Closed suction surgical wound drainage after orthopaedic surgery

Abstract

Background

Closed suction drainage systems are frequently used to drain fluids, particularly blood, from surgical wounds. The aim of these systems is to reduce the occurrence of wound haematomas and infection.

Objectives

To evaluate the effectiveness of closed suction drainage systems for orthopaedic surgery.

Search methods

We searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (March 2006), and contacted the Cochrane Wounds Group. We also searched the Cochrane Central Register of Controlled Trials (The Cochrane Library 2006, Issue 1), and MEDLINE (1966 to March 2006). Articles of all languages were considered.

Selection criteria

All randomised or quasi‐randomised trials comparing the use of closed suction drainage systems with no drainage systems for all types of elective and emergency orthopaedic surgery.

Data collection and analysis

Two authors independently assessed trial quality, using a nine item scale, and extracted data. Where appropriate, results of comparable studies were pooled.

Main results

Thirty‐six studies involving 5464 participants with 5697 surgical wounds were identified. The types of surgery involved were hip and knee replacement, shoulder surgery, hip fracture surgery, spinal surgery, cruciate ligament reconstruction, open meniscectomy and fracture fixation surgery.

Pooling of results indicated no statistically significant difference in the incidence of wound infection, haematoma, dehiscence or re‐operations between those allocated to drains and the un‐drained wounds. Blood transfusion was required more frequently in those who received drains. The need for reinforcement of wound dressings and the occurrence of bruising were more common in the group without drains.

Authors' conclusions

There is insufficient evidence from randomised trials to support the routine use of closed suction drainage in orthopaedic surgery. Further randomised trials with larger patient numbers are required for different operations before definite conclusions can be made for all types of orthopaedic operations.

Plain language summary

Wound drains in orthopaedic surgery (surgery on the joints or limbs)

Drainage systems may be used to drain fluid, mostly blood, from surgical wounds after an operation. The drains can either be open or closed. Closed drains use vacuum bottles and plastic conduits to draw fluid away from the wound by producing a negative pressure. They have been evaluated within trials for a number of elective and emergency orthopaedic operations. Closed surgical drainage systems reduce the need for reinforcement of the dressing and bruising around the wound, but increase the need for blood transfusion. Wound drains do not appear to have any effect on the occurrence of wound infections, large blood collections within the wound (haematomas), or the need for additional surgery to correct wound healing complications.

Closed suction wound drains in orthopaedic surgery are of doubtful benefit.

Background

Surgical drainage systems may be either open drains or closed suction drains. An open drain is when an artificial conduit is left in the wound to allow drainage of fluids to the exterior. Closed suction drains consist of a perforated drainage tube placed within the wound and connected to a drainage bottle. Reducing the pressure within the drainage bottle encourages drainage of fluid from the wound. This review applies purely to closed suction drainage of surgical wounds in orthopaedic surgery (surgery of the musculoskeletal system).

Drains may be used in any orthopaedic operation. Common procedures in which they are used include joint replacement surgery such as that for hip, knee or shoulder replacement, fracture fixation and spinal surgery. The aim of using drains is to reduce the formation of haematomas thereby decreasing the likelihood of prolonged drainage from the wound, delayed wound healing or infection of the wound.

A potential adverse effect of surgical drains is that they may become contaminated and act as a conduit for infection into the depths of the wound (Willett 1988). Furthermore, on rare occasions they may be difficult to remove post‐operatively and can even require additional surgery to remove if they have been inadvertently misplaced or sutured to surrounding tissues. Conversely, drains may be displaced or removed prematurely by confused patients pulling on them. Using drains may add additional cost to the procedure but may reduce the number of wound dressings needed.

Objectives

To determine, based on the evidence from randomised controlled trials, if closed suction drainage is preferable to no drainage of the wounds for different types of orthopaedic surgery.

The following null hypothesis was tested:

There is no difference in outcome between closed suction drainage of the wound and no drainage after orthopaedic surgery. Different types of surgical operation were considered separately.

Methods

Criteria for considering studies for this review

Types of studies

All randomised controlled trials which compared closed suction drainage in any orthopaedic surgical operation with no drainage. Quasi‐randomised trials (where the methods of allocating participants to a treatment are not strictly random e.g. date of birth, hospital record number, alternation) and trials in which the treatment allocation was inadequately concealed were considered for inclusion.

Types of participants

Any patient of any age undergoing any form of orthopaedic surgery (elective or acute, including surgery for trauma). This includes those undergoing joint replacement, fracture fixation and spinal surgery.

Types of interventions

Only trials comparing closed suction drains with no drains were included.

Types of outcome measures

Primary outcomes

• Wound infection. This was defined as any surgical wound infection which occurs within the follow‐up period of the study using the criteria as used in each individual study.

Secondary outcomes

• Deep wound infection. This only applied to those operations in which a prosthetic implant has been inserted at the time of surgery and the infection occurred within the follow‐up period of the study. A deep infection was defined as an infection involving tissues or spaces beneath the deep fascia layer.

• Wound haematoma (as defined by each study)

• Wound dehiscence or separation of the suture line (as defined by each study)

• Wound bruising or ecchymosis (as defined by each study)

• Complications related to drains or removal of the drains

• Limb swelling

• Bleeding from the wound and need for reinforcement of the dressing

• Re‐operation for wound healing complications

• Number of patients requiring blood transfusion

• Post‐operative blood transfusion (in units)

• Fall in haemoglobin or haematocrit

• Thromboembolic complications

• Range of limb movement

• Functional assessment

• Patient discomfort

• Mortality

• Length of hospital stay

• Costs

Any other outcomes as mentioned in individual studies were considered for inclusion.

Search methods for identification of studies

We searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (March 2006), the Cochrane Central Register of Controlled Trials (The Cochrane Library 2006, Issue 1), MEDLINE (1966 to March 2006), MEDLINE Pending (March 2006), EMBASE (1988 to March 2006), CINAHL (1982 to March 2006), reference lists of articles and contacted trialists. The Specialised Register is compiled from multiple databases, including regular searches of the Cochrane Central Register of Controlled Trials in The Cochrane Library, MEDLINE (which combines subject specific terms with the optimal trial search strategy (Higgins 2005)), EMBASE and CINAHL, and handsearch results.

In MEDLINE (OVID WEB) the subject specific search was combined with the first two levels of the optimal trial search strategy for randomised trials (Higgins 2005) (seeAppendix 1). Further details about the search strategy can be found in the Group's Module (Madhok 2006). Similar search strategies were used for CINAHL (OVID‐WEB) and EMBASE (OVID‐WEB).

The Cochrane Wounds Group was also approached for trials. Articles of all languages were considered and translated if necessary. This review is considered updated to March 2006.

Data collection and analysis

Selection of trials 
 Two of the authors independently assessed all identified trials for inclusion. Differences were resolved by discussion.

Assessment of methodological quality

In this review, risk of bias is implicitly assessed in terms of methodological quality. 
 
 The main assessment of methodology was the method of randomisation. A further eight aspects of methodology were used (seeTable 1). Though the scores of the individual items were summed, this was to gain an overall impression rather than for quantitative purposes. Where necessary and practical, we contacted trialists for additional data and clarification.

1. Methodological quality assessment tool.

Items and scores

(1) Was there clear concealment of allocation? Score 3 (and code A) if allocation clearly concealed (e.g. numbered sealed opaque envelopes drawn consecutively). Score 2 (and code B) if there was a possible chance of disclosure before allocation. Score 1 (and code B) if the method of allocation concealment or randomisation was not stated or was unclear. Score 0 (and code C) if allocation concealment was clearly not concealed such as those using quasi‐randomisation (e.g. even or odd date of birth).

(2) Were the inclusion and exclusion criteria clearly defined? Score 1 if text clearly states the types of operations involved and the types of patients included and excluded. Otherwise score 0.

(3) Were the outcomes of patients who withdrew or were excluded after allocation described and included in an intention to treat analysis? Score 1 if yes or text states that no withdrawals occurred or data are presented clearly showing 'participant flow' which allows this to be inferred. Otherwise score 0.

(4) Were the treatment and control groups adequately described at entry and, if so, were the groups well matched for co‐morbidity, or appropriate co‐variate adjustment made? Score 1 if at least four admission details given (e.g. age, sex, mobility, function score, mental test score) with either no important difference between groups or appropriate adjustment made. Otherwise score 0.

(5) Were the care programmes other than the trial options identical? Score 1 if text states they were or if this can be inferred. Otherwise score 0.

(6) Were all the outcome measures, particularly the definition of wound sepsis, clearly defined in the text with a definition of any ambiguous terms encountered? Score 1 if yes. Otherwise score 0.

(7) Were the outcome assessors blind to assignment status? Score 1 if assessors of wound healing were blinded to study groups. Otherwise score 0.

(8) Was there active follow‐up of patients (review at scheduled times), as opposed to passive (simple reporting of incidents as they occurred) with a minimum follow‐up of three months? Score 1 if yes. Otherwise score 0.

(9) Was loss to follow‐up reported and if so were less than five per cent of patients lost to follow‐up? Score 1 if yes. Otherwise score 0.

Data collection 
 Data for the outcomes listed above were independently extracted by two of the authors and each trial assessed independently without masking of the study names for its quality of methodology. Differences were resolved by discussion.

Data synthesis 
 For each study, relative risks and 95% confidence intervals were calculated for dichotomous outcomes, and mean differences and 95% confidence intervals were calculated for continuous outcomes. Where appropriate, results of comparable groups of trials were pooled using both the fixed‐effect and the random‐effects models. Heterogeneity between comparable trials was investigated by visual inspection of the forest plot and tested using a standard chi‐squared test with additional consideration of the I‐squared statistic (Higgins 2003). The results for the random‐effects model are presented when there is significant heterogeneity (P < 0.10; I‐squared = 50% or more) in the results of individual trials.

Many of the studies included in this review had a unit of analysis problem. This arose due to two situations. In the first situation, all the participants in a study had bilateral surgery and the side to get the drain was decided randomly. Hence, those studies had a paired design. In the second situation some of the participants had more than one wound. Participants were randomly assigned to the drain or no‐drain group and outcomes were measured for all wounds on that individual. Hence, those studies had a cluster randomised design. We planned to perform a sensitivity analysis to examine the effect on the results of taking into account the cluster randomised design of some studies and the paired design of others.

Results

Description of studies

Details of the 36 included studies are given in the 'Characteristics of Included Studies' table. The studies involved a total of 5464 participants with 5697 surgical wounds. For the purpose of this review, the unit of analysis was the number of surgical wounds and not the number of participants randomised. Two studies (Kim 1998a, Kim 1998b) involved patients having bilateral surgery. In Paessler 1997, ten of the patients had bilateral surgery; in Walmsley 2005, 25 patients had bilateral surgery; in Sasso 1998, four patients had two operations each; in Ravikumar 2001, two patients had bilateral surgery and in Gonzalez DV 2004, two patients had bilateral operations. Lang 1998 involved 202 patients but some cases had more than one surgical wound, therefore the total number of wounds included was 287.

Types of surgery 
 
 Hip replacement (THR) 
 Murphy 1993 reported on 40 patients, Widman 2002 on 22 patients and Johansson 2005 on 105 patients. Kim 1998a reported on 48 patients who had bilateral hip replacement. Patients were randomised to one side having a drain inserted and no drain inserted in the contralateral side. Walmsley 2005 reported on 552 patients (577 operations). Ravikumar 2001 reported on 23 patients having hip replacement surgery (25 operations). Gonzalez DV 2004 reported on 102 patients (104 operations).

Knee replacement (TKR) 
 Adalberth 1998 reported on 49 patients, Holt 1997 on 136 patients, Esler 2003 on 100 patients, Leb 1995 on 100 patients and Jenny 2001 on 60 patients. Kim 1998b reported on 69 patients having bilateral total knee arthroplasty.

Unicompartmental knee replacement 
 Confalonieri 2004 reported on 78 patients.

Hip and knee replacements 
 Ritter 1994 reported on 140 hip replacements and 275 knee replacements. Crevoisier 1998 reported on 66 hip replacements and 32 knee replacements. Niskanen 2000 reported on 58 hip replacements and 39 knee replacements. Mengal 2001 reported on 152 hip replacements and 104 knee replacements. Ovadia 1997 reported on 30 hip replacements and 58 knee replacements. Nixon 2000 reported on 1293 hip and knee replacements.

