Systematic review and meta-analysis on the rate of postoperative venous thromboembolism in orthopaedic surgery in Asian patients without thromboprophylaxis

Abstract

Background:

Postoperative venous thromboembolism (VTE) is a common life-threatening complication after surgery. This review analysed the rate and mortality of VTE after orthopaedic surgery in Asia.

Methods:

Inclusion criteria were: prospective study; deep vein thrombosis (DVT) diagnosed by venography or ultrasonography; hip fracture surgery (HFS), total hip arthroplasty (THA) or total knee arthroplasty (TKA); and no thromboprophylaxis. The pooled proportion was back-calculated by Freeman–Tukey variant transformation, using a random-effects model.

Results:

Twenty-two studies (total population 2454) published from1979 to 2009 were included. Using venography, the pooled rates of all-site, proximal, distal and isolated distal DVT were 31·7, 8·9, 22·5 and 18·8 per cent respectively. With duplex ultrasonography, the respective rates were 9·4, 5·9, 5·9 and 5·8 per cent. After THA or HFS, using venography, the pooled rates of all-site and proximal DVT were 25·8 and 9·6 per cent; with ultrasonography, the respective rates were 10·8 and 7·2 per cent. In TKA groups, using venography, the pooled rates of all-site and proximal DVT were 42·5 and 8·7 per cent; with ultrasonography, the respective rates were 9·5 and 5·2 per cent. The overall pooled rates of symptomatic DVT and symptomatic pulmonary embolism (PE) were 4·5 and 0·6 per cent. No patient died from PE (pooled rate 0·2 per cent).

Conclusion:

None of these Asian patients undergoing orthopaedic surgery died from VTE. Pooled rates of proximal and symptomatic DVT were lower than in Western reports. Copyright © 2011 British Journal of Surgery Society Ltd. Published by John Wiley & Sons, Ltd.

Introduction

In Western countries, venous thromboembolism (VTE) is a major preventable perioperative complication; the greatest risk follows major orthopaedic surgery. The benefit of perioperative thromboprophylaxis has been well established in many cost–efficacy analyses1–3. It is generally believed that the rate of VTE is low in Asia and prophylaxis is seldom used4. However, there are opposing views4, 5. It is important to know the rate of postoperative VTE before implementing routine prophylaxis. If the VTE rate is low, prophylaxis will result in unnecessary costs and complications.

This study aimed to review the pooled rate of deep vein thrombosis (DVT), pulmonary embolism (PE) and death from PE after hip and knee surgery in Asian patients. Changing rates of VTE over time and variation among Asian ethnic groups were also explored.

Methods

A literature search was undertaken using MEDLINE, Embase, the Cochrane Library and KoreaMed for articles published between January 1979 and August 2009. The keywords used were: Asia, Asian, operative, postoperative, venous thromboembolism, deep vein thrombosis, pulmonary embolism, orthopaedic, total hip replacement/ arthroplasty, total knee replacement/arthroplasty, hip fracture and the names of Asian countries. Western Asian and Arabian reports were not included in this systematic review. Reference lists were searched from any articles identified.

The abstract or full text from the relevant article was reviewed, and selection for the review was done according to the following criteria: prospective study design; rate of DVT diagnosed by routine venography or ultrasonography in all the studied population; and inclusion of consecutive patients who underwent hip fracture surgery (HFS), total hip arthroplasty (THA) or total knee arthroplasty (TKA). Studies that employed medical thromboprophylaxis or pneumatic compression devices were excluded. Studies using basic means to reduce VTE such as early mobilization, elastic bandages or compression stockings were allowed. Trials that compared prophylaxis with no prophylaxis were included, but only data concerning the control population (without prophylaxis) were analysed. Non-English publications were included and translated. When more than one study shared the same author(s) or came from the same institute with overlapping recruitment, only the latest publication was included to prevent double counting.

