysterectomy should be decided by the woman in discussion with her surgeon in light of the relative benefits and hazards.

Plain language summary

Surgical approach to hysterectomy for benign gynaecological diseases
Abdominal hysterectomy involves removal of the uterus through an incision on the lower abdomen. Vaginal hysterectomy involves removal of the uterus via the vagina, with no abdominal incision. Laparoscopic hysterectomy involves ‘keyhole surgery’ with small incisions on the abdomen. In laparoscopic hysterectomy, the uterus is removed with the aid of a surgical telescope (laparoscope) inserted through the umbilicus (belly button) and instruments inserted through two or three further keyholes. Laparoscopic hysterectomy may be further subdivided depending on the extent of the surgery performed laparoscopically compared to that performed vaginally. More recently, laparoscopic hysterectomy can be performed with the use of a so-called robot which is operated from a distance by the surgeon.

Vaginal hysterectomy should be performed in preference to abdominal hysterectomy, where possible. This review found that vaginal hysterectomy meant quicker return to normal activities, fewer infections and episodes of raised temperature after surgery, and a shorter stay in hospital compared to abdominal hysterectomy.

Laparoscopic hysterectomy meant quicker return to normal activities, less blood loss and a smaller drop in blood count, a shorter stay in hospital, and fewer wound infections and episodes of raised temperature after surgery compared to abdominal hysterectomy, but laparoscopic hysterectomies have a greater risk of damaging the bladder or ureter (the tube leading to the bladder from the kidney) and are longer operations.

No benefits were found for laparoscopic versus vaginal hysterectomy. Laparoscopic hysterectomies are longer operations associated with a higher rate of substantial bleeding.

The authors concluded that vaginal hysterectomy should be performed in preference to abdominal hysterectomy, where possible. Where vaginal hysterectomy is not possible, a laparoscopic approach may avoid the need for an abdominal hysterectomy. Risks and benefits of different approaches may however be influenced by the surgeon’s experience. More research is needed, particularly to examine the long-term effects of the different types of surgery.

Description of the condition
Hysterectomy is the surgical removal of the uterus and is the most frequently performed major gynaecological surgical procedure with millions of procedures performed annually throughout the world (Garry 2005). Hysterectomy can be performed for benign and malignant indications. Approximately 90% of hysterectomies are performed for benign conditions, such as fibroids causing abnormal uterine bleeding (Flory 2005).

The first reported elective hysterectomy was performed through a vaginal approach by Conrad Langenbeck in 1813. The first elective abdominal hysterectomy, a subtotal operation (where the cervix was conserved), was performed by Charles Clay of Manchester in 1863 (Sutton 1997). These approaches remained the only two options until the latter part of the 20th century. The first laparoscopic-assisted vaginal hysterectomy (LAVH) was performed by Harry Reich in 1989 (Reich 1989). He also reported the first total laparoscopic hysterectomy (TLH) in 1993.

Description of the intervention
Approaches to hysterectomy may be broadly categorised into three options, abdominal hysterectomy (AH); vaginal hysterectomy (VH); and laparoscopic hysterectomy (LH) where at least some of the operation is conducted laparoscopically (Garry 1994).

The AH has traditionally been the surgical approach for gynaecological malignancy, when other pelvic pathology is present such as endometriosis or adhesions, and in the context of an enlarged uterus. It remains the ‘fallback option’ if the uterus cannot be removed by another approach.

The vaginal approach (VH) was originally used only for prolapse but has become more widely utilised for menstrual abnormalities such as dysfunctional uterine bleeding (DUB), when the uterus is a fairly normal size. Compared to AH, VH was (and still is) regarded as less invasive and seemed to have the advantages of fewer blood transfusions, less febrile morbidity (fever), and less risk of injury to the ureter, but the disadvantages are more bleeding complications and greater risk of bladder injury (Harris 1996).

The term ‘laparoscopic hysterectomy’ (LH) usually refers to a hysterectomy where at least part of the operation is undertaken laparoscopically (Garry 1994). This approach requires general laparoscopic surgical expertise. The proportion of hysterectomies performed by LH has gradually increased and, although the surgery tends to take longer, its proponents argue that the main advantages are the possibility to diagnose and treat other pelvic diseases such as endometriosis, to carry out adnexal surgery including the removal of the ovaries, the ability to secure thorough intraperitoneal haemostasis (direct laparoscopic vision enables careful sealing of bleeding vessels at the end of the procedure), and a more rapid recovery time from surgery compared to AH (Garry 1998). More recently, three sub categorisations of LH have been described (Reich 2003), as follows.

Laparoscopic assisted vaginal hysterectomy (LAVH) is where part of the hysterectomy is performed by laparoscopic surgery and part vaginally, but the laparoscopic component of the operation does not involve division of the uterine vessels.

Laparoscopic hysterectomy (which we will abbreviate to LH(a)) is where the uterine vessels are ligated laparoscopically but part of the operation is performed vaginally.

Total laparoscopic hysterectomy (TLH) is where the entire operation (including suturing of the vaginal vault) is performed laparoscopically and there is no vaginal component except for the removal of the uterus. TLH requires the highest degree of laparoscopic surgical skill. It has been unclear whether TLH offers any benefit over other forms of laparoscopic hysterectomy.

A total hysterectomy is the removal of the entire uterus including the cervix. When the cervix is not removed this is known as a subtotal or supracervical hysterectomy. Subtotal hysterectomies are most easily performed abdominally or laparoscopically, although it is possible to conserve the cervix in a VH or LAVH (Lethaby 2006).

In common with the overall hysterectomy rate, the proportion of hysterectomies currently being performed by different approaches varies markedly across countries, within countries, and even between individual surgeons working within the same unit. The surgical approach taken at hysterectomy continues to depend upon the experience and biases of the surgeon (Johns 1995). Each gynaecologist will have different indications for the approach to hysterectomy for benign disease, based largely on their own array of surgical skills and the patient characteristics such as uterine size and descent, extra-uterine pelvic pathology, previous pelvic surgery, and other features such as obesity, nulliparity, and the need for oophorectomy. Even though vaginal hysterectomy has been widely considered to be the operation of choice for abnormal uterine bleeding, the VALUE study has shown that, in 1995 in the UK, 67% of the hysterectomies performed for this indication were abdominal hysterectomies (Maresh 2002). Previous caesarean section, for example, is often considered to be a contraindication for vaginal hysterectomy. However, this is not supported by evidence as analysis of cumulative data of four studies available on the subject did not find a significant difference in complication rates in hysterectomy patients following caesarean section (Agostini 2005). Mäkinen 2001 reported a prospective study on the learning curve in 10,110 hysterectomies for benign indications, of which 5875 were abdominal, 1801 were vaginal, and 2434 were laparoscopic hysterectomies. As far as injuries to adjacent organs were concerned, the surgeons’ experience significantly correlated inversely with the occurrence of urinary tract injuries in laparoscopic hysterectomy and the occurrence of bowel injuries in vaginal hysterectomy. Encouraging vaginal surgery amongst gynaecologists has been shown to be an effective method of increasing vaginal hysterectomy rates. Finland had a vaginal hysterectomy rate as low as 7% in the 1980s. Following annual meetings on gynaecological surgery where vaginal and laparoscopic surgery were encouraged, and individual training provided, the vaginal hysterectomy rate increased to 39% in 2004 (Brummer 2008). In the same period of time the ureter injuries decreased, which represents an impressive national learning curve.

How the intervention might work
Injuries to adjacent organs are of concern in hysterectomy and their rates of occurrence differ with the various approaches to hysterectomy and surgical experience level (Mäkinen 2001). Furthermore, operation times differ with the different approaches to hysterectomy. In general it is presumed that the laparoscopic approach is followed by a quicker recovery as compared with open surgery.

Apart from the surgical approach to hysterectomy, other aspects of the surgical technique may have an effect on the outcome of surgery. Examples of this include total versus subtotal (where the cervix is not removed) hysterectomy (Lethaby 2006); Doderlein VH or LAVH versus standard VH or LAVH; techniques to support the vaginal vault; bilateral elective oophorectomy versus ovarian conservation (Orozco 2008); and other strategies used mainly by those conducting laparoscopic surgery with the aim of reducing the likelihood of complications, including the use of vaginal delineators, rectal probes, and illuminated ureteric stents. These other aspects are not be within the scope of this review (other than for assessing trial quality), which will focus simply on benefits and harms of the different surgical approaches.

Why it is important to do this review
It was interesting to note that in 1998 there was not a single randomised controlled trial (RCT) comparing AH and VH (Garry 1998). The introduction of the newer approaches to hysterectomy (LAVH, LH(a) and TLH) has stimulated a much greater interest in the proper scientific evaluation of all forms of hysterectomy. The findings of various randomised controlled trials are summarised in this systematic review.

Benign Gynaecological Disease
Benign Gynaecological Disease

Objectives

The aim of the review was to assess the most beneficial and least harmful surgical approach to hysterectomy when considering abdominal hysterectomy (AH), vaginal hysterectomy (VH), and laparoscopic hysterectomy (LH) for women with benign gynaecological conditions.