Shoulder surgery 
 Gartsman 1997 reported on 100 patients undergoing rotator cuff repair, 100 patients undergoing anterior shoulder joint reconstruction for instability and 100 patients having a shoulder arthroplasty.

Iliac crest bone graft donor site 
 Sasso 1998 involved 108 patients (112 cases) having a traumatic spinal fracture treated by bone graft harvested from the anterior or posterior iliac spine. Four patients had two separate procedures and were therefore randomised twice.

Lumbar laminectomy 
 Payne 1996 reported on 205 cases of single level lumbar laminectomy.

Extensive lumbar spine surgery 
 One study of 83 participants involved patients having lumbar spine surgery operations which were all more extensive than a single level unilateral decompression (Brown 2004).

Spinal scoliosis surgery 
 Blank 2003 reported on 30 adolescent patients having posterior spinal fusion and instrumentation for scoliosis. 
 
 Anterior cruciate ligament reconstruction 
 Paessler 1997 reported a study of 139 cases (129 patients) having open reconstruction of the anterior cruciate ligament of the knee. Three studies reported on arthroscopically assisted anterior cruciate ligament surgery. Dhawan 2003 reported on 21 patients, Karahan 2005 reported on 27 patients and Straw 2003 reported on 47 patients.

Open meniscectomy 
 Browett 1978 included 100 patients having open meniscectomy of the knee.

Hip fracture surgery 
 Five studies on hip fractures were identified. Cobb 1990 reported on 70 patients, Duranthon 2000 on 86 patients, Lautenbach 2000 on 51 patients, Tjeenk 2005 on 200 patients and Varley 1995 on 177 patients. Cobb 1990, Tjeenk 2005 and Varley 1995 included patients having surgery for both intracapsular and extracapsular hip fractures. Duranthon 2000 reported only on fractures treated by bipolar hemiarthroplasty. Lautenbach 2000 did not specify the exact type of operations used. 
 
 Non‐emergency fracture fixation or bone grafting 
 Lang 1998 involved 202 patients having 287 surgical wounds for a variety of fracture fixations, or bone grafting donor or fracture sites. Compound fractures and contaminated wounds were excluded.

Excluded studies and studies awaiting assessment 
 Twenty‐three studies were excluded for the reasons listed in the 'Characteristics of excluded studies' table. This review did not consider studies of the number of drains used, their placement or the timing of their removal. A small number of randomised trials on this topic were found and detailed in the 'Characteristics of excluded studies' table.

One study of 494 knee replacements (Hill 2000) is awaiting further information before it can be assessed.

The authors of this review would be pleased to receive any additional information about any of the included studies or those awaiting assessment.

Risk of bias in included studies

The results of the methodological quality assessment (Table 1) for the included studies were as detailed below.

1 2 3 4 5 6 7 8 9 Total Study

Hip replacement surgery (THR) 
 2 1 1 1 1 0 0 1 0 ‐ 7 ‐ Gonzalez DV 2004 
 3 1 0 1 1 1 1 0 1 ‐ 9 ‐ Johansson 2005 
 3 1 1 1 1 1 0 1 1 ‐ 10 ‐ Kim 1998a 
 2 0 0 0 1 1 0 0 1 ‐ 5 ‐ Murphy 1993 
 2 1 1 0 1 1 0 0 0 ‐ 6 ‐ Ravikumar 2001 
 3 1 0 1 1 1 1 1 1 ‐ 10 ‐ Walmsley 2005 
 2 0 0 1 0 0 0 0 0 ‐ 3 ‐ Widman 2002

Knee replacement surgery (TKR) 
 3 1 0 1 1 1 0 1 0 ‐ 8 ‐ Adalberth 1998 
 2 0 0 1 1 1 0 1 0 ‐ 6 ‐ Esler 2003 
 0 1 0 1 1 1 0 0 1 ‐ 5 ‐ Holt 1997 
 2 0 0 0 1 1 0 1 1 ‐ 6 ‐ Jenny 2001 
 3 1 1 1 1 1 0 1 0 ‐ 9 ‐ Kim 1998b 
 1 0 0 0 0 0 0 1 0 ‐ 2 ‐ Leb 1995

Unicompartmental knee replacement 
 3 1 1 1 1 1 0 1 1 ‐ 10 ‐ Confalonieri 2004

Hip and knee replacement surgery 
 2 0 0 1 1 1 0 0 1 ‐ 6 ‐ Crevoisier 1998 
 1 1 1 1 1 1 0 0 0 ‐ 6 ‐ Mengal 2001 
 1 1 1 0 1 1 0 0 0 ‐ 5 ‐ Niskanen 2000 
 1 0 0 0 0 0 0 1 0 ‐ 2 ‐ Nixon 2000 
 2 1 1 1 1 0 1 0 1 ‐ 8 ‐ Ovadia 1997 
 1 1 0 0 1 1 0 0 0 ‐ 4 ‐ Ritter 1994 
 
 Shoulder surgery 
 3 1 1 0 1 1 0 0 1 ‐ 8 ‐ Gartsman 1997

Iliac crest donor site 
 2 1 0 0 1 1 0 1 0 ‐ 6 ‐ Sasso 1998

Lumbar laminectomy 
 2 1 0 0 1 0 0 0 1 ‐ 5 ‐ Payne 1996

Lumbar spine surgery 
 3 1 0 1 1 1 0 0 0 ‐ 7 ‐ Brown 2004

Spinal scoliosis surgery 
 1 1 0 0 1 1 0 0 1 ‐ 5 ‐ Blank 2003

Anterior cruciate ligament reconstruction 
 2 1 0 0 1 1 1 1 0 ‐ 7 ‐ Dhawan 2003 
 2 0 0 0 0 0 0 0 1 ‐ 3 ‐ Paessler 1997 
 0 1 0 1 1 0 1 1 1 ‐ 6 ‐ Straw 2003 
 2 1 0 1 1 1 1 0 0 ‐ 7 ‐ Karahan 2005

Open meniscectomy 
 0 1 0 1 1 1 0 0 0 ‐ 4 ‐ Browett 1978

Hip fracture surgery 
 0 0 1 1 1 0 0 0 0 ‐ 3 ‐ Cobb 1990 
 3 1 1 0 1 1 0 0 1 ‐ 8 ‐ Duranthon 2000 
 1 0 0 0 0 1 0 0 0 ‐ 2 ‐ Lautenbach 2000 
 3 1 1 1 1 1 0 0 1 ‐ 9 ‐ Tjeenk 2005 
 3 0 0 1 1 1 0 1 0 ‐ 7 ‐ Varley 1995

Non‐emergency fracture fixation or bone grafting 
 3 1 0 1 1 0 0 0 0 ‐ 6 ‐ Lang 1998

Regarding the method of randomisation, most studies used sealed envelopes, generally opened at the time of wound closure (Adalberth 1998; Brown 2004; Confalonieri 2004; Duranthon 2000; Esler 2003; Jenny 2001; Johansson 2005; Kim 1998b; Kim 1998a; Lang 1998; Murphy 1993; Ovadia 1997; Payne 1996; Ravikumar 2001; Sasso 1998; Tjeenk 2005; Walmsley 2005; Widman 2002 and Varley 1995). Gartsman 1997 used a table of random numbers at the time of wound closure, with the number being unknown to the surgeon until that time. Gonzalez DV 2004 and Karahan 2005 used a randomisation chart and Dhawan 2003 used a table of random numbers. Crevoisier 1998 used a pre‐sealed box which was opened in the operating theatre. Paessler 1997 used the toss of a coin. Holt 1997 was quasi‐randomised, using the even or odd number for the date of which the operation was undertaken. Browett 1978, Cobb 1990 and Straw 2003 were also quasi‐randomised, using the even or odd numbers of the patient's record number. Seven studies (Blank 2003; Lautenbach 2000; Leb 1995; Mengal 2001; Niskanen 2000; Nixon 2000 and Ritter 1994) did not give the method of randomisation.

Four studies (Dhawan 2003; Ovadia 1997; Straw 2003; Walmsley 2005) used a blinded method of assessment of wound healing using independent members of staff to assess wound healing and limb movement.

Effects of interventions

The outcome measures reported by each study are listed in the 'Characteristics of included studies' table. Where pertinent and when data were available, results, sub‐grouped by type of operation, are shown graphically in the Analyses.

Wound infection 
 The occurrence of any type of wound infection was reported in 30 studies as detailed in Analysis 01.01. The outcome for deep wound infection was reported in 24 studies (Analysis 01.02). Pooled results for the different types of surgery failed to show a statistically significant difference between groups. The most number of participants were for Total Hip Replacement (THR), with a wound infection incidence of 29/711 (4.1%) for drained wounds versus 34/704 (4.8%) for wounds without drains (relative risk (RR) 0.89, 95% confidence intervals (CI) 0.56 to 1.43). For Total Knee Replacement (TKR) the figures were 3/475 (0.6%) versus 4/467 (0.9%) (RR 0.80, 95% CI 0.24 to 2.60).

In addition to the data in the Analyses, the following outcomes were reported:

Hip replacement surgery 
 Ravikumar 2001 used the ASEPSIS scoring system for wounds at the second, fourth and seventh day post‐surgery. They reported without data that the average score was higher (representing poorer wound healing) for the no‐drain group and that this difference was statistically significant. Murphy 1993 also used the ASEPSIS score at day 2, 4, 5, 7 and 10 post‐operatively. No statistically significant difference was found between groups. 
 
 Knee replacement surgery 
 Leb 1995 reported without defining terms that no patients in either group had 'major wound healing complications' and 1/24 in the drained group and 2/76 in the un‐drained group had 'minor wound healing complications'. 
 
 Spinal scoliosis surgery 
 Blank 2003 reported that 0/18 patients with the drained wounds had wound healing complications, whilst 3/12 of the un‐drained group had wound healing complications. The exact nature of these complications was not specified. 
 
 Hip fracture surgery 
 For the study of Cobb 1990, wounds were inspected at seven days after surgery and assessed for signs of inflammation (heat, tenderness, redness and swelling). There was a tendency towards more signs of inflammation in the drained group but the difference was not statistically significant and the total number of patients showing signs of inflammation was similar (21/35 versus 20/35). Varley 1995 used the ASEPSIS scoring system. This is a grading of wound erythema (redness), exudates, isolation of bacteria and need for further treatment. A higher score represented impaired wound healing. Wounds were assessed at two, five and eight days after surgery. Mean scores for the drained versus un‐drained group were 1.02 versus 1.26 at day two, 1.05 versus 1.13 at day five and 1.33 versus 2.05 at day eight. The difference between the two groups was statistically significant at day 8 (P = 0.018 t‐test).

Wound haematoma 
 This was reported in 13 studies (Analysis 01.03). None of the different surgical procedures studied showed any statistically significant difference for this outcome related to the use of drains. The most reported results were for THR with six studies giving an overall incidence of 5/485 (1.0%) with drains versus 4/485 (0.8%) in the un‐drained group (RR 1.20, 95% CI 0.39 to 3.66).

In addition to the data in the Analyses, the following outcomes were reported:

Hip replacement surgery 
 Kim 1998a using an ultrasound assessment of wound haematomas, reported that at day six or seven 13/48 (27.1%) in the drain group and 26/48 (54.2%) in the un‐drained group had large haematomas; this result was statistically significant (P < 0.05). Widman 2002 used erythrocyte scintigraphy to evaluate the degree of haematoma. They reported a median relative volume of 330% for the 10 drained results versus 623% for the un‐drained wounds. This difference was reported to be not statistically significant.

Hip and knee replacement surgery 
 Crevoisier 1998 used a 'haematoma score' to grade the degree of ecchymosis and swelling on a scale of 0‐19, with 0 being none. The mean haematoma scores for the drained THRs and TKRs were 2.5 and 3.7 respectively versus 2.6 and 3.0 for those wounds that were allocated to the no drain group. The difference between the drain and un‐drained group was reported as not being statistically significant (P = 0.39). In Crevoisier 1998, the haematoma was operatively drained in 2/33 THRs and 0/16 TKRs in the drain group as opposed to none in the un‐drained group. Needle aspiration of the haematoma was undertaken in 1/33 THRs and 4/16 TKRs in the drain group versus 2/33 THRs and 2/16 TKRs in the un‐drained group.

Hip fracture surgery 
 Lautenbach 2000 stated without statistical analysis that the volume of haematoma as measured by ultrasound was smaller in the un‐drained group. Varley 1995 also used ultrasound assessment for 140 of the patients. The mean size of haematoma was 20 ml in both the drain and un‐drained groups.

