Summary of main results

We included eight studies in this updated review, with a total of 3680 participants. From these studies, we found that the incidence of deep venous thrombosis (DVT), diagnosed by compression ultrasound, venography, or both, in participants with a leg injury who were immobilized in a plaster cast or brace for at least one week and received no thromboprophylaxis (or placebo) was 4.3% to 40%. This was significantly higher than for participants who received daily subcutaneous injections of low molecular weight heparin (LMWH) during the entire period of immobilization (0% to 37%).

Comparable results were seen in the following groups of participants: patients with a below‐knee cast, surgically treated patients, conservatively treated patients (not surgically treated), patients with fractures, patients with soft‐tissue injuries, and patients with proximal or distal DVT. The odds ratios (OR) between the individual participant groups were similar, with an overlap of the confidence intervals (CI). Therefore, it was not possible to indicate a participant or patient group where prophylaxis was not indicated. No clear differences were found for PE between LMWH and the control groups. The studies found less symptomatic venous thromboembolism (VTE) in the LMWH groups compared with the control groups. One death (in the control group) was reported in the included studies. Major adverse events were rare; the main adverse events reported were cases of minor bleeding.

Overall completeness and applicability of evidence

The studies included in this review all assessed participants with lower‐limb immobilization. Some studies only included conservatively treated participants, whereas, others also included surgically treated participants. Studies included participants with both fractures and soft‐tissue injuries. These differences in inclusion criteria may have led to a lower external validity.

In addition to using Cochrane's 'Risk of bias' tool to assess methodological quality, we also assessed the validity and quality of the included randomized controlled trials using the scoring scheme provided on the website of Cochrane Netherlands (www.cochrane.nl). The overall quality of the eight included studies according to this scoring system, was rated 'good', although some remarks should be added.

The true thromboembolic rate in unprotected patients remains unknown, as high‐risk patients were excluded from participation in all eight studies, underestimating the incidence of thromboembolism in unprotected patients and the potentially beneficial effect of LMWHs. Although there is consensus among surgeons that thromboprophylaxis should be initiated in patients with a moderate to high risk for thromboembolism, there is no uniform definition of this patient group, leading to practice variation. Several attempts have been made to stratify the risk for VTE of immobilized patients (Gearhart 2000; Zagrodnick 1990). Knudson 2004 developed a scheme, based on analysis of 1602 episodes of VTE using the National Trauma Data Bank, to identify trauma patients with a high risk of thrombosis. Lower‐limb fracture and age over 40 years were among the factors used for selection. However, this study also stated that with this scheme, 90% of patients with VTE would be identified, and 10% would be missed. By contrast, in another study, patients under 40 years of age, with soft‐tissue injuries also developed DVT (Kock 1995). Up to now, no stratification method has proven its superiority or has gained general acceptance. In most studies in this review, investigators did not include patients with a high risk of DVT. So even in patients with a low to average risk of DVT, LMWH still provided significant protection.

In five studies, the included participants had a wide variety of trauma, ranging from fractures to soft‐tissue injuries, and tendon ruptures (Jorgensen 2002; Kock 1995; Kujath 1993; Lassen 2002; Van Adrichem 2017). Since the risk of DVT is found to be related to the presence and type of fracture, the extent of soft‐tissue injury, and type and duration of surgery, heterogeneity might have been introduced. However, randomization should have minimized this effect. Three studies focused on participants with a specific trauma (Bruntink 2017; Lapidus 2007a; Lapidus 2007b). We also observed a high drop‐out rate.

Five studies used ultrasound (Bruntink 2017; Kock 1995; Kujath 1993, Lapidus 2007a; Van Adrichem 2017), and three used venography (Jorgensen 2002; Lapidus 2007b; Lassen 2002), to diagnose DVT. The latter is considered the 'gold standard', but is rarely used in routine practice as the first line of investigation for DVT. Duplex ultrasonography and compression ultrasonography have a lower sensitivity and specificity compared to venography, especially for diagnosing calf vein thrombosis (CBO 2008; Lensing 1989). This might be the reason for the differences in the results between Kock 1995 and the studies that used venography (Jorgensen 2002; Lapidus 2007b; Lassen 2002), since there were no other major differences reported in participant characteristics or intervention. However, it does not explain the differences between the results of Kock 1995, Kujath 1993 and Lapidus 2007a. By only examining participants reporting symptoms, Van Adrichem 2017 only reported clinically‐relevant DVTs, therefore underestimating the number of DVTs in the study population. Due to this difference in study protocol, we only used the results of this study in the analyses for pulmonary embolism (Analysis 1.9) and symptomatic VTE (Analysis 1.10).

Lassen 2002 included participants who received up to four days of LMWH (32% of participants). Another study treated all participants with LMWH for one week prior to randomization (Lapidus 2007b). This might equalize the effects in the two groups and lead to an underestimation of the treatment effect. Therefore, Lapidus 2007b only focused on the duration of treatment, and not on the indication of treatment itself.

The dose of 3500 anti‐Xa IU of tinzaparin that Jorgensen 2002 used might have been too low, since another study showed equal antithrombotic effect using 4500 anti‐Xa IU of tinzaparin compared with 40 mg of enoxaparin, which is the standard dose in orthopedic surgery (Eriksson 2001). This possibly creates an underestimation of the effect of prophylaxis.