Methods

Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) where one surgical approach to hysterectomy was compared with another.

Types of participants
Inclusions: women undergoing hysterectomy for benign disease (including uterine fibroids).
Exclusions: women with gynaecological cancer. Where trials included both women with benign and malignant disease, authors were requested for a breakdown of data in order to include only women with benign disease. Trials were excluded if this information was not forthcoming. There were no such trials.

Dropouts were defined as cases where: participation was refused or hysterectomy was cancelled after randomisation, the assigned procedure was refused, or randomised cases were excluded from analysis by the researchers. Losses to follow up were not regarded as dropouts.

Types of interventions
The surgical approach to removal of the uterus where at least one approach was compared with another. Approaches were, for example, AH, VH, and LH.

AH involves removal of the uterus through an incision on the lower abdomen. VH involves removal of the uterus via the vagina, with no abdominal incision. The distinction between the subcategories of LH was made based on whether ligation of the uterine vessels was undertaken laparoscopically and whether suturing of the vaginal vault was undertaken vaginally (see Table 1). Thus LH was further subdivided in the analysis into LAVH (where the laparoscopic component did not involve ligation of the uterine vessels), LH(a) (where the uterine vessels were ligated laparoscopically but there was still some vaginal component), TLH (where the entire hysterectomy was completed laparoscopically with no vaginal component other than the removal of the uterus), and non-categorisable LH (where there was insufficient information or the types of LH were too heterogeneous to otherwise sub categorise). There are two other classifications of LH (Nezhat 1995; Richardson 1995) and these are summarised in Table 2 and Table 3.

Subtotal versus total hysterectomy is the scope of another Cochrane review (Lethaby 2006) and trials making this comparison were excluded from the present review. Trials evaluating different surgical approaches to subtotal hysterectomy were also excluded. However, if a minority of the trial women had a subtotal hysterectomy and the comparison was made between any of the three approaches outlined above then the trial was included.

LH subcategories: LAVH is where the laparoscopic component does not involve ligation of the uterine vessels; LH(a) is where the uterine vessels are ligated laparoscopically but there is still some vaginal component; and TLH is where the entire hysterectomy is completed laparoscopically with no vaginal component other than the removal of the uterus. Non-categorisable LH was where there was insufficient information, or the types of LH were too heterogeneous to otherwise sub categorise.

Types of outcome measures
The following outcome measures were defined as the primary outcomes (Johnson 2005b; Kluivers 2008).

Primary outcomes

Return to normal activities

Satisfaction and quality of life

Intra-operative visceral injury
°

Bladder injury
°

Ureter injury
°

Urinary tract (bladder or ureter) injury
°

Bowel injury
°

Vascular injury

Major long-term complications
°

Fistula
°

Pelvi-abdominal pain
°

Urinary dysfunction
°

Bowel dysfunction
°

Pelvic floor condition (prolapse)
°

Sexual dysfunction

Secondary outcomes

Operation time

Other intra-operative complication
°

(Sequelae of) bleeding

Substantial bleeding

Haemoglobin or haematocrit drop

Transfusion

Pelvic haematoma
°

Unintended laparotomy for approaches not involving routine laparotomy

Short-term outcomes and complications
°

Length of hospital stay
°

Infections

Vaginal cuff

Abdominal wall or wound

Urinary tract infection

Febrile episodes or unspecified infections
°

Thromboembolism

Costs

Note: data on the cost of treatment were sought but it was intended to describe these data qualitatively and not to include the information in the meta-analysis since ‘cost’ could be defined differently in different studies depending upon whether studies incorporate the cost of sequelae. Different healthcare systems could produce markedly different results.

Search methods for identification of studies
The Trials Search Co-ordinator of the Cochrane Menstrual Disorders and Subfertility Group was involved in the definition of search terms and the searches.

Electronic searches
The search for trials will be repeated every two years and the review updated if new trials are found.

There were no language restrictions applied.

See the following appendices Appendix 1, Appendix 2; Appendix 3 and Appendix 4 for the strategies used for the electronic data bases searched.

The National Research Register (NRR) is a register of ongoing and recently completed research projects funded by or of interest to the United Kingdom’s National Health Service, as well as entries from the Medical Research Council’s Clinical Trials Register, and details on reviews in progress collected by the NHS Centre for Reviews and Dissemination. The register was searched for any trials with the following keywords.
1. Hysterectomy
2. Abdominal
3. Vaginal
4. Laparoscopic assisted
5. Laparo-vaginal
6. Laparoscopic
7. 1 and 2 or 3 or 4 or 5 or 6

The Clinical Trials register, a registry of federally and privately funded US clinical trials, was searched for the same keywords.

Searching other resources
The citation lists of relevant publications, review articles, abstracts of scientific meetings, and included studies were also searched.

Data collection and analysis
Selection of studies
The selection of trials for inclusion in the initial review was performed by at least two of four review authors (ET, EC, AL, NJ) after employing the search strategy described previously. Selection of trials for the update was performed by two different review authors (TN, KK). Differences of opinion were resolved by consensus after consultation with one or two other review authors.

Trials were excluded from the review if they made comparisons other than those specified above. These trials were detailed in the table ‘Characteristics of excluded trials’.

Data extraction and management
Data extraction

The following data were collected from the included studies.

Trial characteristics
°

Method of randomisation, in order of preference, as follows:

third party randomisation, for example by pharmacy, computer, or telephone;

true randomisation by carer, for example by opaque numbered envelope or register;

not stated.
°

Study design:

blinding;

duration of follow up;

type of follow up.
°

Size of study:

number of women recruited;

number of women randomised;

number of women excluded;

number of women withdrawn and lost to follow up;

number of women analysed.
°

Study setting:

single centre or multicentre;

location;

timing and duration;

source of funding stated or not.
°

Analyses:

whether a power calculation was performed and adhered to;

whether ‘intention-to-treat’ analysis was performed by authors, was possible from the data but not performed by authors, not possible or uncertain.
°

Criteria for hysterectomy:

indications specified;

data broken down by indications for hysterectomy.

Characteristics of the study participants
°

Baseline characteristics:

age;

parity;

indication for hysterectomy;

investigative work up, for example pelvic ultrasound scan, endometrial sampling;

previous treatments;

exclusion criteria.
°

Treatment characteristics:

pre-operative preparation, for example pre-operative medical treatment;

level of training of surgeons.

Interventions
°

Total or subtotal hysterectomy
°

Subcategory in case of LH (i.e. LAVH, LH(a), and TLH)
°

Use of technique to support the vaginal vault
°

Proportion undergoing bilateral elective oophorectomy versus ovarian conservation
°

Other strategies to reduce the likelihood of complications
°

Absence of co-interventions in treatment and control groups
°

If the trial compared a surgical approach performed by one (group of) surgeon(s) with another surgical approach performed by a second (group of) surgeon(s)

Outcomes
°

Operating time
°

Immediate complications of surgery

Surgical injury:

urinary tract (bladder or ureter) injury;

bladder injury;

ureter injury;

bowel injury;

vascular injury.

Bleeding

Unintended laparotomy for approaches not involving routine laparotomy
°

Short-term outcomes

Pain

Sequelae of bleeding:

haemoglobin/haematocrit drop;

transfusion;

pelvic haematoma.

Infection:

vaginal cuff;

abdominal wall or wound;

urinary tract infection (UTI);

febrile episodes or unspecified infection.

Thrombo-embolism

Perioperative mortality
°

Recovery from surgery

Length of hospital stay

Return to normal activities
°

Long-term outcomes

Fistula

Pelvi-abdominal pain

Urinary dysfunction

Bowel dysfunction

Pelvic floor condition (prolapse)

Sexual dysfunction

Satisfaction, quality of life
°

Costs

Data management
All data for the meta-analysis were extracted independently by at least two review authors (from ET, EC, AL, NJ, TN, KK). Differences of opinion were resolved by consensus after consultation with another review author. Additional information on trial methodology or actual original trial data was sought from the corresponding authors of trials in which the eligibility criteria were apparently met: when aspects of methodology were unclear, or where data were in a form unsuitable for meta-analysis. Reminder correspondence was sent if a reply was not received within four weeks, and again at the time of updating the review.

Assessment of risk of bias in included studies
See Characteristics of included studies; Figure 1; Figure 2
Included studies were assessed independently by three review authors (ET, AL, KK) for the following quality criteria and methodological details. This information is presented in a table describing the included studies and provides the context for assessing the reliability of results.

Methodological quality summary: review authors’ judgements about each methodological quality item for each included study.

Methodological quality graph: review authors’ judgements about each methodological quality item presented as percentages across all included studies.