Non‐emergency fracture fixation or bone grafting 
 For the 22 haematomas reported in Lang 1998, only two in the drain group required surgery to evacuate.

Incidence of haemarthrosis or joint effusion 
 For two studies on knee surgery the drain was inserted in the knee joint. Paessler 1997 reported without numbers that the frequency of aspiration of the knee after cruciate ligament repair was not significantly different between the two groups (P = 0.46). Browett 1978 reported on the size of any effusion when the knee was inspected at 10 days after surgery. They reported an increased prevalence of tense and moderate effusions for those knees that were not drained (31/50 versus 10/50) as opposed to small or no effusions in the drained knees (18/50 versus 39/50). The difference between groups was reported to be statistically significant (P < 0.01). These differences were no longer apparent at three weeks. 
 
 Wound dehiscence 
 This was reported in six studies (Analysis 01.04). None of the different surgical procedures studied showed any statistically significant difference for this outcome related to the use of drains. 
 
 Skin edge necrosis 
 Necrosis at the edge of the wound was reported in two studies (Analysis 01.05). Both studies reported this outcome was more common in un‐drained wounds. 
 
 Bruising/ecchymosis 
 This was reported in six studies. Summary of the results (Analysis 01.06) indicates an increased occurrence of bruising for those wounds that did not have drains.

In addition to the data in the Analyses, the following outcomes were reported:

Hip replacement surgery 
 Kim 1998a reported a statistically significant increase in the area of ecchymosis for those wounds which were not drained in comparison to the drained wounds. 
 
 Knee replacement surgery 
 Holt 1997 noted the mean area of ecchymosis was 28 cm² for the drained cases versus 91 cm² for the non‐drained cases (P < 0.0001).

Erythema of the wound 
 This was reported in three studies. Summary of the results (Analysis 01.07) indicates an increased occurrence of erythema for those wounds that did not have drains.

Reinforcement of the dressing 
 This was reported in 10 studies. Summary of the results (Analysis 01.08) indicates a statistically significant increased occurrence for those wounds that did not have drains. The figures for Esler 2003 are the number of patients that needed a change of dressing and the figures for Ritter 1994 are the number of patients in which the limb was immobilised for excessive bleeding post‐operatively. Blank 2003 reported on the number of wounds with saturated dressings. Niskanen 2000 reported on the need for considerable reinforcement of the dressing. In this study there were also a greater number of patients in the un‐drained group who required 'significant' reinforcement of the dressing (8/27 versus 17/31 for the drained versus un‐drained THRs and 5/20 versus 8/19 for the TKRs).

In addition to the data in the Analyses, the following outcomes were reported:

Hip replacement surgery 
 Kim 1998a reported a statistically significant increase (P < 0.001) in the amount of blood draining onto the dressing for the un‐drained wounds.

Knee replacement surgery 
 Leb 1995 reported that drainage persisted from the wound for a mean of 5.5 days in the drained group versus 4.7 days for the un‐drained group. The difference was stated to be not statistically significant. Esler 2003, reported on the mean blood volume in the dressing which was 55 ml for the drained versus 119 ml for the un‐drained group, a difference that was reported to be statistically significant. 
 
 Hip and knee replacement surgery 
 Nixon 2000 stated without data there was no significant difference in the ooze from the wounds at day three and five post‐operatively. 
 
 Persistent discharge or oozing from the wound 
 This was reported in seven studies. Summary of the results (Analysis 01.09) indicates a tendency towards more persistent discharge for the un‐drained wounds, with the results being statistically significant only for TKR (RR 0.34, 95%CI 0.14 to 0.80). 
 
 Swelling of the limb 
 
 Hip replacement surgery 
 Gonzalez DV 2004 reported on the change in thigh circumference which was 1.6 cm for the drained group versus 1.7 cm for the un‐drained group, a difference that was not statistically significant. 
 
 Knee replacement surgery 
 Adalberth 1998 and Holt 1997 reported no significant difference between groups for the swelling of the limb. Jenny 2001 measured the knee circumference at day 2, 7 and 14 post‐operatively. There was no difference between groups. Esler 2003 also reported no statistically significant difference in the mean knee circumference at day 3 and 10 post‐operatively. 
 
 Unicompartmental knee replacement 
 Confalonieri 2004 reported on thigh, knee and calf circumference on day 3, 1 week, 1 month and 4 months after surgery. Swelling was less for the drained group at 3 days after surgery (P value 0.01) but there was no statistically significant difference thereafter. 
 
 Hip and knee replacement surgery 
 Mengal 2001 reported a 'non‐statistically significant' increase in the limb swelling for the wounds without drains. Ovadia 1997 stated that one of the 18 hip replacement patients in the drained group had excessive swelling around the wound, as opposed to none in the un‐drained group. 
 
 Anterior cruciate ligament reconstruction 
 Dhawan 2003 measured the thigh girth at day 1 to 7 and reported no significant difference whilst Straw 2003 measured the knee circumference at two weeks, four weeks and six weeks. There was a significant increase for the un‐drained knees at two weeks (P = 0.02), but not thereafter. Karahan 2005 reported on thigh and leg circumference on day 1, 3, 5 and 7 after surgery. This was less for the drained group which was reported to be statistically significant for the day 7 measurements only. 
 
 Re‐operations for wound healing complications 
 Re‐operations for complications directly related to wound healing were mentioned in 20 studies (Analysis 01.10). Overall there was no statistically significant difference in the re‐operation rate between groups. The most reported results were for THR, with a re‐operation rate of 4/525 (0.8%) for drained wounds versus 3/534 (0.6%) for wounds without drains (RR 1.27, 95% CI 0.37 to 4.30).

Regarding the indications for the re‐operations for the different surgical procedures: 
 
 Hip replacement surgery 
 Ravikumar 2001 reported one patient in the drained group required debridement and a later revision arthroplasty for sepsis, whilst three patients in the un‐drained group required re‐operation: two for wound debridement and one for drainage of the haematoma.

Knee replacement surgery 
 Kim 1998b reported two re‐operations for sepsis. Mengal 2001 mentioned one case in the drained group, which required arthrotomy to remove a broken drain.

Iliac crest donor site 
 Sasso 1998 mentioned that two patients in the drain group and one in the un‐drained group required surgery for evacuation of haematoma and debridement.

Hip fracture surgery 
 Cobb 1990 stated that two of the drained group required re‐operation, one for a partial wound dehiscence and one for removal of a drain that was stitched in place. Varley 1995 also reported that one drain was inadvertently stitched in and required surgery to remove. Seven other patients in this study also required re‐operations but none for wound healing complications. 
 
 Open meniscectomy 
 Browett 1978 stated one case in the drained group required arthrotomy to remove a broken drain. 
 
 Non‐emergency fracture fixation or bone grafting 
 Lang 1998 reported that all the re‐operations were for either sepsis or evacuation of haematoma.

Transfusion requirements 
 Thirteen studies reported on the number of patients needing transfusion (Analysis 01.11). Transfusion was required more frequently in those patients with the drained wounds. For THR transfusion was required in 168/417 (40.3%) patients in the drain group versus 132/421 (31.4%) in the un‐drained group (RR 1.28, 95% CI 1.07 to 1.52). Most of the figures for the mean volume of blood transfused could not be combined due to only three studies reporting data with standard deviations (Analysis 01.12).

Hip replacement surgery 
 Gonzalez DV 2004 reported a mean of 1.6 units were transfused in the drain group versus 1.5 in the un‐drained group. Widman 2002 reported on the mean volume of blood transfused of 823 ml for the drained group versus 235 ml for the un‐drained group (P value reported as 0.03). 
 
 Knee replacement surgery 
 Holt 1997 reported the mean volume of blood transfused was 283 ml for the drained cases versus 278 ml for the un‐drained cases; again this difference was not statistically significant. Jenny 2001 reported that the total number of units transfused was 34 in the drained group versus 29 in the un‐drained group (P value reported as 0.03). 
 
 Hip and knee replacement surgery 
 Crevoisier 1998 reported mean transfusion requirements of 700 ml for both the drained and undrained THRs. For the knee replacements the figures were 280 ml in the drained group versus 263 ml in the un‐drained group. The difference between groups was reported to be not statistically significant. Mengal 2001 reported the mean volume of blood transfused was 540 ml versus 585 ml for the hip replacements with and without drains and 275 ml versus 495 ml for the knee replacement patients. The difference between groups for the knee replacement was reported to be statistically significant. Niskanen 2000 reported a mean transfusion for the hip replacements of 1.8 units for both the drained and un‐drained groups and a mean of 2.3 units for the drained knee replacements versus 1.4 for the undrained groups, a difference that was reported as not statistically significant. Nixon 2000 reported without data that there was an increased transfusion requirement in the drained group. Ovadia 1997 also gave the total number of units of blood transfused which were 13 and 21 units for the 50 drained total hip and knee replacements and 3 and 5 units for the un‐drained groups (P value reported as 0.005). Ritter 1994 reported the mean volume of blood transfused was 118 ml versus 93 ml for the hip replacement patients and 160 ml versus 157 ml for the knee replacements, with and without drains respectively. 
 
 Shoulder surgery 
 Gartsman 1997 stated there was 'no difference' in the transfusion requirements between groups.

Calculated total blood loss 
 
 Hip replacement surgery 
 Johansson 2005 calculated the total blood loss, which showed no difference between groups (Analysis 01.13). Murphy 1993 also estimated total blood loss, which was reported as a mean of 1455 ml for the drained group versus 1134 ml for the un‐drained group. The difference between these figures was reported as being statistically significant at the level of 0.05. Widman 2002 gave figures for the total recorded blood loss of 1759 ml for the drained group versus 624 ml for the un‐drained group (P value reported as 0.0001). 
 
 Knee replacement surgery 
 Esler 2003 estimated the total blood loss as a mean of 568 ml for the drained group versus 119 ml for the un‐drained group, a difference that was reported as being statistically significant. Jenny 2001 estimated the total blood loss as 1431 ml in the drained group versus 1264 ml in the un‐drained group (P value reported as 0.13). 
 
 Hip and knee replacement surgery 
 Mengal 2001 reported the total blood loss for the drained versus un‐drained THRs was estimated as 1942 ml versus 1766 ml, and for TKRs the figures were 1590 ml versus 1983 ml. The increased blood loss for the un‐drained knee replacements was reported to be statistically significant. 
 
 Anterior cruciate ligament reconstruction 
 Paessler 1997 determined the total fluid loss into the drain within the knee joint and from any aspirations of the knee that were undertaken. The mean was 146.9 ml for the 57 drained cases and 18.4 ml for the 82 un‐drained cases (P < 0.0001). 
 
 Change in haemoglobin or haematocrit 
 
 Hip replacement surgery 
 Ravikumar 2001 and Walmsley 2005 both reported no significant difference between groups for the post‐operative haemoglobin levels. Gonzalez DV 2004 reported on the fall in haematocrit, which was increased for the drained wounds (Analysis 01.15). 
 
 Knee replacement surgery 
 Adalberth 1998 and Esler 2003 also reported no significant difference between groups for the post‐operative haemoglobin levels. Holt 1997 reported a mean fall in haemoglobin of 28.5 g/litre for the drained group versus 33.2 g/litre for the un‐drained group, a result that was not statistically significant. Leb 1995 used the change in haematocrit as an estimation of blood loss, which suggested a significant increase for the drained group (P = 0.02). 
 
 Unicompartmental knee replacement 
 Confalonieri 2004 reported on the haemoglobin levels at 1 day and 1 week after surgery. There was no statistically significant difference between the two groups. The haematocrit was measured at day 3 and 1 week with no statistically significant difference between groups. 
 
 Hip and knee replacement surgery 
 Mengal 2001 and Ritter 1994 reported no significant difference between groups for the post‐operative haemoglobin levels. Niskanen 2000 reported no difference in the fall in haematocrit between groups. Ovadia 1997 reported the mean haemoglobin levels at two and six days post‐operatively. There was no significant difference between groups for either the THRs (99 g/litre versus 102 g/litre at day two) or TKRs (95 g/litre versus 100 g/litre at day two). Crevoisier 1998 reported there was no statistically significant difference in the change in haematocrit from before and after surgery between the groups. 
 
 Lumbar spine surgery 
 Brown 2004 reported no significant difference between groups for the post‐operative haemoglobin levels or for the change in haematocrit from before and after surgery. 
 