Quality of the evidence

The quality of evidence according GRADE varied by outcome and ranged from low to moderate with reasons for downgrading being risk of bias due to attrition and performance bias (DVT), imprecision and risk of bias due to attrition and performance bias (PE, symptomatic VTE and adverse outcomes) and low number of events and imprecision (mortality due to other causes).

See also summary of findings Table for the main comparison

Potential biases in the review process

Two review authors independently carried out study selection, data extraction, and study quality assessment in order to reduce bias and subjectivity. We are confident that all potential studies are included in this review. However, the possibility remains that relevant data exist which have not been published or were not found in the search.

The total number of included studies in this review was eight. Since ten studies are required to perform a funnel plot analysis, reliable funnel plot analysis could not be performed.

By only examining patient reporting symptoms Van Adrichem 2016 only reported clinical relevant DVTs, therefore underestimating the number of DVTs in the study population. Due to this difference in study protocol the results of this study were only used in the analyses for 'pulmonary embolism' (Analysis 1.9) and 'symptomatic VTE' (Analysis 1.10).

Agreements and disagreements with other studies or reviews

As early as 1944, the first study on deep venous thrombosis (DVT) following leg injuries was published, reporting incidence rates of 7% to 18% (Bauer 1944). Nonfatal and fatal pulmonary embolism (PE) complicate DVT of the lower extremities. Fatal PE used to be a common cause of death in hospitals, because of the often clinically occult nature of DVT.

The discussion on the use of LMWH in immobilization of the lower limb focuses on the following issues: the reduction of symptomatic VTE, the reduction of asymptomatic DVT, the relevance of asymptomatic DVT, and the incidence of complications. With meta‐analyses, we were able to show that the use of LMWH in immobilization of the lower limb following leg injury results in a reduction of symptomatic VTE and of (asymptomatic) DVT.

Over 80% of DVTs diagnosed were located distally. Calf vein thrombosis propagates and becomes proximal in between 0% and 25% of patients, accounting for a mean of 10%. Up to 10% of proximal DVTs embolize massively, and are potentially fatal (Anonymous 1986; Schellong 2007). Opinions differ about the risk of post‐thrombotic syndrome (PTS) after distal DVT. It is stated that the risk is considerably lower than in proximal DVT. Results are inconsistent for the relationship between the location of the initial thrombus and the subsequent development of PTS. Some prospective studies reported rates of PTS after distal DVT that were as high as 20% to 80% (McLafferty 1998; Schulman 1986). Hence, distal DVT appears to be associated with a substantial risk of subsequent PTS. Further research is indicated to elicit the exact role of distal DVT in the development of PTS (Kahn 2006).

The incidence of complications in the review seemed to be low compared with data in the literature. Major bleeding was reported in 0.27% (two out of 1469 participants) and minor bleeding in up to 7.8% of participants (Van Adrichem 2017). In contrast, Reilmann 1993 reported up to 14% of participants with hematomas due to injections. Another study even described it up to 28%, although some of the participants were not treated with LMWH, but with unfractionated heparin (Zagrodnick 1990). In our seven included studies, no cases of heparin‐induced thrombocytopenia were described, which is in concordance with the observation that this condition is seldom seen in combination with LMWH (Bloemen 2012).

The eight studies used different types of LMWH. The total number of participants was insufficient to evaluate which type of LMWH to choose. Evidence published so far indicates that any differences between LMWH preparations, if they exist, must be extremely small (Geerts 2004). It is unlikely that a properly sized and designed study comparing the various LMWHs will ever take place. Sample size calculations quickly reach over 10,000 if one considers appropriate definitions of 'non‐inferiority' when comparing the various antithrombotic regimens (Vaitkus 2004).

Venography is considered the most accurate method of diagnosing DVT but is now rarely used in clinical practice (Abelseth 1996; Bergqvist 2002). It is an invasive procedure, and there is a reported incidence of serious adverse reactions to the contrast media in the range of 0.4% to 2% (Lensing 1990). Duplex ultrasound is the most common non‐invasive test used to diagnose venous thrombosis of the extremities. Compared with venography, ultrasound has been shown to be reliable in the diagnosis of proximal symptomatic DVT, with a sensitivity and specificity of over 90%. However, for distal thrombosis, a sensitivity of 73% can be reached when combining compression ultrasound with color‐doppler ultrasound (CBO 2008). Considering the properties of the imaging methods used to diagnose PE, a recent systematic review showed a sensitivity of 86% for the ventilation‐perfusion scintigraphy and a specificity of 46% compared with pulmonary angiography. For the CT scan, the percentages were 85% sensitivity and 94% specificity (Hayashino 2005). Consequently, the number of instances of missed DVT and PE seems to be acceptable when including studies with different diagnostic procedures, and should not influence our conclusions to a major extent. However, an underestimation of the incidence of VTE can occur when using ultrasound, CT‐scan, or ventilation‐perfusion scanning.

Increased incidence of DVT in an untreated population calls for measures, and thromboprophylaxis seems necessary if immobilization in a cast or brace is needed. Post‐traumatic DVT can lead to PE or long‐term damage in the form of PTS (Kakkar 1994). Low molecular weight heparin has been shown to be effective in reducing the incidence of VTE. The possible complications of major bleeding events (2/750 participants or 0.3%), and heparin‐induced thrombopenia (none in this review) have been shown to be extremely rare, and do not outweigh the beneficial effect of the reduction of thromboses.