The risk of bias table summarises the data on the randomisation and allocation process, blinding, strategy in the case of dropouts, pre-definition of outcome measures, and eventual obvious methodological problems of the included studies. In the table, ‘Yes’ represents a judgement of good quality, ‘Unclear’ denotes that the issue was not reported (or in case of allocation by sealed opaque envelopes could not be judged), and ‘No’ represents a risk of bias. For an adequate sequence generation, the sequence of randomisation had to be generated beyond the influence of the researchers. For adequate allocation concealment, the sequence of randomisation had to be unknown to the researcher until after the randomisation. Sealed opaque envelopes were judged as ‘Unclear’. Since blinding of the surgeon was impossible in hysterectomy techniques, blinding was judged as ‘Yes’ in the case of an adequate attempt to blind the patient. For a judgement ‘Yes’ with regard to incomplete data, dropouts had to be included in the data analysis wherever possible. A study was judged as free of selective reporting where the outcome measures were obviously pre-defined, that is the primary outcome was defined or a sample size calculation had been performed for one of the outcome measures. Data on differences in the experience of surgical teams for different procedures and funding from pharmaceutical or surgical instrumentation companies were collected to assess other possible sources of bias.

Measures of treatment effect
Statistical analysis was performed in accordance with the guidelines from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). The data were analysed using an intention-to-treat model, where data were available.

Dichotomous data were expressed as odds ratios with 95% confidence intervals and combined for meta-analysis with RevMan software using the Peto-modified Mantel-Haenszel method. An increase in the odds of a particular outcome is displayed graphically in the meta-analyses to the right of the centre line, and a decrease in the odds of an outcome is displayed graphically to the left of the centre line.

Continuous data were combined for meta-analysis with RevMan software using the mean difference (MD) with 95% confidence interval (CI). The mean and standard deviation (SD) were used when available or calculated from 95% CIs. When only the median and (interquartile) ranges were reported, or when measures of variation were missing, these results were presented as descriptive data in a separate table. Outcome variables that were reported only graphically were not included in the study.

Statistical heterogeneity between the results of different studies was examined by inspecting the scatter in the data points on the graphs, the overlap in their CI and, more formally, by checking the results of Chi2 tests and I2 statistics. The outcomes were pooled statistically where no clinical heterogeneity was apparent. A fixed-effect model was used where statistical heterogeneity was absent. Where statistical heterogeneity was apparent after pooling of data, this was noted and statistically significant results interpreted cautiously after further analysis using a random-effects statistical model.

Dealing with missing data
The included studies were assessed for number of women lost to follow up and exclusions from analysis after randomisation (dropouts).

Missing variables were not imputed for meta-analysis.

Assessment of heterogeneity
Where statistical heterogeneity was apparent after pooling of data, this was noted and statistically significant results were interpreted cautiously after further analysis using a random-effects statistical model.

Assessment of reporting biases
The included studies were assessed for pre-defined primary outcome measures or a power calculation, or both, to assess reporting bias.

Data synthesis
A fixed-effect model was used to calculate a pooled estimate of effect in meta-analyses. If significant statistical heterogeneity was confirmed by the Chi2 test (P < 0.1) and the I2 statistic (I2 > 50%) it was planned to use a random-effects model.

Sensitivity analysis
A sensitivity analyses was performed to examine the stability of the results in relation to the following factors.

Exclusion of trials that were judged as ‘unclear’ or ‘no’ with regard to adequate sequence generation in the risk of bias table.

Exclusion of trials comparing a surgical approach performed by one surgeon (or group of surgeons) with another surgical approach performed by a second (group of) surgeon(s).

The effect of analysing studies of LH subcategories compared to studies of LH pooled as an overall category.

Results

Description of studies
See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of ongoing studies.

Fifty-five trials were identified. Nine of these were initially identified as published abstracts from conference proceedings. The first authors of these studies were contacted in an attempt to obtain details that were not reported; two studies were included (Darai 2001; Miskry 2003) and two excluded (Møller 2001; Park 2003). Five studies that had been listed as ‘Studies awaiting assessment’ in the first publication of the review have been excluded in the present update. Either no replies from the authors were received to our repeated request for more information (Davies 1998; Pabuccu 1996; Petrucco 1999) or the women had already been included in another study (Cucinella 2000; Hahlin 1994). Fourteen further studies were excluded from the review; the reasons for their exclusion are listed in the ‘Characteristics of excluded studies’ table. The authors were able to extract data from the remaining 34 trials of which: three compared VH versus AH (Benassi 2002; Miskry 2003; Silva Filho 2006); 19 compared LH versus AH (including one LH-BSO versus AH-BSO (Raju 1994) and one LAVH versus minilaparotomy AH (Muzii 2007)); six compared LH versus VH (Agostini 2006; Darai 2001; Morelli 2007; Richardson 1995; Soriano 2001; Summitt 1992); two compared LAVH versus TLH (Drahonovsky 2006; Long 2002); one compared both LH versus AH and LH versus VH (Garry 2004); and three compared LH versus AH versus VH (Hwang 2002; Ottosen 2000; Ribiero 2003). Two studies have been described in two papers each. Where Ollson 1996 is mentioned in the review, the data from Ellstrom 1998b have been used where applicable. The eVALuate trial population was studied in two papers (Garry 2004; Sculpher 2004) and study quality was summarised under Garry 2004.

Study design
All of the included trials had a parallel-group design. Twenty-five of the trials were single-centre studies (five from Italy; three from Sweden; three from Taiwan; two each from the UK, USA, Brazil, France, and Germany; and one each from the Czech Republic, Finland, the Netherlands, and Hong Kong). Of the nine multicentre trials, three trials recruited from two centres (Darai 2001 based in France; Langebrekke 1996 based in Norway; Miskry 2003 based in the UK). Three trials recruited from three centres (Summitt 1998 based in the USA; Lumsden 2000 based in the UK; Muzii 2007 based in Italy). One trial from Italy (Marana 1999) recruited from four centres; one Swedish trial recruited from five centres (Persson 2006); and a trial based in the UK with additional centres in South Africa (Garry 2004) recruited from 30 centres.

Participants
The 34 included studies involved 4495 women, the majority in the age range 41 to 50 years.

All of the included studies recruited women who needed a hysterectomy for benign causes; six studies specifically included women who underwent hysterectomy for symptomatic uterine fibroids (Benassi 2002; Ferrari 2000; Hwang 2002; Long 2002; Ribiero 2003; Tsai 2003).

VH versus AH
Benassi 2002 specifically included women with symptomatic enlarged fibroid uteri and excluded women with prolapse, vaginal stenosis, neoplasia, previous pelvic surgery, and taking hormone treatments within the six months prior to surgery. Miskry 2003 excluded women with uterine size greater than 14-weeks gestation, malignancy, adnexal pathology, reduced uterine mobility, or reduced vaginal access, and any woman requiring concomitant prolapse or incontinence surgery. Silva Filho 2006 included women with myoma and uterine size < 300 cm3 and excluded women with uterine prolapse, need for associated procedures, and suspicion of extra-uterine disease.

LH versus AH (including LH-BSO versus AH-BSO, and LAVH versus minilaparotomy-AH)
Thirteen of the 23 studies that compared LH with AH specifically included women who were scheduled for an abdominal hysterectomy or who had contraindications for a vaginal hysterectomy (Ellstrom 1998; Harkki-Siren 2000; Falcone 1999; Ferrari 2000; Kluivers 2007; Lumsden 2000; Marana 1999; Muzii 2007; Ollson 1996; Seracchioli 2002; Summitt 1998; Tsai 2003; Yuen 1998). Contraindications to vaginal hysterectomy included: the size of the uterus greater than 12-weeks pregnancy (Kluivers 2007), greater than 14-weeks pregnancy (Lumsden 2000; Seracchioli 2002); uterine volume greater than 200 ml (Ferrari 2000), greater than 300 gm (Seracchioli 2002), greater than 280 gm (Marana 1999) or 200 gm (Schutz 2002); limited vaginal access (Ferrari 2000; Kluivers 2007; Marana 1999; Muzii 2007); lack of uterine descent (Kluivers 2007; Marana 1999; Muzii 2007) or immobile uteri (Ferrari 2000); previous pelvic surgery or a history of pelvic inflammatory disease (Ferrari 2000; Marana 1999; Muzii 2007); presence of moderate or severe endometriosis or adnexal disease, or both (Muzii 2007).

Thirteen studies excluded women according to their uterine size or width: uterine size greater than 12-weeks pregnancy (Langebrekke 1996), greater than 14-weeks pregnancy (Harkki-Siren 2000; Lumsden 2000; Perino 1999; Raju 1994), greater than 16-weeks pregnancy (Marana 1999; Tsai 2003; Yuen 1998), and greater than 18-weeks pregnancy (Kluivers 2007; Summitt 1998). Ellstrom 1998 and Ollson 1996 excluded women with a uterus width greater than 11 cm, whilst Harkki-Siren 2000 excluded women if the uterine width was greater than 10 cm.

women were excluded for various physiological and anatomical reasons: pubic arch of at least 90 degrees (Summitt 1998), uterine prolapse (Harkki-Siren 2000; Raju 1994; Seracchioli 2002), pelvic floor relaxation (Seracchioli 2002), and immobile uteri (Ferrari 2000). Medical reasons were: morbidly obese (Harkki-Siren 2000; Raju 1994), suspicious adnexal mass or malignant disease (Falcone 1999; Marana 1999; Langebrekke 1996; Persson 2006; Seracchioli 2002; Summitt 1998), severe pelvic disease including adhesions and endometriosis (Ferrari 2000; Harkki-Siren 2000; Ollson 1996; Summitt 1998), concomitant incontinence procedure, pelvic reconstruction or colporrhaphy required (Falcone 1999; Kluivers 2007; Summitt 1998), or if the women had any serious diseases including cardiopulmonary disease, bleeding disorders, psychiatric disorders (Harkki-Siren 2000; Langebrekke 1996; Persson 2006; Seracchioli 2002; Summitt 1998) or an absolute contraindication to laparoscopy (Muzii 2007).