 Hip fracture surgery 
 Cobb 1990 and Duranthon 2000 reported no significant difference between groups for the post‐operative haemoglobin levels. Tjeenk 2005 reported on the fall in haemoglobin with no statistically significant difference between group (Analysis 01.14). Varley 1995 reported a mean fall in the haemoglobin of 27 g/litre for the drained group versus 24 g/litre for the un‐drained group. The difference was not reported to be statistically significant.

Thromboembolic complications 
 The occurrence of deep vein thrombosis (DVT) was reported in nine studies (Analysis 01.16). No statistically significant difference was seen between groups. Most outcomes were reported for THR with six studies reporting 9/570 (1.6%) for the drained wounds versus 2/566 (0.4%) for the wounds without drains (RR 3.30 95% CI 0.91 to 11.90). Holt 1997 used vascular imaging for all patients at discharge from hospital and at six weeks to assess the incidence of DVT, hence the higher reported incidence in this study. Five studies reported on the incidence of pulmonary embolism, with no statistically significant difference between groups (Analysis 01.17).

Mortality 
 Mortality related to the use of drains was reported in only three studies with no statistically significant difference between groups (Analysis 01.18).

Range of movement 
 
 Hip replacement surgery 
 Both Kim 1998a and Walmsley 2005 reported that there was no statistically significant difference between groups for the range of movements at the hip. 
 
 Knee replacement surgery 
 Adalberth 1998, Esler 2003, Holt 1997, Jenny 2001, Kim 1998b and Leb 1995 all reported that there was no statistically significant difference between groups for the range of movements at the knee. 
 
 Unicompartmental knee replacement 
 Confalonieri 2004 reported on knee flexion at 1 and 4 months, with no statistical significant difference between the two groups. 
 
 Hip and knee replacement surgery 
 Crevoisier 1998, Niskanen 2000 and Ovadia 1997 (TKR only) reported that there was no statistically significant difference between groups for the range of movements in the joints. Ritter 1994 reported on range of movement post‐operatively for the patients treated by knee replacement and the mean range was 72 degrees for both groups. 
 
 Anterior cruciate ligament reconstruction 
 Dhawan 2003 reported that there was no statistically significant difference between groups for the range of movements in the knee. Straw 2003 reported on range of knee movement at two weeks, four weeks and six weeks. There was a significant reduced range for the un‐drained knees at two weeks (P value = 0.02), but not thereafter. Karahan 2005 reported on knee flexion and extension on days 1, 3, 5 and 7 after surgery. There was less restriction of movement for the drained group, which was reported to be statistically significant (P value = 0.04) at day 7 for knee flexion. 
 
 Open menisectomy 
 Browett 1978 reported that there was no statistically significant difference between groups for the range of movements in the knee.

Functional assessment 
 
 Hip replacement surgery 
 Kim 1998a and Walmsley 2005 both reported that there were no statistically significant differences in the hip scores between groups. 
 
 Knee replacement surgery 
 Esler 2003 reported that the time taken to achieve a straight leg raise after knee replacement was a mean of 4.4 days for the drained group versus 3.4 for the un‐drained group, the P value for the difference was reported as 0.02. Holt 1997 reported no statistically significant difference for the time to straight leg raising and arthrofibrosis between groups. Leb 1995 reported no difference in the Insall knee scores between groups. 
 
 Hip and knee replacement surgery 
 Mengal 2001 stated there was no difference in the functional recovery between groups. 
 
 Anterior cruciate ligament reconstruction 
 Straw 2003 reported no statistically significant difference in muscle strength between the drained and un‐drained knees at three months. 
 
 Open menisectomy 
 Browett 1978 reported no significant difference between groups in the time to achievement of straight leg raising or quadriceps power between groups at three weeks follow up. 
 
 Pain 
 
 Hip replacement surgery 
 Kim 1998a stated there was no statistically significant difference between groups for pain. Ravikumar 2001 stated without data or statistical analysis that pain scores two and four days postoperatively were 'higher' in the drained group. 
 
 Knee replacement surgery 
 Esler 2003, Jenny 2001, Kim 1998b and Leb 1995 also stated there was no statistically significant difference between groups for pain. Holt 1997 reported no significant difference for the mean amount of morphine used post‐operatively. 
 
 Unicompartmental knee replacement 
 Confalonieri 2004 reported a mean pain score at 3 days, 1 week, 1 month and 4 months after surgery with no statistical significant difference between the two groups. Neither was there any difference for analgesic use. 
 
 Anterior cruciate ligament reconstruction 
 Dhawan 2003 used a visual analogue pain scale at day 1, 3 and 5 after surgery and reported a statistically significant increase in the pain scores (more pain) for the drained group. Paessler 1997 used a visual analogue scale to assess pain at 24 hours after knee ligament surgery. They also reported a significant increased pain for those allocated to the drain group (mean score for drained cases 5.6 versus 4.4 for un‐drained cases, P < 0.0001). Frequency of analgesic consumption between the two groups was reported as showing no significant difference (P = 0.27). Straw 2003 used a visual analogue scale at two weeks and reported there was no statistical significant difference between groups. Karahan 2005 reported on a visual analogue score at day 1, 3, 5 and 7 after surgery. An increased pain score was reported in the un‐drained group (P value 0.07, 0.001, 0.001 and 0.001 at day 1, 3, 5 and 7 respectively). 
 
 Open menisectomy 
 Browett 1978 reported on the difference between the drain and the un‐drained group for the mean number of parenteral injections of opiate required (1.3 versus 1.6), mean number of oral analgesic medications issued (11.1 versus 13.8) and mean number of days for which oral analgesia was required (4.8 versus 5.6). Although the authors of this study report that there was a trend to more analgesia consumption in those allocated to the un‐drained group, the difference was not statistically significant.

Cost analysis 
 For THR surgery Kim 1998a compared the cost of using drains with the control group. The cost of the drains was $135 per patient. There was an increase in the dressings required for those treated without drains which amounted to $50.

For TKR surgery Holt 1997 reported the cost of treatment for the two groups ‐ for the drain group, the mean cost was $17,269 versus $17,428 for the un‐drained group.

Hospital stay 
 Five studies reported usable data on hospital stay, with no difference between groups (Analysis 01.19). In addition, the studies of Walmsley 2005 (THR), Esler 2003 (TKR), Holt 1997, (TKR), Jenny 2001 (TKR), Crevoisier 1998 (hip and knee replacement) and Brown 2004 (lumbar spine surgery) all stated there were no significant difference in the length of hospital stay between drain and un‐drained cases.

Other outcomes 
 
 Hip and knee replacement surgery 
 Niskanen 2000 reported on C reactive protein levels as a measure of inflammation, at days three and eight post‐operatively and there was no significant difference between groups. 
 
 Lumbar spine surgery 
 Brown 2004 reported on the patient's temperature at day 1 and 2. The mean temperature was reported to be higher by 1 degree (P value 0.04) in the un‐drained group. By day 2 the difference was not significant. 
 
 Open menisectomy 
 Browett 1978 reported there was no difference in the mean time of return to work or sport between groups for 62 of the study patients who responded to a postal questionnaire. 
 
 Hip fracture surgery 
 Duranthon 2000 reported on the patient's temperature at day 1, 2, 5 and 10 post‐operatively. There was no significant difference between groups.

Discussion

The methodology of the studies identified and included in the analysis varied considerably. Some trials scored poorly for the methodology assessment, although in some of these cases the low score may reflect a poor reporting of trial methods rather than poor trial methodology. In addition, because few studies used a blinded assessment of outcome, there may be biases in the reporting of subjective outcome measures. In addition the follow‐up period for a number of the studies was short, often involving only the time the patient was in hospital and this may have resulted in under‐reporting of outcomes.

Because of the variety of surgical procedures performed, the characteristics of the participants included in the different studies varied considerably, from adolescent patients having major spinal surgery, to young, predominantly male patients undergoing cruciate ligament reconstruction and meniscectomy, to arthritic patients having hip and knee replacements and the very elderly, mainly female patients undergoing surgery for a hip fracture. Because of this, data for all the individual surgical procedures has been presented separately, with less reliance placed on the summated data for all the surgical procedures. However despite this difference in surgical procedures, there was homogeneity between results for many of the reported outcomes, particularly the important outcomes of wound infection, haematoma and re‐operations.

There was a unit of analysis problem for some of the studies included in this review. Six of the studies (Gonzalez DV 2004; Lang 1998; Paessler 1997; Ravikumar 2001; Sasso 1998; Walmsley 2005) had individual patients with more than one surgical wound (cluster randomised study) while two of the studies, Kim 1998b and Kim 1998a were a paired design with the side to get the drain decided at random. These study designs could be accommodated in RevMan using the inverse variance method if the relevant information was available from the individual studies. However, the statistical analyses reported for the individual studies did not take into account the non‐independence of data from the same individual and hence we did not have the necessary information from the studies to use the inverse variance method. As an alternative, we performed sensitivity analyses to investigate the impact of the study designs on the results from the meta‐analysis. For the cluster randomised studies, if the average cluster size is close to one, we would expect that there would be very little impact on the meta‐analysis if we took the study design into account. The average cluster size for five of the studies (Gonzalez DV 2004, Paessler 1997, Ravikumar 2001, Sasso 1998, Walmsley 2005) is close to 1. A sensitivity analysis was undertaken for these studies and even when we took the worst case scenario (intra class correlation coefficient = 1.0), this had little impact on the results that are currently presented. The average cluster size for Lang 1998 was a little larger (= 1.4) so this could potentially have more impact on the results. However, Lang 1998 was not combined with any other studies in the sub‐group analysis. The reader should be aware that the 95% confidence intervals from Lang's study reported in this review are narrower than the 95% confidence intervals would be if the clustering had been accounted for. A sensitivity analysis was also performed to assess the impact of the paired study design on the outcome of the meta‐analysis. The review authors used the maximum possible correlation for a particular study given the observed data and the results from that meta‐analysis were compared to those from a meta‐analysis assuming zero correlation (i.e. a parallel group analysis). Taking into account the pairing of the studies did not change the reviewers' conclusions.

The primary outcome measures for an investigation of drains should be the incidence of wound infections. The results of this analysis indicate no statistically significant difference for either the wound infection rate or the incidence of deep wound infection. Care should be taken before making definite conclusions for these outcomes, as the overall numbers of wound infections was small. Furthermore there may be an under‐reporting of many of these events ‐ the follow‐up for some of the studies was short and inadequate to assess the true prevalence of wound healing complications. Because the occurrence of wound infection is low, a large sample size is required with numbers in excess of 2000 participants within each group before it is possible to make any conclusions on the presence of absence of benefit for wound drains on the occurrence of wound infections.

There was a possible trend to more wound haematomas in the undrained group as shown in Analysis 01.03. However this outcome is easily influenced by the method of assessment of this outcome. Most of the haematomas were reported in the studies of Tjeenk 2005 and Lang 1998. Lang 1998 reported most of the haematomas occurred in the un‐drained group but only two patients, both in the drained group had haematomas considered large enough to require surgery to evacuate. Wound dehiscence (Analysis 01.04), which may occur as a consequence of a haematoma showed no difference in frequency between groups.

An outcome in favour for the use of drains was the reduction in the amount of blood leaking through the wound as demonstrated by the lower number of times the dressing needed to be reinforced in the drained group. This may cause some inconvenience, require additional nursing care and provoke some distress for the patient. In addition there was a reported reduction in persistent discharge from the wound for the drained cases for TKR. Erythema and bruising of the wound were also reduced for the drained wounds, although these outcomes were only reported in limited studies and these findings did not seem to have any significant detrimental effect for the patient.

No difference between drains and un‐drained wounds were found for the outcome of re‐operation for wound healing complications (Analysis 01.10). The overall number of operations was small, 28 in total, and the type of operations when specified were either for treatment of wound infection, drainage of haematoma or removal of a trapped drain (four reported cases).

Blood transfusion requirements tended to be lower for those patients in the un‐drained group and this may be a consequence of there being no loss of blood into the drain. This was a consistent finding for both hip and knee replacement surgery. There were insufficient studies or reporting of data to make any firm conclusions for other types of surgery. Pooling of the results in the Analyses indicated that the use of surgical drains increased the numbers of patients who required transfusion from 31% to 40% for THR and from 31% to 50% for TKR (Analysis 01.11). Statistical analysis for the outcomes of mean volume of blood transfused could only be done to a limited extent because of the failure to report on standard deviations for this outcome, but again showed increased blood transfusion for the drained wounds. The outcomes of change in haemoglobin or haematocrit were less useful and may be influenced by the extent to which the patient was transfused.