LH versus VH
Three of the six studies that compared LH with VH included women if their uterine size was larger than 280 gm (Darai 2001; Soriano 2001) or below the midpoint between the pubis and umbilicus (Agostini 2006). The remaining three studies excluded women if their uterine size was greater than 12-weeks (Morelli 2007) or 16-weeks pregnancy (Richardson 1995; Summitt 1992).

Exclusions for physiological and anatomical reasons were: pubic arch of at least 90 degrees (Summitt 1992), narrow vagina (Darai 2001), and immobile uteri (Darai 2001; Summitt 1992). Medical reasons were: suspicious adnexal mass or malignant disease (Agostini 2006; Darai 2001; Morelli 2007; Richardson 1995; Soriano 2001), severe pelvic disease including adhesions and endometriosis (Richardson 1995; Soriano 2001), concomitant incontinence procedure, pelvic reconstruction or colporrhaphy required (Morelli 2007; Summitt 1992), or if the women had any serious diseases including cardiopulmonary disease, bleeding disorders (Agostini 2006; Morelli 2007; Summitt 1992). Agostini 2006 excluded patients who refused bilateral oophorectomy or vaginal surgery and virgin patients.

VH versus LH (vLH as it was called in the trial) and AH versus LH (aLH as it was called in the trial)
Garry 2004 included women scheduled for hysterectomy for non-malignant conditions. The same exclusion criteria were used for both arms of the trial: a uterine mass greater than the size of 12-weeks pregnancy, suspected malignant disease of the genital tract, uterine prolapse, serious medical illness precluding surgery, and requirement for bladder or other pelvic support surgery.

LH versus AH versus VH
Two of the three trials (Hwang 2002; Ribiero 2003) specifically included women with uterine fibroids. Ottosen 2000 included women with leiomyomas <15 cm in diameter; Hwang 2002 included women with a myoma diameter larger than 8 cm and the second myoma less than 5 cm, or two myomata both at least 6 cm in diameter but less than 8 cm (a maximum of three myomata); Ribiero 2003 included women with fibroids or adenomyosis. Ottosen 2000 excluded those with a uterine mass larger than 16-weeks gestational size, previous dense adhesions, narrow vagina, or inaccessible uterus. Hwang 2002 excluded those with indications of adenomyosis, uterine prolapse, chronic pelvic pain, dysfunctional uterine bleeding, cervical dysplasia or pelvic inflammatory disease (PID). Ribiero 2003 excluded women: with uterine volume greater than 400 ml; taking anti-inflammatory drugs; with diabetes mellitus, coagulation disorders, and autoimmune disease.

LAVH versus TLH

Drahonovsky 2006 included women with benign uterine disease and excluded women in whom: laparoscopy was contraindicated, there was suspicion of malignancy, the uterine size was beyond the 3rd month of gestation at clinical examination or more then 120 x 80 x 80 mm at ultrasound scan, there was a necessity of an accessory surgical procedure, or urinary incontinence or prolapse stage was beyond 1st degree.

In Long 2002, women were included if they had contraindications for vaginal hysterectomy (a uterine weight >280 gm, previous pelvic surgery, PID, need for adnexectomy, lack of uterine descent, and limited vaginal access). If their uterine volume was greater than 16-weeks pregnancy (or weight greater than 700 gm) they were excluded.

(Note that according to Condous 2007, a uterus of 12-weeks gestation corresponds to a uterus of approximately 220 gm.)

Interventions
Surgical procedures

VH versus AH

Four trials compared VH with AH (Benassi 2002; Miskry 2003; Ottosen 2000; Silva Filho 2006); one included a laparoscopic arm as well (Ottosen 2000). Hysterectomies were performed by standard technique for each route.

LH versus AH
Twenty-three trials included a comparison of laparoscopic hysterectomy (LH) with abdominal hysterectomy (AH). These included four trials that randomised women to LH, AH, and VH (Garry 2004; Hwang 2002; Ottosen 2000; Ribiero 2003). Raju 1994 compared LH and bilateral salpingo-oophorectomy (LH-BSO) with AH-BSO. Ellstrom 1998 stratified the two randomised groups (LH and AH) into total and subtotal hysterectomies. Muzii 2007 performed minilaparotomy for AH (with a moving surgical field or window and three separate retractors).

LH versus VH
Ten trials included a comparison of laparoscopic hysterectomy (LH) with vaginal hysterectomy (VH), including the four trials randomising women to LH, AH, and VH. Garry 2004 was a very large RCT comparing LH (called vLH in the trial) with VH and LH (called aLH in the trial) with AH; it was essentially two concurrent RCTs as part of the same study.

LAVH versus TLH
Drahonovsky 2006 and Long 2002 compared two types of laparoscopic hysterectomy, which was LAVH versus TLH in both studies.

Although all the trials used variations of the terms ‘laparoscopic assisted vaginal hysterectomy’ (LAVH) or ‘laparoscopic hysterectomy’, their definition varied according to what stages of the hysterectomy were completed laparoscopically and the point at which the operation continued vaginally. We included all trials with hysterectomies that had some laparoscopic component in the larger LH category. Using the Richardson 1995 ‘Staging of laparoscopic hysterectomy’ table (see Additional Table 2) we were able to categorise 26 of the 29 included studies that involved LH according to the amount of laparoscopic content. We also subcategorised these 22 trials involving LH as either LAVH, LH(a), or TLH, depending on the extent of the surgery performed either laparoscopically or vaginally (see Additional Table 1). If any trial included women undergoing different Richardson LH stages in the LH arm, we arbitrarily categorised the stage firstly, as the stage to which the surgeons had intended to go; secondly, if that information was not available, to the LH stage that most women underwent surgery; or thirdly, the most advanced LH stage that women underwent. According to Richardson staging, one trial involved stage zero LH (Ottosen 2000), four trials were stage two (Agostini 2006; Kunz 1996; Marana 1999; Raju 1994), three trials were stage three (Ferrari 2000; Muzii 2007; Tsai 2003), nine trials were stage four where the uterine artery was transected laparoscopically (Darai 2001; Ellstrom 1998; Ollson 1996; Persson 2006; Schutz 2002; Soriano 2001; Summitt 1992; Summitt 1998; Yuen 1998), and nine trials were stage five (Falcone 1999; Hwang 2002; Harkki-Siren 2000; Kluivers 2007; Langebrekke 1996; Morelli 2007; Perino 1999; Ribiero 2003; Seracchioli 2002). For three trials we were unable to sub categorise the LH procedures and we described these as ‘non-categorisable LH’: Richardson 1995 had LHs of all stages from 0 to 5, and two trials (Garry 2004; Lumsden 2000) did not stipulate the LH stages performed. In Long 2002 the LAVH treatment arm was a stage three whilst the TLH arm was a stage five. Drahonovsky 2006 did not provide information on the LAVH and TLH procedures.

Antibiotic prophylaxis and anticoagulant therapy
In 24 of the trials the use of antibiotic prophylaxis was reported. Twenty-one trials prescribed the following antibiotics pre-operatively only (intravenous unless otherwise stated): cefazoline 2 gm (Darai 2001; Soriano 2001; Summitt 1992; Summitt 1998); cephalosporine 2 gm (Kunz 1996; Langebrekke 1996); metronidazole 500 mg (Harkki-Siren 2000); cephalosporine and metronidazole (Ellstrom 1998; Ollson 1996; Richardson 1995); cefuroxime 1.5 gm and metronidazole 1 gm rectally (Ottosen 2000); cefuroxime 1.5 gm and metrinodazole 1g (Persson 2006); cefotaxime 2 gm (Benassi 2002); co-amoxiclav 1.2 gm (Miskry 2003); ampicillin 2 gm (Seracchioli 2002); piperacillin 2 gm (Lumsden 2000); cefoxitin (Agostini 2006); cefoxitin 1.5 g (Drahonovsky 2006); amoxicillin clavulanate 2.2 gm (Kluivers 2007); cefalotin 1 gm (Silva Filho 2006); and first or second-generation cephalosporin (Muzii 2007).

Long 2002 prescribed intravenous cefazolin 1 gm pre and post-operatively. Raju 1994 gave Amoxil clavulanate (Augmentin) by bolus intravenous injection during and for seven days following the operation. Hwang 2002 prescribed cephalosporin 1 g every 8 hours combined with aminoglycoside 80 mg every 12 hours for one day after surgery.