There were no definite differences or trends demonstrated between the drained and un‐drained groups for the outcomes of deep vein thrombosis, pulmonary embolism, mortality, swelling of the limb, range of movement, pain, function and hospital stay. However, due to the variable reporting and small number of reported cases for some of these outcomes, significant differences between trial groups cannot be definitely excluded. Cost analysis for the use of drains was not adequately reported and only mentioned in two studies (Holt 1997; Kim 1998a).

This review did not compare the use of re‐infusion drains, in which the blood drained into the suction bottles is re‐infused back into the patient. Theoretically this type of drainage system may reduce the fall in haemoglobin or the need for blood transfusion after surgery. The search strategy for this review found a number of studies which compared a re‐infusion drain with a standard closed suction drain. Review of these studies was felt to be outside the scope of this review. Future studies may compare a re‐infusion drain with no drain and such studies will be considered for inclusion in this review.

Authors' conclusions

Implications for practice.

Based on the randomised trials undertaken to date, there is insufficient evidence to support the routine use of closed suction drainage in orthopaedic surgery. The use of drains was associated with a more frequent need for transfusion. Wound drains do not appear to have any effect on the incidence of wound infections or haematomas, but bruising of the wound and the need for reinforcement of dressings is less for drained wounds.

Implications for research.

Further randomised trials of surgical drains in orthopaedic surgery are appropriate, particularly for surgical procedures that have not been adequately studied such as fracture surgery and spinal surgery. Such studies should have a larger number of patients, appropriate trial methodology, longer follow‐up and full reporting of outcomes. Until these studies are completed, the routine use of a closed suction drainage system in orthopaedic surgery will remain open to question.

What's new

Date	Event	Description
25 July 2008	Amended	Converted to new review format.

History

Protocol first published: Issue 4, 1999
 Review first published: Issue 4, 2001

Date	Event	Description
22 May 2007	New citation required and conclusions have changed	For the first update of the review in Issue 3, 2007 the following changes were made: 1. Two studies (Nixon 2000 and Widman 2002), previously awaiting assessment, are now included. 2. Fourteen newly identified studies (Blank 2003; Brown 2004; Confalonieri 2004; Della Valle 2004; Dhawan 2003; Esler 2003; Karahan 2005; Kim 1998b; Johansson 2005; Murphy 1993; Jenny 2001; Straw 2003; Tjeenk 2005; and Walmsley 2005) are included. 3. One previously excluded study (Tiemessen 1997) is now included under the title of Lautenbach 2000, after new information on the study was received. 4. One previously included study (Beer 1991) is now excluded after statistical advice ‐ the method of randomisation was considered to be inadequate. 5. One included study (Kim 1998) is renamed Kim 1998a. 6. One study (Nicolajsen 1996), previously awaiting assessment, is now excluded. 7. Fifteen newly‐identified studies (Ashraf 2001; Berman 1990; Gehrig 2005; Graupe 1996; Gerngross 1992; Kohler 2000; Mac 1993; Mah 1996; Martin 2004; Padala 2004; Sen 2005; Seyfert 2002; Tatari 2005; Trammell 1991 and Tsumara 2005) now excluded. 8. Extra reference added to Paessler 1997. 9. Additional data outcome table for deep wound infection, bruising/ecchymosis, erythema of wound, skin edge necrosis, pulmonary embolism, mean units blood transfused, calculated total blood loss, fall in haemoglobin, fall in haematocrit, mortality and hospital stay created. 10. Changes made to results, discussion and conclusions of the review.

Notes

The title of the published protocol was "Closed suction surgical wound drainage for preventing complications after orthopaedic surgery".

Appendices

Appendix 1. Search strategy for MEDLINE (OVID WEB)

Search strategy

1. Drainage/ 2. Suction/ 3. ((suction or wound) adj3 (drain?1 or drainage)).tw. 4. or/1‐3 5. Orthopedics/ 6. exp Orthopedic Procedures/ 7. exp Osteoarthritis/su 8. exp Fractures/su 9. Laminectomy/ 10. exp Joints/su [Surgery] 11. exp Musculoskeletal Diseases/su [Surgery] 12. (arthroplast$ or meniscectom$ or laminectom$ or arthrotom$ or internal$ fix$ or orthop#edic or bone graft$).tw. 13. exp Joint Prosthesis/ 14. or/5‐13 15. and/4,14

Data and analyses

Comparison 1. Drains versus no drain.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 All wound infections	30	5370	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.61, 1.15]
1.1 Hip replacement	12	1415	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.56, 1.43]
1.2 Knee replacement	9	942	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.24, 2.60]
1.3 Unicompartmental Knee Replacement	1	78	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.4 Hip and knee replacement	1	1292	Risk Ratio (M‐H, Fixed, 95% CI)	0.5 [0.13, 1.99]
1.5 Shoulder surgery	1	300	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.6 Iliac crest donor site	1	112	Risk Ratio (M‐H, Fixed, 95% CI)	1.04 [0.34, 3.21]
1.7 Lumbar laminectomy	1	200	Risk Ratio (M‐H, Fixed, 95% CI)	1.88 [0.17, 20.44]
1.8 Lumbar Spine Surgery	1	83	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.9 Spinal Scoliosis Surgery	1	30	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.10 Anterior cruciate ligament reconstruction	1	47	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.11 Hip fracture surgery	5	584	Risk Ratio (M‐H, Fixed, 95% CI)	0.62 [0.34, 1.14]
1.12 Non emergency surgical fixation of fractures	1	287	Risk Ratio (M‐H, Fixed, 95% CI)	2.16 [0.55, 8.47]
2 Deep wound infection	24	4394	Risk Ratio (M‐H, Fixed, 95% CI)	0.70 [0.29, 1.68]
2.1 Hip replacement	11	1375	Risk Ratio (M‐H, Fixed, 95% CI)	1.10 [0.26, 4.61]
2.2 Knee replacement	8	882	Risk Ratio (M‐H, Fixed, 95% CI)	0.2 [0.01, 4.09]
2.3 Unicompartmental Knee Replacement	1	78	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.4 Hip and knee replacement	1	1292	Risk Ratio (M‐H, Fixed, 95% CI)	0.5 [0.05, 5.50]
2.5 Lumbar laminectomy	1	200	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.6 Lumbar Spine Surgery	1	83	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.7 Spinal Scoliosis Surgery	1	30	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.8 Anterior cruciate ligament reconstruction	1	47	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.9 Hip fracture surgery	4	407	Risk Ratio (M‐H, Fixed, 95% CI)	0.77 [0.17, 3.48]
3 Wound haematoma	13	2172	Risk Ratio (M‐H, Fixed, 95% CI)	0.78 [0.52, 1.17]
3.1 Hip replacement	6	970	Risk Ratio (M‐H, Fixed, 95% CI)	1.20 [0.39, 3.66]
3.2 Knee replacement	4	446	Risk Ratio (M‐H, Fixed, 95% CI)	2.2 [0.54, 8.88]
3.3 Lumbar Spine Surgery	1	83	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
3.4 Spinal Scoliosis Surgery	1	30	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
3.5 Hip fracture surgery	3	356	Risk Ratio (M‐H, Fixed, 95% CI)	0.71 [0.41, 1.24]
3.6 Non emergency surgical fixation of fractures	1	287	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.21, 1.20]
4 Wound dehiscence	6	992	Risk Ratio (M‐H, Fixed, 95% CI)	1.02 [0.42, 2.49]
4.1 Hip replacement	1	66	Risk Ratio (M‐H, Fixed, 95% CI)	1.0 [0.15, 6.68]
4.2 Knee replacement	2	92	Risk Ratio (M‐H, Fixed, 95% CI)	2.5 [0.50, 12.42]
4.3 Shoulder surgery	1	300	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
4.4 Hip fracture surgery	2	247	Risk Ratio (M‐H, Fixed, 95% CI)	1.03 [0.15, 7.30]
4.5 Non emergency surgical fixation of fractures	1	287	Risk Ratio (M‐H, Fixed, 95% CI)	0.15 [0.01, 2.96]
5 Skin edge necrosis	2	216	Risk Ratio (M‐H, Fixed, 95% CI)	0.13 [0.02, 0.98]
5.1 Knee replacement	1	138	Risk Ratio (M‐H, Fixed, 95% CI)	0.11 [0.01, 2.03]
5.2 Unicompartmental Knee Replacement	1	78	Risk Ratio (M‐H, Fixed, 95% CI)	0.14 [0.01, 2.68]
6 Bruising/ecchymos	6	573	Risk Ratio (M‐H, Random, 95% CI)	0.48 [0.27, 0.86]
6.1 Hip replacement	2	121	Risk Ratio (M‐H, Random, 95% CI)	0.53 [0.15, 1.85]
6.2 Knee replacement	3	374	Risk Ratio (M‐H, Random, 95% CI)	0.47 [0.20, 1.10]
6.3 Unicompartmental Knee Replacement	1	78	Risk Ratio (M‐H, Random, 95% CI)	0.25 [0.03, 2.14]
7 Erythema of wound	2	312	Risk Ratio (M‐H, Fixed, 95% CI)	0.19 [0.11, 0.34]
7.1 Hip replacement	1	96	Risk Ratio (M‐H, Fixed, 95% CI)	0.27 [0.08, 0.92]
7.2 Knee replacement	1	138	Risk Ratio (M‐H, Fixed, 95% CI)	0.17 [0.08, 0.34]
7.3 Unicompartmental Knee Replacement	1	78	Risk Ratio (M‐H, Fixed, 95% CI)	0.2 [0.05, 0.85]
8 Large draining onto the dressing or need for reinforcement	10	1486	Risk Ratio (M‐H, Random, 95% CI)	0.40 [0.25, 0.65]
8.1 Hip replacement	4	533	Risk Ratio (M‐H, Random, 95% CI)	0.61 [0.36, 1.05]
8.2 Knee replacement	5	553	Risk Ratio (M‐H, Random, 95% CI)	0.27 [0.08, 0.84]
8.3 Lumbar Spine Surgery	1	83	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.13, 0.81]
8.4 Spinal Scoliosis Surgery	1	30	Risk Ratio (M‐H, Random, 95% CI)	0.22 [0.03, 1.89]
8.5 Non emergency surgical fixation of fractures	1	287	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.25, 1.32]
9 Persistent discharge from the wound	7	635	Risk Ratio (M‐H, Random, 95% CI)	0.63 [0.28, 1.44]
9.1 Hip replacement	5	322	Risk Ratio (M‐H, Random, 95% CI)	0.92 [0.29, 2.86]
9.2 Knee replacement	3	235	Risk Ratio (M‐H, Random, 95% CI)	0.34 [0.14, 0.80]
9.3 Unicompartmental Knee Replacement	1	78	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.04, 3.07]
10 Re‐operation for wound healing complications	20	2955	Risk Ratio (M‐H, Fixed, 95% CI)	1.21 [0.65, 2.26]
10.1 Hip replacement	6	1059	Risk Ratio (M‐H, Fixed, 95% CI)	1.27 [0.37, 4.30]
10.2 Knee replacement	4	373	Risk Ratio (M‐H, Fixed, 95% CI)	0.2 [0.01, 4.09]
10.3 Unicompartmental Knee Replacement	1	78	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
10.4 Shoulder surgery	1	300	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
10.5 Open meniscectomy	1	100	Risk Ratio (M‐H, Fixed, 95% CI)	3.0 [0.13, 71.92]
10.6 Iliac crest donor site	1	112	Risk Ratio (M‐H, Fixed, 95% CI)	1.73 [0.16, 18.57]
10.7 Lumbar Spine Surgery	1	83	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
10.8 Spinal Scoliosis Surgery	1	30	Risk Ratio (M‐H, Fixed, 95% CI)	0.10 [0.01, 1.74]
10.9 Hip fracture surgery	4	533	Risk Ratio (M‐H, Fixed, 95% CI)	1.30 [0.33, 5.20]
10.10 Non emergency surgical fixation of fractures	1	287	Risk Ratio (M‐H, Fixed, 95% CI)	6.48 [0.79, 53.13]
11 Patients requiring transfusion	13	1681	Risk Ratio (M‐H, Random, 95% CI)	1.22 [1.06, 1.41]
11.1 Hip replacement	5	838	Risk Ratio (M‐H, Random, 95% CI)	1.28 [1.07, 1.52]
11.2 Knee replacement	4	267	Risk Ratio (M‐H, Random, 95% CI)	1.58 [1.12, 2.21]
11.3 Lumbar spine surgery	1	83	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.01, 7.77]
11.4 Spinal scoliosis surgery	1	30	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.81, 1.14]
11.5 Hip fracture surgery	3	463	Risk Ratio (M‐H, Random, 95% CI)	1.12 [0.92, 1.38]
12 Mean units blood transfused	3	184	Mean Difference (IV, Fixed, 95% CI)	0.46 [0.08, 0.84]
12.1 Hip replacement	1	105	Mean Difference (IV, Fixed, 95% CI)	0.60 [‐0.07, 1.27]
12.2 Knee replacement	1	49	Mean Difference (IV, Fixed, 95% CI)	0.34 [‐0.27, 0.95]
12.3 Spinal Scoliosis Surgery	1	30	Mean Difference (IV, Fixed, 95% CI)	0.47 [‐0.23, 1.17]
13 Calculated total blood loss (mls)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
13.1 Hip replacement	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
14 Fall in haemoglobin (mg/dl)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
14.1 Hip fracture surgery	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
15 Fall in haematocrit	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
15.1 Hip replacement	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
16 Deep vein thrombosis	9	1790	Risk Ratio (M‐H, Fixed, 95% CI)	1.66 [0.81, 3.39]
16.1 Hip replacement	6	1136	Risk Ratio (M‐H, Fixed, 95% CI)	3.30 [0.91, 11.90]
16.2 Knee replacement	5	603	Risk Ratio (M‐H, Fixed, 95% CI)	1.09 [0.44, 2.67]
16.7 Hip fracture surgery	1	51	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
17 Pulmonary embolism	4	986	Risk Ratio (M‐H, Fixed, 95% CI)	1.57 [0.41, 5.95]
17.1 Hip replacement	3	833	Risk Ratio (M‐H, Fixed, 95% CI)	2.37 [0.35, 15.94]
17.2 Knee replacement	2	153	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.14, 6.88]
18 Mortality	3	881	Risk Ratio (M‐H, Fixed, 95% CI)	1.35 [0.72, 2.55]
18.1 Hip replacement	2	681	Risk Ratio (M‐H, Fixed, 95% CI)	1.57 [0.72, 3.43]
18.2 Hip fracture surgery	1	200	Risk Ratio (M‐H, Fixed, 95% CI)	1.0 [0.33, 3.00]
19 Hospital stay (days)	5	519	Mean Difference (IV, Fixed, 95% CI)	0.39 [‐0.06, 0.84]
19.1 Hip replacement	2	134	Mean Difference (IV, Fixed, 95% CI)	0.73 [‐0.13, 1.59]
19.2 Knee replacement	2	107	Mean Difference (IV, Fixed, 95% CI)	‐0.40 [‐1.54, 0.74]
19.3 Unicompartmental Knee Replacement	1	78	Mean Difference (IV, Fixed, 95% CI)	0.46 [‐0.14, 1.06]
19.4 Hip fracture surgery	1	200	Mean Difference (IV, Fixed, 95% CI)	‐1.0 [‐6.65, 4.65]