In Ollson 1996, antibiotics were used in the laparoscopic arm of the study but they were not routinely given for the abdominal hysterectomies.

The use of low molecular weight heparin was reported in nine trials: three trials prescribed heparin pre-operatively (Benassi 2002; Darai 2001; Soriano 2001) and six post-operatively (Drahonovsky 2006; Kluivers 2007; Langebrekke 1996; Miskry 2003; Ottosen 2000; Silva Filho 2006).

Anaesthesia and post-operative medication
Twenty-one trials specifically stated that all hysterectomies were completed under general anaesthesia (GA). In three trials, GA was used for all LHs but the choice of regional or general anaesthesia was left to the anaesthesiologists and patients for the AH or VH (Summitt 1992; Summitt 1998), or was not reported for AH (Muzii 2007). In Ottosen 2000, 109 of the 120 included women were operated on using GA, three had spinal blockade, and eight had spinal blockade in combination with epidural blockade. Benassi 2002 used GA for AH procedures and spinal anaesthetic for VH. Five trials did not report the anaesthetic technique used. Silva Filho 2006 described epidural anaesthesia for all VH and AH procedures.

Fifteen trials reported on the type of post-operative pain relief given to women. In six trials morphine was used, two via intramuscular morphine sulphate injections (Raju 1994; Soriano 2001), three via a programmable infusion pump (Ellstrom 1998; Falcone 1999; Yuen 1998), and in Ollson 1996 details of how the morphine was administered were not reported. In Hwang 2002 intravenous meperidine 50 mg was prescribed every four hours. Long 2002 administered lysine aspirin intravenously. Muzii 2007 prescribed ketorolac 30 mg once or twice daily and additionally on request on the operative and first post-operative days.

The use of oral or rectal analgesics was reported in 12 trials: Summitt 1992 and Summitt 1998 discharged women with 16 tablets of acetaminophenoxycodone; Raju 1994 gave rectal diclofenac immediately after surgery, followed by coproxamol or codidramol; Ellstrom 1998 and Hwang 2002 prescribed paracetamol; Soriano 2001 gave 2 gm propacetamol and 100 mg ketoprofen, started 30 to 60 minutes before completion of the operation and then every six hours for 24 hours followed by acetaminophen (paracetamol); Falcone 1999 gave oxycodone 5 to 10 mg every 4 to 6 hours as needed, then 325 to 650 gm acetaminophen (paracetamol) every 4 to 6 hours as needed; Kunz 1996 and Drahonovsky 2006 prescribed tramadol hydrochloride (100 and 50 mg respectively); and Marana 1999 and Perino 1999 prescribed ketorolac every six hours for the first 24 hours. The use of anti-emetic drugs was reported in three trials (Ellstrom 1998; Summitt 1992; Summitt 1998).

Risk of bias in included studies
An overview of the risk of bias is provided. Only one study fulfilled all criteria for adequate management of risk of bias (Garry 2004). The two studies on two different laparoscopic hysterectomy techniques has a high risk of bias (Drahonovsky 2006; Long 2002).

Allocation
Randomisation and allocation concealment
Eight studies randomised by computer and used sealed opaque envelopes for allocation concealment (Agostini 2006; Ferrari 2000; Hwang 2002; Miskry 2003; Muzii 2007; Ottosen 2000; Raju 1994; Summitt 1998). Two trials randomised by computer and used a telephone for allocation concealment (Garry 2004; Schutz 2002). Langebrekke 1996 used a table of random digits for randomisation and used sealed opaque envelopes for allocation of concealment. Ten trials used a computer-generated randomisation code (Benassi 2002; Darai 2001; Falcone 1999; Lumsden 2000; Marana 1999; Seracchioli 2002; Soriano 2001; Summitt 1992; Tsai 2003; Yuen 1998) and one trial used a random numbers table (Richardson 1995) but none of these latter 11 trials reported whether allocation was concealed. Four trials used sealed opaque envelopes for allocation of treatment. Persson 2006 numbered the envelopes according to a random list, and Kluivers 2007 sealed the envelopes after which they were shuffled and numbered by a third party. The other two trials did not report the randomisation method (Harkki-Siren 2000; Ollson 1996). Seven trials did not report the randomisation method or if it was concealed (Drahonovsky 2006; Ellstrom 1998; Kunz 1996; Long 2002; Morelli 2007; Perino 1999; Ribiero 2003). The methodological quality of the Long 2002 trial was as follows: women were randomised to treatment groups before a large number (66) of the women were excluded. Therefore, the women in each treatment group may not have been a true representation of the original randomised groups.

Blinding
One trial reported sham abdominal dressings until discharge from hospital in VH (Miskry 2003).

One trial reported blinding of the interviewer one month after surgery (Silva Filho 2006).

Incomplete outcome data
Dropouts

Twenty-five trials reported no dropouts. Nine trials reported dropouts, with the dropout rate ranging from 1.7% to 12%. Table 4 lists the trials that reported dropouts with the dropout circumstances. In six trials the dropouts were excluded from the data analysis (Long 2002; Lumsden 2000; Morelli 2007; Persson 2006; Summitt 1998; Yuen 1998) whereas the other three either included the data in the analysis where possible (Falcone 1999; Kluivers 2007) or performed a sensitivity analysis for the missing data (Garry 2004). Four trials had women withdraw pre-operatively: Falcone 1999 (4 out of 48), Garry 2004 (34 out of 1380), Morelli 2007 (20 out of 420), and Persson 2006 (1 out of 119). In the Lumsden 2000 study, seven women withdrew pre-operatively and case records were not available for three more. Two and one women respectively refused their assigned procedure in the Summitt 1998 and Kluivers 2007 studies; in the Yuen 1998 study, four women declined their assigned operation and a further two women refused to participate post-operatively. In the Long 2002 trial, excluded post-randomisation were: three women undergoing conversion to laparotomy, seven with incomplete records, and three with combined procedures. A further 53 were excluded because they did not have indications of uterine fibroids or adenomyosis. In the Persson 2006 trial, five patients allocated to AH and one to LH withdrew after giving informed consent prior to the operation or withdrew in the post-operative period before the five-week follow up.

Intention-to-treat
Twenty-five trials reported no dropouts. Of the nine RCTs reporting dropouts, three reported analysis by intention to treat (ITT), defined as all randomised women reported upon according to the group of randomised allocation (Falcone 1999; Garry 2004; Kluivers 2007). Six RCTs reporting dropouts did not report ITT analysis of all randomised women (Long 2002; Lumsden 2000; Morelli 2007; Persson 2006; Summitt 1998; Yuen 1998). One further trial that had no dropouts did not analyse by ITT but according to the treatment received, which was different to the assigned treatment in two cases: the operation was converted from LH to AH and these women were analysed in the AH group (Tsai 2003).

Selective reporting
In 15 studies it was not clear whether the outcome measures had been pre-defined since the primary outcome was not reported and no sample size had been performed. Another four studies did not report that a power calculation was performed for sample size. Garry 2004 performed the largest trial (n = 1380) and used major complications for power calculation. The recruitment target was met in the LH versus AH arm but not in the LH versus VH arm.

Outcomes
All of the trials assessed the operation times and intra or post-operative complications. Lumsden 2000 and Garry 2004 split the complications into major and minor. The majority (27 trials) assessed blood loss or haemoglobin change. Ellstrom 1998 reported on the difference in erythrocyte volume fraction. Febrile morbidity was measured in 13 trials, pulmonary function in one trial (Ellstrom 1998), and 14 trials reported any operations that were converted to abdominal surgery (Darai 2001; Drahonovsky 2006; Garry 2004; Kluivers 2007; Marana 1999; Morelli 2007; Muzii 2007; Ottosen 2000; Persson 2006; Richardson 1995; Seracchioli 2002; Soriano 2001; Summitt 1992; Summitt 1998).

Post-operative pain was assessed in 16 trials, with Ellstrom 1998 listing it as a primary outcome. Thirty-one trials assessed the length of post-operative hospital stay and nine included an analysis of costs. Recovery time or the time needed to return to normal activities or work was assessed in 14 trials. An assessment of health status was reported in nine trials, three trials included sexual activity or body image in the analysis (Garry 2004; Long 2002; Morelli 2007).

Other potential sources of bias
Surgeon’s experience
The surgeon’s experience or level of training was reported in 20 of the trials. Eleven of the trials used the authors of the trial or surgeons of senior registrar grade to perform all the operations. Five of these trials specified that the same group of surgeons performed operations for both interventions (Benassi 2002; Hwang 2002; Lumsden 2000; Seracchioli 2002; Silva Filho 2006). In five trials, surgeons for one intervention were different to those performing the other intervention: Ollson 1996 (LH carried out by 2 out of 5 senior registrar grade surgeons trained in LH, AH carried out by 2 out of 10 senior registrar grade surgeons trained in AH); Langebrekke 1996 (LH performed exclusively by the two authors, AH performed by any skilled gynaecologist in the department); Raju 1994 (LAVH performed by one of the authors, AH by one of the authors or a senior registrar grade surgeon); Kluivers 2007 (LH was performed or supervised (resident 39%) by 3 out of 10 experienced gynaecologists (at least 100 LHs), AH performed or supervised by all 10 gynaecologists; and Long 2002 (one surgeon performed all LAVH, another performed all TLH). Residents were the first surgeon in 39% of LH and 88% of AH. In Agostini 2006 the five surgeons were experienced in vaginal surgery but laparoscopic experience was not reported. Drahonovsky 2006 reported that all surgeons of the department participated in the LAVH and TLH procedures. In Ottosen 200015 gynaecological surgeons with assistants performed the operations, their experience varied and there were cases of residents performing operations under supervision. In Schutz 2002 71% of LH were performed by the attending physician and 29% by a resident under supervision, and 40% of AH were performed by the attending physician and 60% by the resident under supervision. One trial (Summitt 1998) used only gynaecological residents to perform all the operations with the assistance of the attending physician. It is unlikely that any of the latter three trials used the same group of surgeons for both intervention groups. In three other trials it was unclear if the surgeons performing the operations were different: Darai 2001 (all experienced in laparoscopic and vaginal surgery but no mention of who performed each intervention); Perino 1999 (LH by team of three laparoscopic surgeons with experience of more than 100 LHs, no details provided for AH arm); and Falcone 1999 (one of the senior authors performed all the LH operations with the assistance of a pelvic surgery fellow or resident, but no mention of the AH group). In four of the trials, surgeons of all grades and experience carried out the operations. In Garry 2004 each surgeon recruited to the trial had to have performed 25 of each procedure, however cases could be used for teaching if the main assistant was the designated surgeon.

Source of funding
Twelve studies reported their sources of funding. Three of these studies received funding from pharmaceutical or surgical instrumentation companies: Falcone 1999 received part of the funding from Ethicon Endosurgery Inc; Harkki-Siren 2000 received a part of its funding from the Research Foundation of the Orion Corporation; Summitt 1998 received all of its funding from US Surgical Corporation, USA.

Effects of interventions
Meta-analysis results
Where outcomes for specific comparisons included in the meta-analysis are not mentioned below, no data were available from the included trials. For results that were not statistically significant, the summary statistics and CIs have not been reported in the text but can be found in the meta-analysis graphs.

Where there were differences in the subcategories (for example TLH) these have explicitly been reported. All other subcategory meta-analyses were similar to meta-analysis of the pooled groups.

Primary outcomes
Return to normal activities
VH versus AH

For VH versus AH, patients returned to normal activities sooner after VH (MD 9.5 days, 95% CI 6.4 to 12.6 days; 176 women, 3 trials, Analysis 1.1) although statistical heterogeneity was present (Chi2 P value 0.02, I2 = 75.3%); similar results were obtained with a random-effects model.

LH versus AH

Return to normal activities was also quicker after LH than after AH (MD 13.6 days, 95% CI 11.8 to 15.4 days; 520 women, 6 trials, Analysis 2.1) although statistical heterogeneity was present (Chi2 P value 0.004, I2 = 71.2%); similar results were obtained with a random-effects model.

LH versus VH

For LH versus VH there was no difference in return to normal activities (140 women, 2 trials, Analysis 5.1).

Intra-operative visceral injury
VH versus AH

There were no statistically significant differences in bladder, ureter, or urinary tract injuries for the comparison VH versus AH (239 women, 3 trials, Analysis 1.4). No bowel or vascular injuries occurred in either group.

LH versus AH

Where bladder and ureter injuries were pooled as ‘urinary tract injury’, there was a significant increase in urinary tract injury for LH versus AH (OR 2.41, 95% CI 1.21 to 4.82; 2090 women, 12 trials, Analysis 2.5). There were no statistically significant differences in bladder, ureter, bowel, or vascular injuries for the comparison LH versus AH.

LH versus VH

There were no significant differences in urinary tract injuries between LH and VH (1205 women, 7 trials, Analysis 5.4). When regarding the LH subcategories, there were statistically significantly more urinary tract injuries for TLH versus VH (OR 3.69, 95% CI 1.11 to 12.24; 440 women, 2 trials, Analysis 6.4). There were no statistically significant differences in bladder, ureter, bowel, or vascular injuries for the comparison LH versus VH.

TLH versus LAVH

There were no statistically significant differences in bladder, ureter, urinary tract, or vascular injury for the comparison TLH versus LAVH (186 women, 2 trials, Analysis 7.1). No bowel injuries occurred in either group.

Major long-term complications
VH versus AH

No urinary dysfunction occurred in either group (80 women, 1 trial, Analysis 1.5).

LH versus AH

No significant differences were found in the following long-term complications: fistula formation (245 women, 2 trials, Analysis 2.6), and urinary dysfunction (246 women, 2 trials, Analysis 2.6).

LH versus VH

No significant differences were found in the following long-term complications: fistula formation (56 women, 1 trial, Analysis 5.5), and urinary dysfunction (80 women, 1 trial, Analysis 5.5).

TLH versus LAVH

No significant differences were found in the following long-term complication: sexual dysfunction (that is dyspareunia or failure to orgasm) (101 women, 1 trial, Analysis 7.2).

Secondary outcomes
Satisfaction and quality of life
VH versus AH
For VH versus AH, Silva Filho 2006 found significantly better quality of life after VH in the SF-36 sub scales for functional capacity, physical aspects, and pain; and a higher rate of patients in VH who would choose the same treatment again. There were no significant differences in patient satisfaction between VH versus AH (Benassi 2002).

LH versus AH

For LH versus AH, Garry 2004 demonstrated that quality of life (measured by the SF12 scoring system) was significantly better for LH at six weeks; body image was significantly improved for LH versus AH at six weeks and four months, but not 12 months; and sexual frequency was significantly higher at six weeks following LH. Kluivers 2007 found a significant treatment effect favouring LH in the RAND-36 scale for vitality in the first 12 weeks post-operatively.

There were no significant differences in patient satisfaction between LH and AH (Lumsden 2000).

LH versus VH

Morelli 2007 found a significant higher score on the physical component score of SF-12 for LH versus VH at six weeks post-operatively.

Operation time
VH versus AH

Three trials in the meta-analysis of VH versus AH showed a significant difference, two in favour of VH (259 women, 3 trials, Analysis 1.6). Because the direction of the treatment effect differed amongst studies, the results were not pooled.

LH versus AH

AH had a significantly shorter operation time than LH (MD 11.8 minutes, 95% CI 8.6 to 14.9 minutes; 1047 women, 11 trials, Analysis 2.7). In the subcategory of trials where LAVH was compared with AH, one trial showed a significantly shorter operation time in LAVH (Tsai 2003), whilst other subcategories of LH took significantly longer than AH operations (LH(a) versus AH: MD 30.6 minutes, 95% CI 25.6 to 35.7 minutes; 420 women, 5 trials, Analysis 2.7; TLH versus AH: MD 22.7 minutes, 95% CI 14.6 to 30.8 minutes; 161 women, 2 trials, Analysis 2.7).

LH versus VH

VH had a significantly shorter operation time than LH (MD 39.3 minutes, 95% CI 38.7 to 39.9 minutes; 741 women, 6 trials, Analysis 5.6) and, although statistical heterogeneity was present (Chi2 P value 0.0005, I2 = 77 %), similar results were obtained with a random-effects model.

TLH versus LAVH

LAVH had a significantly shorter operation time than TLH (MD 23.3 minutes, 95% CI 10.0 to 40.6; 101 women, 1 trial, Analysis 7.3).

Intra-operative complications (other than visceral injury)
VH versus AH

No significant differences in mean blood loss were found between VH and AH (140 women, 2 trials, Analysis 1.8).

LH versus AH

No significant differences were found in the number of women with substantial bleeding between LH and AH (1266 women, 5 trials, Analysis 2.5). For the subcategories, LH(a) was associated with significantly fewer blood transfusions than AH (OR 0.50, 95% CI 0.26 to 0.95; 641 women, 8 trials, Analysis 3.11).

LH versus VH

Substantial bleeding was higher for LH versus VH (OR 2.76, 95% CI 1.02 to 7.42; 904 women, 2 trials, Analysis 5.4). There were no differences in the number of unintended laparotomies (1290 women, 8 trials, Analysis 5.4).

TLH versus LAVH

There were no differences in number of unintended laparotomies (189 women, 2 trials, Analysis 7.1).

Short-term outcomes and complications
VH versus AH

Hospital stay was significantly shorter in VH compared to AH (MD 1.1 day, 95% CI 0.9 to 1.2 days; 295 women, 4 trials, Analysis 1.13) although statistical heterogeneity was present (Chi2 P value < 0.00001, I2 = 95.0%); similar results were obtained for these outcomes using a random-effects model). For VH versus AH, there were significantly fewer febrile episodes or unspecified infections in VH (OR 0.42, 95% CI 0.21 to 0.83; 295 women, 4 trials, Analysis 1.10). There were no significant differences in the need for blood transfusion, mean blood loss, haemoglobin drop, occurrence of pelvic haematoma, or vaginal cuff infection, UTI and chest infection for VH versus AH.