1.1. Analysis.

Comparison 1 Drains versus no drain, Outcome 1 All wound infections.

1.2. Analysis.

Comparison 1 Drains versus no drain, Outcome 2 Deep wound infection.

1.3. Analysis.

Comparison 1 Drains versus no drain, Outcome 3 Wound haematoma.

1.4. Analysis.

Comparison 1 Drains versus no drain, Outcome 4 Wound dehiscence.

1.5. Analysis.

Comparison 1 Drains versus no drain, Outcome 5 Skin edge necrosis.

1.6. Analysis.

Comparison 1 Drains versus no drain, Outcome 6 Bruising/ecchymos.

1.7. Analysis.

Comparison 1 Drains versus no drain, Outcome 7 Erythema of wound.

1.8. Analysis.

Comparison 1 Drains versus no drain, Outcome 8 Large draining onto the dressing or need for reinforcement.

1.9. Analysis.

Comparison 1 Drains versus no drain, Outcome 9 Persistent discharge from the wound.

1.10. Analysis.

Comparison 1 Drains versus no drain, Outcome 10 Re‐operation for wound healing complications.

1.11. Analysis.

Comparison 1 Drains versus no drain, Outcome 11 Patients requiring transfusion.

1.12. Analysis.

Comparison 1 Drains versus no drain, Outcome 12 Mean units blood transfused.

1.13. Analysis.

Comparison 1 Drains versus no drain, Outcome 13 Calculated total blood loss (mls).

1.14. Analysis.

Comparison 1 Drains versus no drain, Outcome 14 Fall in haemoglobin (mg/dl).

1.15. Analysis.

Comparison 1 Drains versus no drain, Outcome 15 Fall in haematocrit.

1.16. Analysis.

Comparison 1 Drains versus no drain, Outcome 16 Deep vein thrombosis.

1.17. Analysis.

Comparison 1 Drains versus no drain, Outcome 17 Pulmonary embolism.

1.18. Analysis.

Comparison 1 Drains versus no drain, Outcome 18 Mortality.

1.19. Analysis.

Comparison 1 Drains versus no drain, Outcome 19 Hospital stay (days).

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Adalberth 1998.

Methods	Method of randomisation: sealed envelopes opened at the time of wound closure. Number lost to follow‐up: 17/90 (19%) Methodological quality score: 8 Length of follow‐up: 4 months
Participants	Orthopaedic hospitals in Uppsala, Sweden 49 patients undergoing total knee replacement Mean age: 72 years (range 56‐86 years) Percentage male: 46%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Mean units blood transfused 2. Number of patients transfused 3. Thromboembolic complications 4. Length of hospital stay 5. Length of surgery 6. Range of movement 7. Swelling of the limb
Notes	The study also included a third group of patients allocated to an auto‐transfusion system. The results for this group of patients are not considered in this review. The results for five patients who developed post‐operative complications, five cases which did not follow the trial protocol and one patient in which data were lost were not given. Originally 30 cases were randomised to each group, after these exclusions there were 24 patients in the un‐drained group and 25 in the drained group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Blank 2003.

Methods	Method of randomisation: not stated Number lost to follow‐up: none Methodological quality score: 5 Length of follow‐up: six days
Participants	Orthopaedic Hospital in Philadelphia, USA 30 patients undergoing posterior spinal fusion for idiopathic scoliosis Mean age: 14 years (range 11‐18 years) Percentage male: 13%
Interventions	Wound drainage with single two suction drain versus no drain
Outcomes	1. Wound infections or haematoma 2. Transfusion 3. Re‐operation 4. Saturation of the dressing 5. Wound healing 6. Wound evaluation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Browett 1978.

Methods	Method of randomisation: hospital number even or odd. Number lost to follow‐up: none at 3 weeks. Postal questionnaire after discharge at unspecified time, response from 62/100 Methodological quality score: 4 Length of follow‐up: 3 weeks
Participants	Orthopaedic hospitals in Harlow, Essex, United Kingdom 100 patients undergoing open meniscectomy Mean age: 33 years (range not stated) Percentage male: 79%
Interventions	Wound drainage with a suction drain versus no drain Drain inserted into the knee joint
Outcomes	1. Size of effusion into knee 2. Range of movement 3. Quadriceps power 4. Return to work and sport 5. Pain
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Brown 2004.

Methods	Method of randomisation: sealed envelope opened at the time of wound closure. Number lost to follow‐up: not stated Methodological quality score: 7 Length of follow‐up: 1 year
Participants	Orthopaedic Hospital in Miami, USA 83 patients undergoing extensive lumbar spine surgery Mean age: 67 years (range not stated) Percentage male: not stated
Interventions	Wound drainage with a single suction drain versus no drain
Outcomes	1. Wound infection 2. Wound haematoma 3. Transfusion 4. Haemoglobin on day 1, 3 and discharge 5. Temperature day 1 and day 2 6. Haematocrit day 1,2 and 3 7. Number of patients transfused 8. Hospital stay 9. Dressing drainage 10. Re‐operations
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Cobb 1990.

Methods	Method of randomisation: even or odd hospital number. Number lost to follow‐up: none Methodological quality score: 3 Length of follow‐up: 1 week
Participants	Orthopaedic hospitals in Middlesex, UK 70 patients undergoing surgery for a hip fracture. Intra and extracapsular fracture included. Operations were sliding hip screw (40), hemiarthroplasty (29), multiple screws (1). Mean age: 81 years (range 45‐97) Percentage male: 10%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infection 2. Partial wound dehiscence 3. Signs of wound inflammation 4. Re‐operations 5. Change in haemoglobin
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Confalonieri 2004.

Methods	Method of randomisation: opening a sealed envelope with randomly altered numbers at the time of surgery Number lost to follow‐up: none Methodological quality score: 10 Length of follow‐up: 4 months
Participants	Orthopaedic hospital in Milan, Italy 78 patients undergoing unicompartmental knee replacement Mean age: 68 years (range not stated) Percentage male: 44%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infection 2. Ooze from the wound 3. Other minor wound healing complications (erythema, skin edge necrosis, ecchymosis) 4. Re‐operations 5. Change in haemoglobin 6. Change in haematocrit 7. Thigh, knee and calf circumference at day 3, 1 week, 1 month and 4 months 8. Pain score at day 3, 1 week, 1 month and 4 months 9. Knee flexion at 1 month and 4 months 10. Analgesic requirements at day 1,2 and 3 11. Hospital stay
Notes	Additional information provided by trialists for confirmation of intention to treat analysis and no patient lost to follow‐up and outcomes of re‐operation. The labeling of the tables in the published article was incorrect and the trialists confirmed the correct labelling.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Crevoisier 1998.

Methods	Method of randomisation: opening a pre‐sealed box at the end of the operation before the wound was sutured. Number lost to follow‐up: none Methodological quality score: 6 Length of follow‐up: until discharge from hospital
Participants	Orthopaedic hospitals in Interlaken, Switzerland 66 patients undergoing a hip replacement and 32 patients undergoing a knee replacement Mean age: 72 years (range 36‐87) Percentage male: 41%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Superficial wound infection 2. Deep wound infection 3. Wound haematoma 4. Transfusion requirements 5. Range of movement 6. Hospital stay 7. Fall in haematocrit
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Dhawan 2003.

Methods	Method of randomisation: table of random numbers. Time of randomisation not stated. Number lost to follow‐up: 2/21 (9%). Methodological quality score: 7 Length of follow‐up: 7 days
Participants	Orthopaedic hospitals in Washington, USA 21 patients undergoing arthroscopically assisted anterior cruciate ligament reconstruction Mean age: 28 years (range 19‐36 years) Percentage male: 90%
Interventions	Wound drainage with a suction drain versus no drain (drain placed within the knee joint)
Outcomes	1. Thigh girth 2. Knee flexion 3. Knee extension 4. Pain using visual analogue scale on day 1, 3 and 5
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Duranthon 2000.

Methods	Method of randomisation: envelopes opened in the operating theatre at the time of surgery Number lost to follow‐up: none Methodological quality score: 8 Length of follow‐up: 10 days
Participants	Orthopaedic hospitals in Paris, France 86 patients undergoing bipolar hemiarthroplasty via a posterior approach for treatment of a hip fracture. Mean age: 81 years (range not stated) Percentage male: 22%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Number of patients transfused 2. Re‐operations 3. Change in haemoglobin 4. Patients' temperature
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Esler 2003.

Methods	Method of randomisation: sealed envelopes opened at the time of wound closure. Number lost to follow‐up: not stated. Methodological quality score: 6 Length of follow‐up: 60 months
Participants	Orthopaedic hospitals in Bournemouth, England 100 patients undergoing cemented total knee replacement Mean age: 73 years (range 50‐88 years) Percentage male: 45%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Deep wound infection 2. Number of patients transfused 3. Mean fall in haemoglobin 4. Length of hospital stay 5. Need for manipulation of the knee 6. Range of knee flexion 7. Swelling of the limb 8. Re‐operations 9. Estimated blood loss 10. Blood loss into drains and dressings 11. Knee circumference 12. Time to straight leg raising 13. Soft tissue ecchymosis 14. Number with dressing change 15. Pain scores 16. Post operative pyrexia
Notes	the absence of any re‐operations for wound healing complications were inferred from the published report.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Gartsman 1997.