LH versus AH

Hospital stay was significantly shorter in LH compared to AH (hospital stay MD 2.0 days, 95% CI 1.9 to 2.2 days; 1007 women, 10 trials, Analysis 2.14) although statistical heterogeneity was present (Chi2 P value < 0.00001, I2 = 95.0%); similar results were obtained for these outcomes using a random-effects model. For LH versus AH, there were significantly fewer wound or abdominal wall infections in LH (OR 0.31, 95% CI 0.12 to 0.77; 530 women, 6 trials, Analysis 2.12) and significantly fewer febrile episodes or unspecified infections (OR 0.67, 95% CI 0.51 to 0.88; 2138 women, 15 trials, Analysis 2.12). Although LH and AH showed no significant difference in the need for blood transfusion, LH was associated with a significantly lower mean blood loss (MD 45.3 ml, 95% CI 17.9 to 72.7 ml; 693 women, 7 trials, Analysis 2.9) and smaller drop in haemoglobin (MD 0.55 g/L, 95% CI 0.28 to 0.82 gm/L; 288 women, 3 trials, Analysis 2.10). There were no significant differences in the occurrence of pelvic haematoma, vaginal cuff infection, UTI, or chest infection, and thromboembolic events.

LH versus VH

There were no significant differences in hospital stay for LH versus VH (685 women, 5 trials, Analysis 5.12). For LH versus VH, there was a significantly higher need for blood transfusion in LH (OR 2.07, 95 % CI 1.12 to 3.81; 1249 women, 7 trials, Analysis 5.10). There were no significant differences in the occurrence of pelvic haematoma, vaginal cuff infection, UTI, chest infection, febrile episodes or unspecified infection, and thromboembolic events.

TLH versus LAVH

There were no significant differences in hospital stay for TLH versus LAVH (101 women, 1 trial, Analysis 7.8). For TLH versus LAVH, there were significantly more febrile episodes or unspecified infections in TLH (OR 3.77, 95% CI 1.05 to 13.51; 186 women, 2 trials, Analysis 7.6). There were no significant differences in the need for blood transfusion, mean blood loss, and haemoglobin drop for TLH versus LAVH. There were no significant differences in occurrence of pelvic haematoma or vaginal cuff infection for TLH versus LAVH.

Sensitivity analyses
Exclusion of trials susceptible to inadequate sequence generation during the randomisation process

Exclusion of nine trials with unclear or detrimental sequence generation (Drahonovsky 2006, Ellstrom 1998; Kunz 1996; Long 2002; Morelli 2007; Ollson 1996; Perino 1999; Ribiero 2003; Silva Filho 2006) altered the results as follows: bleeding and transfusion in LH versus VH were no longer significantly different; and estimated blood loss, transfusion, and drop in haemoglobin in LH(a) versus AH were no longer significantly different.

Exclusion of trials susceptible to ‘surgeon effect’

Exclusion of the four trials in which surgeons for one intervention were unequivocally different to those performing the other intervention (Kluivers 2007; Langebrekke 1996; Ollson 1996; Raju 1994) did not alter the statistical significance of any meta-analysis results.

Data from included trials that were not in the meta-analysis
Only outcomes reaching statistical significance will be mentioned below (a full summary of results is presented in Data and analyses: Tables 01 to 08).

Primary outcomes
Return to normal activities
LH versus AH

Median duration of return to normal activities was significantly shorter for LH in three trials (Langebrekke 1996; Persson 2006; Raju 1994).

Secondary outcomes
Operation time
VH versus AH

Hwang 2002 found a significantly shorter median operating time for VH (74 minutes) versus AH (98 minutes).

LH versus AH

In five trials, AH had a significantly shorter median operation time than LH (Falcone 1999 (P < 0.001); Ferrari 2000 (P = 0.001); Muzii 2007; Persson 2006 (P < 0.0001); Raju 1994 (P < 0.0001)). In Drahonovsky 2006, median operating time was significantly shorter for LAVH (85 minutes) versus TLH (111 minutes) (P < 0.001).

LH versus VH

Hwang 2002 found a significantly shorter median operating time for VH (74 minutes) versus LH (109 minutes).

Intraoperative complications
LH versus AH

For LH versus AH, median estimated operative blood loss was significantly lower for AH in one trial (Falcone 1999), and for LH in two trials (Kluivers 2007; Yuen 1998). Median haemoglobin drop was significantly lower for LH versus AH in one trial (Schutz 2002).

LH versus VH

For LH versus VH, significantly more women experienced blood loss > 500 cc (Agostini 2006).

TLH versus LAVH

Drahonovsky 2006 found less blood loss for TLH versus LAVH.

Short term outcomes
VH versus AH

Benassi 2002 found a significant lower percentage of patients demanding analgesics after VH.

LH versus AH

For LH versus AH, LH was associated with significantly lower pain scores in a number of trials: on post-operative days 0, 1, 2 and 3 (Marana 1999), day 1 and 2 (Muzii 2007), day 2 (Ollson 1996), day 4 (Schutz 2002); and on coughing (Ellstrom 1998). TLH was associated with significantly less severe post-operative pain than AH (Perino 1999). Recovery from pain was significantly faster for LH (Raju 1994). Concerning analgesic use, LH was associated with: significantly less opiate use (Garry 2004; Kluivers 2007) and oral and rectal analgesia (Langebrekke 1996); shorter duration of analgesic use overall (Raju 1994) and of patient-controlled analgesic use (Falcone 1999); fewer patients requiring intramuscular narcotics on the day of surgery (Summitt 1998); and less analgesic use after the first 24 hours (Ferrari 2000). Median duration of hospital stay was significantly shorter for LH in six trials (Falcone 1999; Ferrari 2000; Langebrekke 1996; Persson 2006; Raju 1994; Yuen 1998).

LH versus VH

For LH versus VH, Morelli 2007 found significantly lower pain scores on day zero for LH. In Summitt 1992, LH was associated with significantly greater use of oral pain tablets on post-operative day two.

TLH versus LAVH

For TLH versus LAVH, TLH was associated with significantly greater use of tramadol during hospitalisation (Drahonovsky 2006).

Cost
LH versus AH

No trial found a significant difference in the overall cost of LH versus AH, but only five RCTs examined comparative cost in any detail (Ellstrom 1998; Falcone 1999; Lumsden 2000; Raju 1994; Summitt 1998).

LH versus VH

The mean total hospital cost was significantly higher for LH than for VH (Summitt 1992).

Discussion

Summary of main results
Our review found a number of statistically significant advantages of VH over AH. VH was associated with quicker return to normal activities, earlier discharge from hospital, and VH was less painful. There were conflicting data on which was the quickest operation to perform and this presumably relates to the prior experience with these procedures of the surgeons involved in the trials. LH offered a number of statistically significant advantages over AH. These were quicker return to normal activities, less post-operative pain, fewer wound or abdominal wall infections, fewer febrile episodes or unspecified infections, smaller drop in haemoglobin, earlier discharge from hospital, and improved quality of life at six weeks and four months after surgery; the cost was more urinary tract injuries and longer operating time. LH had a number of statistically significant disadvantages compared to VH. These were longer operating time, higher rate of substantial bleeding, greater use of oral pain tablets on day two, and a higher hospital cost. TLH was associated with statistically significantly more urinary tract injuries compared to VH. TLH was associated with significantly more febrile episodes or unspecified infections and longer operation time compared to LAVH.

Speed of recovery is determined by the avoidance of an abdominal procedure; AH is associated with lengthier recovery than all other approaches to hysterectomy. Avoidance of AH appears to be important to minimise post-operative pain and to avoid abdominal wall infections and infections of unspecified origin or general apyrexial illness post-operatively.

Although regarded as very important, the quality of life data do not lend themselves easily to meta-analysis (due to the use of diverse tools, time frames, and statistical analysis). Data on quality of life can show the impact of surgery and complications on patient’s lives, and thus can be a leading argument in the discussion about the best way to perform a hysterectomy (Kluivers 2008). Only a few studies in the meta-analysis have used quality of life as an outcome measure. The available data indicate that the laparoscopic and vaginal procedures performed better or equally compared with AH as far as the quality of life in the first weeks after the procedure was concerned. In the decision on an approach to hysterectomy, the advantage of better quality of life should be offset against disadvantages. Meta-analysis of quality of life data would benefit from the use of well validated instruments applied in a standardised manner in future studies (Kluivers 2008b; Kluivers 2008c).

Urinary tract damage, in particular ureteric injury, remains the major concern related to the laparoscopic approach (Garry 2004; Garry 1995; Harkki-Siren 1997). However, this meta-analysis of RCTs was underpowered to detect a clinically significant increase in the incidence of bladder and ureter damage from a laparoscopic approach. Much of the data for an increased incidence of urinary tract injury has come from non-randomised studies. Only large case series usually have the power to detect such a rare complication, but RCTs remain the least biased way to assess the benefits and harms of an intervention. When bladder and ureter injuries in our meta-analysis were pooled under a single category ‘urinary tract injury’, a significant increase in urinary tract injury was detected for LH versus AH (OR 2.4, 95% CI 1.2 to 4.8) and TLH versus VH (OR 3.7, 95% CI 1.1 to 12.2).

Operating time is overall longer for LH versus AH, and LH versus VH. However, LAVH had a significantly shorter operating time than TLH. This suggests that operating time seems to be governed by the proportion of the surgery performed laparoscopically and the greater proportion performed laparoscopically, the lengthier the operation.

Overall completeness and applicability of evidence
It is particularly difficult to address the issues surrounding effectiveness and complications in surgical procedures where the skill base of surgeons is not only variable but different between surgeon experience of ‘traditional’ operations and ‘laparoscopic’ operations. This is likely to be especially relevant to the rates at which complications, such as ureteric damage, occur. There is no good way of taking into account the risk of such rare complications in surgeons who are beyond their learning curve. This is not just a hysterectomy issue but pervades many aspects of surgical therapy and surgical innovations. It does not apply to the same extent where drug therapy interventions are being studied, in which the efficacy is much less dependent on the skill of the investigator providing the treatment. Much of the Cochrane methodology is developed based on the medical model of intervention.

Until the last few years, the vast majority of hysterectomies were performed abdominally (Reich 2003; Vessey 1992), although in some countries there is a tendency to perform fewer abdominal hysterectomies (Brummer 2008; Spilsbury 2006). In the current state of gynaecological practice and training, all training gynaecologists tend to become thoroughly trained in abdominal hysterectomy techniques but there is huge variation in their learning curve position in relation to vaginal and laparoscopic hysterectomy techniques.

In clinical practice as well as in the trials included in this review, VHs will be mostly performed under optimum conditions only, whereas AH remains the default intervention for all more difficult cases. Each gynaecologist (as has been the case since AH became the alternative to VH, in 1863) will have his or her own indications for the choice of approach to hysterectomy for benign disease. These choices may be influenced to some extent by the results from scientific evidence (for example this review) but the decisions will also be largely based on their own array of surgical skills and the patient characteristics. Whether there will be more of a consensus in the future than there has been to date, regarding these indications for route of hysterectomy, is less certain. To reach this consensus, however, should probably not be the ultimate goal since the prudent decision for one approach to hysterectomy over the other may be very justified and may lead to better outcomes after all.

One concern is the statistical heterogeneity of the trials included in this review. The heterogeneity in such outcomes as operating time, even when the ‘traditional’ hysterectomy techniques VH versus AH are compared, directly relates to the fact that some surgeons are better trained in and thus perform faster either type of hysterectomy. This heterogeneity might be expected to be even more apparent when LH is compared with either AH or VH. Concerning the heterogeneity in recovery time, hospital policies on post-operative stay and advice regarding when to resume work can differ, hence the observed differences.

Although much has been written in the scientific literature about various outcomes of hysterectomy, there has been no discussion on what outcomes are of key importance. Surgeons wish to minimise operative complications, healthcare managers wish to minimise costs, but what do patients want? Quality of life is likely to be the most key outcome as it captures the benefit the patient experiences from treatment and takes into account the effects of complications on women’s lives (Chien 2005; Johnson 2005b; Kluivers 2008). Consequently,  the most plausible primary measure of effectiveness is ‘return to normal activity’ (where VH and LH fare most favourably). ‘Major lasting problem’ could be considered as the primary adverse event, but data on all long-term outcomes are sparse in these RCTs. Whether it is reasonable to prioritise outcomes as primary or secondary in advance is controversial. Usual Cochrane policy is to term the most clinically relevant outcome as ‘primary’ rather than the one most obviously affected by the treatments under comparison. There is certainly scope for the authors of individual RCTs to report only the outcomes that they consider to have produced interesting results, resulting in reporting bias. Each single complication is rare and thus a large sample size is needed to capture each one of them individually and powerfully. So researchers tend to pool complications together into composite outcomes, an approach that is not scientifically sound. More importantly, when comparing different types of hysterectomies laparotomy cannot be a complication of abdominal hysterectomy, leading to asymmetry of comparison.

There is currently a much larger database of trial experience involving LAVH than for TLH and this undermines the extent to which conclusions may be drawn about TLH currently.

One vital conclusion from our review must be that VH remains a very good option when it is feasible, since we have not shown any significant disadvantages of VH versus any other approach. In selected cases, even in patients without previous vaginal delivery, VH can be performed (Tohic 2008). The concept that LH allows identification of pelvic disease (such as adhesions and endometriosis) which could otherwise lead to complications with VH, and that the meticulous haemostasis achievable with ‘final-look’ laparoscopy during LH might reduce pelvic haematomas or vaginal cuff infections, have not been borne out in the outcomes in this review. It is uncertain whether the increased detection of unexpected pathology at LH versus VH (Garry 2004) affects subsequent clinical outcomes. Although it has been suggested that LAVH does little more than to combine the complications of laparoscopic surgery with those of vaginal surgery (Reich 2003), this has not been supported in our review. Where oophorectomy is desired, a laparoscopic approach may facilitate this.

Quality of the evidence
Most outcomes for the comparisons LH versus AH, as well as LH versus VH, are mainly based on the large trial by Garry 2004 with a low risk of bias.

With regard to the comparison VH versus AH, the conclusions are based on six trials with comparable sample sizes and low risk of bias.

There was a high risk of bias in the only two studies on different approaches to laparoscopic hysterectomy, and consequently the results and conclusions from this comparison need to be appreciated with caution.

The risk of bias table provides a quick overview of the trial quality variables. The distinction between ‘good trial quality’ and ‘poor trial quality’ is, however, still a quite controversial area with no clear guidelines. In this review, the distinction between good  and poor trial quality has been made as based on the adequacy of allocation concealment. The sensitivity analysis has led to some changes in statistical significance in various variables on bleeding and blood loss. The findings with regard to complications, operation times, and recovery times did not change with exclusion of trials with more detrimental trial quality.

Potential biases in the review process
Three so called multi-arm trials have been included in the review (Hwang 2002; Ottosen 2000; Ribiero 2003) where data have been used twice in different comparisons. There is not an agreed approach to this problem. Since no large effects of correlation and non-independence of data are expected on the resulting conclusions, no special measures have been taken in the review to address this issue. Similar correlation between the two trials and inter-dependence of data might be present in the study by Garry 2004 where the surgeon, and not randomisation, made the decision in which trial a patient was included. Again, this is unlikely to have influenced the results of the review.

Implications for practice
When technically feasible, VH should be performed in preference to AH because of more rapid recovery and fewer febrile episodes post-operatively. Where VH is not possible, LH has some advantages over AH (including less operative blood loss, more rapid recovery, fewer febrile episodes, and wound or abdominal wall infections) but these are offset by longer operating time and more urinary tract (bladder or ureter) injuries. No advantages of LH over VH could be found; LH had longer operation time and more substantial bleeding, and TLH had more urinary tract injuries. Of the three subcategories of LH, there are more RCT data for LAVH and LH(a) than for TLH. The surgical approach to hysterectomy should be decided by a woman in discussion with her surgeon in light of the relative benefits and hazards. These benefits and hazards seem dependant of surgical expertise and may influence the decision.

Implications for research
The various subcategories of LH should be further evaluated against each other. For example, whether TLH has any benefits or harms in comparison to other forms of LH (including LH(a) and LAVH). The increase in the rate of ureteric injury resulting from LH, suggested by very large observational studies, remains to be conclusively proven by RCT data. In recent years, robot-assisted hysterectomy has come into practice, but RCTs are lacking, until now.

Although it is important that RCTs should have the same surgeon (or group of surgeons) carrying out each of the approaches being compared, different levels of expertise with each approach means that such RCTs are always likely to be statistically heterogeneous when considered for pooling in meta-analyses.

We strongly encourage trial authors to report their laparoscopic approach to hysterectomy according to our defined subcategories: LAVH, LH(a), TLH, and LSH ( Table 1). This should minimise the confusion that has prevailed in the first published literature on LH.

There is an absence of data for long-term outcomes in RCTs comparing surgical approached to hysterectomy. RCTs should aim to report long-term outcomes, including urinary, bowel, and sexual function, along with occurrence of fistulae. Quality of life may be regarded as a key outcome in trials on the approaches to hysterectomy for benign disease. To enable meta-analysis of quality of life data, well validated instruments should be applied in a standardised manner.

Acknowledgements

We thank the trial authors for their tremendous effort, and the editors and external peer reviewers of this work.

We thank David Barlow, who was involved in the accomplishment of the protocol and the first review, for all his work .

We thank Jane Clarke, the Managing Editor of Cochrane Menstrual Disorders and Subfertility Group in Auckland, for all her time and efforts in the review.

Random Diseases

You can leave a response, or trackback from your own site.

2 Responses to “Benign Gynaecological Disease”

  1. [...] often, varicose veins occur during pregnancy and then resolve on its own after delivery. However, during this period, varicose vein can cause a [...]

  2. [...] clinical studies employing modern research methods. In this article, we shall take several common gynaecological conditions for discussion on the current evidence of using Chinese medicine in those [...]

Leave a Reply