Methods	Method of randomisation: random number table used at the time of surgery. Number unknown to the surgeon until the time of wound closure Number lost to follow‐up: none Methodological quality score: 8 Length of follow‐up: not stated
Participants	Orthopaedic hospitals in Houston, Texas USA 100 patients undergoing rotator cuff repair, 100 patients undergoing anterior shoulder joint reconstruction for instability and 100 patients having a shoulder arthroplasty. Mean age: 49 years (range 16‐87) Percentage male: 60%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Number of patients transfused 2. Re‐operations 3. Wound infection 4. Wound haematoma 5. Re‐operations for wound healing complications 6. Wound dehiscence 7. Length of hospital stay
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Gonzalez DV 2004.

Methods	Method of randomisation: computer generated table of randomised numbers Number lost to follow‐up: not stated Methodological quality score: 7 Length of follow‐up: 3 months
Participants	Orthopaedic hospitals in Buenos Aires, Argentina 102 patients (104 operations) undergoing a hip replacement Mean age: 63 years (range 24‐87) Percentage male: 53%
Interventions	Wound drainage with 2 suction drains versus no drain
Outcomes	1. Superficial wound infection 2. Deep wound infection 3. Hematoma 4. Transfusion 5. Change in haematocrit 6. Thromboembolic complications 7. Dressing reinforcement 8. Persistent drainage 9. Mortality 10. Re‐operation 11. Mean units blood transfused 12. Hospital stay 13. Change in thigh circumference
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Holt 1997.

Methods	Method of randomisation: even or odd date of operation Number lost to follow‐up: none Methodological quality score: 5 Length of follow‐up: 6 weeks
Participants	Orthopaedic hospitals in Alexandria, VA 136 patients undergoing knee replacement Mean age: 69 years (range 46‐93) Percentage male: 32%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Superficial wound infection 2. Deep wound infection 3. Ecchymosis 4. Deep vein thrombosis 5. Transfusion 6. Fall in haemoglobin 7. Reinforcement of the dressing 8. Arthrofibrosis 9. Range of movement 10. Time to straight leg raising 11. Pain 12. Hospital stay 13. Cost of hospital treatment
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Jenny 2001.

Methods	Method of randomisation: sealed envelopes opened at the time of wound closure. Number lost to follow‐up: none Methodological quality score: 6 Length of follow‐up: 2 weeks
Participants	Orthopaedic hospitals in Illkirch, France 60 patients undergoing total knee replacement Mean age: 70 years (range 62‐84 years) Percentage male: 32%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Total units of blood transfused in each group 2. Number of patients transfused 3. Wound infection 4. Length of hospital stay 5. Re‐operations 6. Estimated blood loss 7. Range of movement on day 2, 7 and 14 8. Swelling of the limb on day 2, 7 and 14 9. Pain on day 2, 7, and 14 10. Stiffness of the knee
Notes	It was inferred from the text that 30 patients were in each group
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Johansson 2005.

Methods	Method of randomisation: sealed envelopes opened at the time of wound closure. Number lost to follow‐up: none Methodological quality score: 9 Length of follow‐up: 2 months
Participants	Orthopaedic Hospitals in Linkoping, Sweden 105 patients undergoing total hip replacement Mean age: 67 years (range 42‐84 years) Percentage male: 50%
Interventions	External wound compression with or without suction drainage
Outcomes	1. Wound infection 2. Blood transfusion 3. Fall in haemoglobin 4. Deep vein thrombosis 5. Persistent wound discharge 6. Re‐operations 7. Calculated blood loss
Notes	The no‐drain group had compression of the wound with an inflatable girdle.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Karahan 2005.

Methods	Method of randomisation: randomisation chart. Number lost to follow‐up: 21% Methodological quality score: 7 Length of follow‐up: 7 days
Participants	Orthopaedic hospital in Istanbul, Turkey 27 patients undergoing arthroscopically assisted anterior cruciate ligament reconstruction Mean age: 29 years (range not stated) Percentage male: 96%
Interventions	Wound drainage with a suction drain versus no drain (The drain was placed at the donor site of the hamstring tendon graft)
Outcomes	1. Knee flexion 2. Knee extension 3. Thigh circumference 4. Leg circumference 5. Visual analogue pain score
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Kim 1998a.

Methods	Method of randomisation: sealed numbered envelopes. Number lost to follow‐up: none Methodological quality score: 10 Length of follow‐up: 12 months
Participants	Orthopaedic hospitals in Seoul, Korea 48 patients undergoing simultaneous bilateral hip replacement. All operations performed by one surgeon Mean age: 48 years (range 22‐76) Percentage male: 21%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Superficial wound infection 2. Deep wound infection 3. Wound haematoma (assessed by ultrasound) 4. Area of erythema 5. Area of ecchymosis 6. Transfusion requirements 7. Deep vein thrombosis 8. Bleeding into the wound dressing and reinforcement 9. Range of movement 10. Pain 11. Hip score 12. Re‐operations
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Kim 1998b.

Methods	Method of randomisation: sealed numbered envelopes. Number lost to follow‐up: none Methodological quality score: 9 Length of follow‐up: 12 months
Participants	Orthopaedic hospitals in Seoul, Korea 69 patients undergoing simultaneous bilateral knee replacement. All operations performed by one surgeon Mean age: 64 years (range 37‐80) Percentage male: 10%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Deep wound infection 2. Re‐operations 3. Need for dressing reinforcement 4. Persistent drainage at 24 hours and 6 days 5. Ecchymosis 6. Time to return of quadriceps function 7. Range of movement 8. Pain
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Lang 1998.

Methods	Method of randomisation: opening a sealed envelope containing a computer generated random number. Number lost to follow‐up: not stated Methodological quality score: 6 Length of follow‐up: minimum of six weeks
Participants	Orthopaedic hospitals in Charlotte, North Carolina, USA 202 patients undergoing a non‐emergency surgery for fixation to a fracture or bone grafting procedure. Compound fractures and contaminated surgical sites excluded as well as surgery to the hands and feet. Total of 287 wounds randomised Mean age: 38 years (range 18‐91) Percentage male: 30%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Superficial wound infection 2. Wound haematoma 3. Wound dehiscence 4. Re‐operation for wound healing 5. Reinforcement of the dressing
Notes	Main report of the study was in an article in 1998. Two reports in 1997 were preliminary results.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Lautenbach 2000.

Methods	Method of randomisation: not stated. Number lost to follow‐up: None Methodological quality score: 2 Length of follow‐up: 2 weeks
Participants	Orthopaedic hospitals in Johannesburg, South Africa 51 patients having surgery for hip fracture (type of surgery not specified) Mean age: not stated (range not stated) Percentage male: not stated
Interventions	Wound drainage with a suction drain versus no drain (the suction drainage group had four different types of drainage size, (3 mm or 6 mm lumen) or a double drainage tube with a bolus of streptokinase or a four hourly bolus of streptokinase
Outcomes	1. Superficial wound infections 2. Deep wound infections 3. Deep vein thrombosis 4. Volume of wound haematoma (as measured by ultrasound)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Leb 1995.

Methods	Method of randomisation: not stated Number lost to follow‐up: not stated Methodological quality score: 2 Length of follow‐up: 3 months
Participants	Orthopaedic hospitals in Cleveland, Ohio, USA 100 patients undergoing primary knee replacement Mean age: not stated (range not stated) Percentage male: not stated
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Major and minor wound healing complications 2. Days of wound drainage 3. Range of movement 4. Grading of knee function 5. Change in haematocrit
Notes	Reported only as a conference abstract
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Mengal 2001.

Methods	Method of randomisation: at the time of wound closure, method not stated. Number lost to follow‐up: not stated Methodological quality score: 6 Length of follow‐up: not stated
Participants	Orthopaedic hospitals in Liege, Belgium 152 patients undergoing total hip replacement and 104 patients undergoing total knee replacement Mean age: 67(hip) and 72 (knee) years (range 44‐89) Percentage male: 38%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infection 2. Thromboembolic complications 3. Mean volume of blood transfused 4. Re‐operations 5. Estimated blood loss 6. Limb swelling 7. Functional recovery
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Murphy 1993.

Methods	Method of randomisation: sealed envelopes opened at the time of wound closure. Number lost to follow‐up: None Methodological quality score: 5 Length of follow‐up: 10 days
Participants	Orthopaedic hospitals in London, England 40 patients undergoing total hip replacement Mean age: not stated (range not stated) Percentage male: not stated
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infection 2. ASEPSIS wound score at 2, 4, 5, 7 and 10 days 3. Blood loss
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Niskanen 2000.

Methods	Method of randomisation: not stated Number lost to follow‐up: not stated Methodological quality score: 5 Length of follow‐up: 2 months
Participants	Orthopaedic hospitals in Lahti, Finland 58 patients undergoing total hip replacement for arthritis and 39 undergoing knee replacement. Mean age: 70 years (range 48‐89) Percentage male: 32%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Superficial wound infection 2. Deep wound infection 3. Wound haematoma 4. Mean units blood transfused 5. Fall in hematocrit 6. Thromboembolic complications 7. Prolonged oozing from the wound 8. Need for reinforcement of the dressing 9. C reactive protein 10. Range of limb movement
Notes	Discrepancies were present between the abstract and full article for the number of superficial would infections and transfusion requirements. The data from the full report were used.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Nixon 2000.

Methods	Method of randomisation: not stated. Number lost to follow‐up: not stated. Methodological quality score: 2 Length of follow‐up: 4 years
Participants	Orthopaedic hospitals in Belfast, Northern Ireland 1293 patients undergoing total hip and knee replacements Mean age: not stated (range not stated) Percentage male: not stated
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infections (superficial and deep) 2. Ooze from the wound
Notes	This large study was only reported as a conference abstract with very limited reporting of outcomes. Contact with the trialists has indicated that the data for this study has largely been lost. Despite this problem we have included this study as there is data available for the outcome of wound infection. In addition this study is one of the few studies with a more prolonged follow‐up of a significant number of participants to be able to give data on the outcome of wound infection after the initial peri‐operative period. The trial report stated that 1293 patients had been randomised. Contact from the trialists indicated that numbers of patients in each group were the same and therefore we have assumed for the analysis 646 patients in each group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Ovadia 1997.

Methods	Method of randomisation: sealed envelopes opened at the time of wound closure. Number lost to follow‐up: none Methodological quality score: 9 Length of follow‐up: until discharge from hospital
Participants	Orthopaedic hospitals in Tel Aviv, Israel 30 patients undergoing total hip replacement and 58 patients undergoing total knee replacement Mean age: 71 years (range not stated) Percentage male: 32%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infection 2. Number of patients transfused 3. Length of hospital stay 4. Range of knee flexion (knee replacements only) 5. Excessive swelling of the limb 6. Serious leakage from the wound 7. Fall in haemoglobin 8. Number of patients transfused 9. Number of units blood transfused
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Paessler 1997.

Methods	Method of randomisation: toss of a coin. Number lost to follow‐up: none Methodological quality score: 3 Length of follow‐up: 2 weeks
Participants	Orthopaedic hospitals in Heidelberg, Germany 129 patients undergoing open anterior cruciate ligament reconstruction via a mini arthrotomy and patella tendon graft. 10 patients had bilateral surgery so 139 cases were included. Mean age: 27 years (range not stated) Percentage male: not stated
Interventions	Wound drainage with a suction drain versus no drain Drain inserted into the knee joint
Outcomes	1. Blood loss (from drains and aspiration) 2. Frequency of aspiration 3. Pain score 4. Demand for analgesia
Notes	Reported only as a conference abstract
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Payne 1996.

Methods	Method of randomisation: Opening of a sealed envelope during surgery. Number lost to follow‐up: 2% Methodological quality score: 5 Length of follow‐up: 2 weeks
Participants	Orthopaedic hospitals in Michigan, USA 205 patients undergoing single level lumbar laminectomy Mean age: not stated (range not stated) Percentage male: not stated
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Superficial wound infection
Notes	Only one outcome measure reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Ravikumar 2001.

Methods	Method of randomisation: sealed envelope opened in the operating theatre at the time of wound closure. Number lost to follow‐up: not stated Methodological quality score: 6 Length of follow‐up: 6 week
Participants	Orthopaedic hospitals in Tunbridge Wells, UK 23 patients (25 wounds) undergoing hip replacement surgery. Mean age: 77 years (range 39‐97) Percentage male: 48%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infection 2. Re‐operations 3. Change in haemoglobin 4. Wound bruising 5. Pain scores 6. Wound haematoma
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Ritter 1994.

Methods	Method of randomisation: not stated Number lost to follow‐up: not stated Methodological quality score: 4 Length of follow‐up: not stated ? for period in hospital only
Participants	Orthopaedic hospitals in Mooresville, USA 275 patients undergoing primary knee replacement and 140 undergoing total hip replacement Mean age: not stated (range not stated) Percentage male: not stated
Interventions	Wound drainage with a suction drain versus no drain Drains placed below the deep fascia and left in place for 24 hours
Outcomes	1. Superficial wound infection 2. Deep wound infection 3. Wound haematoma 4. Transfusion requirements 5. Post‐operative haemoglobin 6. Thromboembolic complications 7. Immobilisation for excessive bleeding 8. Range of movement
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Sasso 1998.

Methods	Method of randomisation: opening an envelope containing a card. Number lost to follow‐up: not stated Methodological quality score: 6 Length of follow‐up: one year
Participants	Orthopaedic hospitals in Chicago, Illinois, USA 108 patients (112 cases) undergoing surgery to a traumatic injury of the spine in which bone graft was taken from the Iliac crest One patient from whom the drain fell out was excluded from the analysis of results Mean age: 38 years (range 13‐79) Percentage male: 68%
Interventions	Wound drainage with a single suction drain versus no drain Drain left in place for 2‐5 days
Outcomes	1. Wound infection (as no implants were used all infections were considered as superficial) 2. Re‐operation
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Straw 2003.

Methods	Method of randomisation: last digit of hospital number Number lost to follow‐up: 2(4%) Methodological quality score: 6 Length of follow‐up: six months
Participants	Orthopaedic hospitals in Derby, UK 47 patients undergoing arthroscopically assisted anterior cruciate ligament reconstruction with a patella tendon autograft. Mean age: 27 years (range 16‐49 years) Percentage male: 61%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infection 2. Knee circumference 3. Range of movement 4. Muscle strength 5. Visual analogue pain score 6. Extension deficit
Notes	Two patients were not followed‐up and were excluded from the presentation of the results which relate to 47 patients.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Tjeenk 2005.

Methods	Method of randomisation: sealed opaque opened at the end of the operation. Number lost to follow‐up: none Methodological quality score: 9 Length of follow‐up: length of hospital stay
Participants	Orthopaedic hospital in The Hague, The Netherlands 200 patients with a hip fracture undergoing either sliding hip screw fixation (88 cases) or hemiarthroplasty (102 cases) Mean age: 82 years (range not stated) Percentage male: 25%
Interventions	Wound drainage with a single suction drain versus no drain
Outcomes	1. Superficial wound infection 2. Deep wound infection 3. Haematoma 4. Transfusion 5. Haemoglobin fall 6. Mortality 7. Re‐operations 8. Hospital stay
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Varley 1995.

Methods	Method of randomisation: opening a numbered sealed envelope immediately prior to wound closure in the operating theatre. Number lost to follow‐up: not stated Methodological quality score: 7 Length of follow‐up: six months
Participants	Orthopaedic hospitals in Nottingham, England 177 patients with a hip fracture undergoing either sliding hip screw fixation (91 cases) or hemiarthroplasty (86 cases) Mean age: 80 years (range not stated) Percentage male: 22%
Interventions	Wound drainage with a suction drain versus no drain
Outcomes	1. Wound infection 2. Wound haematoma (measured by ultrasound) 3. ASEPSIS score of wound healing 4. Wound dehiscence 5. Transfusion requirements 6. Re‐operations 7. Fall in haemoglobin
Notes	Main report of trial in paper of 1995. Also reported as three conference abstracts and one study in 1994 on the use of ultrasound to assess the presence of wound haematomas
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Walmsley 2005.

Methods	Method of randomisation: opening a sealed envelope at the end of the procedure Number lost to follow‐up: not stated Methodological quality score: 10 Length of follow‐up: six months
Participants	Orthopaedic hospital in Fife, Scotland 553 patients undergoing hip replacement surgery (577 wounds) as 25 patients had bilateral surgery. Mean age: 68 years (range 48‐95 years) Percentage male: 37%
Interventions	Wound drainage with a single suction drain versus no drain.
Outcomes	1. Wound infection 2. Deep sepsis 3. Haematoma 4. Transfusion 5. Fall in haemoglobin 6. Thromboembolic complications 7. Mortality 8. Re‐operation 9. Dislocation 10. Hospital stay 11. Harris Hip Score
Notes	Three deep wound infections (1 in drain group, 2 in no drain group) were reported after the six month follow‐up period and were not incuded in the analysis.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Widman 2002.

Methods	Method of randomisation: opening a sealed envelope. Time of randomisation not stated Number lost to follow‐up: not stated Methodological quality score: 3 Length of follow‐up: not stated
Participants	Orthopaedic hospitals in Stockholm, Sweden 22 patients undergoing hip replacement surgery. Mean age: 71 years (range 49‐88 years) Percentage male: 41%
Interventions	Wound drainage with two suction drains versus no drain
Outcomes	1. Wound infection 2. Wound haematoma (measured by erythrocyte scintigraphy) 3. ASEPSIS score of wound healing 4. Total recorded blood loss 5. Transfusion requirements (number of patients transfused and mean volume blood transfused)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Ashraf 2001	This was a comparative study of 206 patients with primary or revision hip and knee arthroplasty with and without drains. A greater tendency of wound oozing was seen in revision hip arthroplasty performed with postoperative suction drainage. The study was excluded as it was a comparative study with no randomisation of patients.
Beer 1991	This was a randomised study of 24 patients undergoing simultaneous bilateral hip replacement and 76 patients undergoing bilateral knee replacement from an Orthopaedic hospital in Columbus, Ohio, USA. The right limb for all cases had wound drainage and the left limb no drainage. There were no wound infections in either group and no difference was reported for the outcome measures of swelling of the limb, range of movements or pain. The study was excluded because of the method of 'randomisation', which was felt to be inadequate and may have resulted in bias to occur within the two patient groups.
Berman 1990	This was a randomised trial for 126 consecutive orthopaedic wounds in which drainage was either with a continuous vacuum system (VariDyne) and an intermittent spring type system (Hemovac). The study was excluded as there was no group in which drains were not used.
Bryan 1969	This study involved patients treated by open meniscectomy of the knee. 45 were treated using suction wound drainage and post‐operative cast immobilisation, 54 to cast immobilisation only, 39 to drainage only and 25 to no cast immobilisation and no drainage. The text of the article stated that allocation to the drainage and cast immobilisation group was 'random', but later in the text it stated that patients could not be allocated to each group at random. The patients were operated on by a number of surgeons, some of which did not use all the four methods of treatment. The study was excluded, as the method of allocation to the different groups was not considered to be a random method. In addition, open meniscectomy is an operation that is now performed less frequently since the introduction of arthroscopy of the knee. The conclusion of the study was that suction drainage conferred no benefit.
Gehrig 2005	This study compared a novel external suction drainage system with a conventional internal suction drainage system in patients undergoing primary total hip or knee arthroplasties. The study was excluded as there was no group in which drains were not used.
Gerngross 1992	This study was a randomised trial of 40 patients having intra‐articular knee drainage after surgery. The study compared three different vacuum pressures. The study was excluded, as there was no group without drains.
Graupe 1996	This study of 44 patients undergoing elective hip joint operation involved randomisation to either a new slit‐suction drainage or regular Redon‐suction drainage. The study was excluded as there was no group without drains.
Kohler 2000	This was a randomised trial of 257 patients comparing five different types of suction drainage. The five types were gravity drainage, high pressure, middle pressure, low pressure and standard Recon drainage. The study was excluded, as there was no group in which no drains were used.
Labek 1998	This study published in German compared within a randomised trial the use of one, two or three drains in total hip replacement. There was no group in which no drains were used, therefore the study was excluded.
Mac 1991	This was a trial of 91 patients undergoing total joint arthroplasty of the knee or hip. Patients were randomised to an autoreinfusion (Constavac) device or standard suction (Hemovac), in which blood drainage is discarded, in patients. The study was excluded as there was no group in which drains were not used.
Mah 1996	This was a randomised study of three types of wound drainage system for 30 uncemented total knee replacements and 30 uncemented total hip replacements. Patients were randomised to either a gravity drainage system, a constant suction system drainage system or a standard charged redivac bottle. The study was excluded as there was no group in which drains were not used.
Martin 2004	This was a comparative study involving total knee arthroplasty. 50 patients had three wound drains and autotransfusion, 50 had no wound drainage and 50 had one intraarticular wound drain. The group with autotransfusion had the highest blood loss. The study was excluded as there was no randomisation of patients.
Nicolajsen 1996	This study was only reported as a conference abstract involving 80 patients with a hip fracture treated operatively. Patients were randomised to have either no drains or closed suction drains. No difference was noted for differences in temperature, sedimentation reaction, haemoglobin or leucocyte levels, signs of wound infection, haematoma or the need for transfusion. The study was excluded as no data for the outcomes listed above or patient details were given.
Padala 2003	This study compared the role of drains and tourniquets in primary total knee replacement for 115 participants. One group of 60 participants were randomised to have surgery with no tourniquet, adrenaline and saline infiltration of the wound and no drain. The other group of 55 participants had a tourniquet, no adrenaline infiltration and a drain. Transfusion requirements were less in the group treated without drains (3% versus 24% participants transfused). This study was excluded as there were three differences in treatment methods between the two groups and it was not a trial of drain versus no drain.
Pomeroy 1995	This study, reported only as a conference abstract, used a 'random selection method' to allocate 65 knee arthroplasty cases to suction drainage and 46 to no drains. Results were reported without data as showing no statistical difference between groups for the complications of infection, swelling, haemoglobin levels and flexion. The study was excluded, as there was inadequate reporting of trial results.
Reilly 1986	This was a comparison of 170 knee replacements with 129 without drainage. It was excluded as it was not a randomised trial.
Sen 2005	This study aimed to evaluate the role of closed suction drainage on the extent of epidural fibrosis after lumbar disc surgery. 79 patients were included in the study, with 41 having suction drainage and 38 having no drains. Epidural fibrosis and clinical outcomes were reported as better in the drain group. The study was excluded as there was no mention that the choice of using drains was randomised.
Seyfert 2002	This was a randomised study of 116 knee replacements. All patients had two wound drains inserted, 57 with and 59 without suction. The use of suction had no significant impact on the post‐operative blood loss and the postoperative course. This study was excluded as there was no group without drains.
Tatari 2005	This study reported experience of 190 patients with the use of closed suction drainage in arthroscopic knee procedures. There were three subgroups: Tourniquet and no drain, infusion pump and drain and infusion pump and no drain. The study was excluded as it was an observation study with no randomisation of participants.
Trammell 1991	This was a trial of 96 patients undergoing total joint arthroplasty or spinal reconstructive surgery. Patients were randomised to receive either a Solcotrans Plus or a Stryker‐CBC ConstaVAC closed‐wound drainage system. The study was excluded as there was no group in which drains were not used.
Tsumara 2006	This was a randomised trial of drains after total knee replacement for 212 participants. One group had the drains clamped after intra‐articular injection of saline with 1:500000 adrenaline and the other had post‐operative blood salvage. Drain clamping was more effective in reducing blood loss. The study was excluded as both groups received closed suction drainage.
Willemen 1991	This was a randomised trial of closed suction drainage after knee arthroplasty in which drains were allocated to be removed at either 24 hours or 48 hours. The study was excluded, as there was no comparison group allocated not to receive drains.
Zamora‐Navas 1999	This randomised trial compared the removal of closed suction drains at 12, 24 or 48 hours for knee arthroplasty. There was no group in which no drains were used, therefore the study was excluded.

Contributions of authors

Martyn Parker initiated and designed the review, read all studies, extracted data, contacted trialists and compiled the first drafts. Rupert Clifton read all the studies for the first update, extracted data and read through and checked all text and results for the initial review. Andrew McKee read all the newly identified studies, extracted data and read through and checked all text and results. Vicki Livingstone provided statistical support and advice regarding the identified studies. Martyn Parker is the guarantor of the review.

Sources of support

Internal sources

• Peterborough and Stamford Hospitals NHS Foundation Trust, Peterborough, UK.

• University of Otago, Dunedin, New Zealand.

External sources

• No sources of support supplied

Declarations of interest

None known.

Edited (no change to conclusions)