To look for trends in the rate of VTE over time, each study was allocated according to the mid-year of patient recruitment. If specified in the paper, the ethnicity of the patients was recorded. If not, and the study originated from a region predominantly comprising one ethnic group, this was used in the classification. Thus, patients in Taiwan and Hong Kong were categorized as Chinese, those in Korea as Korean, those in Japan as Japanese, those in Thailand as Thai, and those in India as Indian. For countries with a multiethnic population, such as Singapore, the ethnicity of the population was categorized according to the name of the country.

Data collection

All data were abstracted by two independent authors using standard reporting forms. The data included: preoperative venous imaging, postoperative venous imaging, investigation for symptomatic DVT (defined examination for symptoms of DVT), investigation for symptomatic PE (all patients with suspected PE had an objective diagnostic test), symptomatic DVT and PE rates, death from PE, death from other causes, interval of patient recruitment, specific inclusion and exclusion criteria, race of studied population, mean (range) age of population, and evidence of follow-up.

The results of imaging by venography and duplex ultrasonography were summarized separately. Proximal DVT was defined as thrombosis in iliac, femoral and/or popliteal veins; distal DVT was defined as thrombosis distal to the popliteal vein, including muscular calf veins. Any preoperative DVT, if identified, was not included in calculation of the postoperative DVT rate. Data from venographic and ultrasound studies were pooled for analysis of the rate of symptomatic VTE and mortality. The diagnosis of symptomatic PE and death from PE was based on the definition used in each study; however, that had to include specific evidence for the diagnosis such as imaging or autopsy result. Zero PE mortality was assumed if the study stated specifically that there was no overall mortality, or no symptomatic PE.

Statistical analysis

An estimation of pooled risk was determined by summation of the number of events that had occurred in all studies and the number of patients at risk. If one of the outcomes had not been determined specifically in a study, the patients in that study were not considered to be at risk for that particular outcome. For example, if an author reported on the rate of total DVT without distinguishing proximal from distal DVT, the data were used to calculate the overall risk of DVT alone. The proportional data (rate) were transformed into quantity (Freeman–Tukey variant of the arcsine square root transformed proportion). The pooled proportion (rate) was calculated as the back-transformation of the weighted mean of the transformed proportions, using the DerSimonian–Laird weighted random-effects model.

Heterogeneity was represented as the I² value. The calculation was performed using StatsDirect version 2.7.8 (StatsDirect, Altrincham, UK). The χ² statistic was used for comparison of risk of DVT over time, and risk among different ethnic groups.

Results

A total of 35 reports on postoperative VTE after orthopaedic surgery were found5–39 (Fig.1). A paper in Korean by Kyu and co-workers21 was translated into English. Nine studies29–32, 34–38, including a multinational study by Piovella and colleagues34, were excluded because of shared authors and timing of studies. A multinational study by Leizorovicz et al.5 did not contain routine investigation for DVT and was excluded. Three other studies were excluded: one provided intermittent pneumatic compression28, one had inadequate diagnosis for DVT33, and one had inadequate results (VTE was not separated into DVT and PE)9. This left 22 studies with a total population of 2454 patients, published between 1979 and 2008, for analysis6–27. The populations included Chinese (575, 23·4 per cent), Korean (572, 23·3 per cent), Japanese (525, 21·4 per cent), Singaporean (413, 16·8 per cent), Thai (313, 12·8 per cent), Indian (41, 1·7 per cent) and Malay (15, 0·6 per cent).

Figure 1. PRISMA diagram showing article search. DVT, deep vein thrombosis; VTE, venous thromboembolism; HFS, hip fracture surgery; THA, total hip arthroplasty; TKA, total knee arthroplasty

A study by Fuji and colleagues12 involved multiple centres within the same country (Japan), whereas the others were single-hospital studies. Of the 22 studies, 13 used venography6, 9, 10, 12, 13, 16, 17, 20–22, 25–27 and nine7, 8, 11, 14, 15, 18, 19, 23, 24 duplex ultrasound imaging for the diagnosis of DVT. One study used radionuclide imaging and duplex ultrasonography in the same population8; only the ultrasound data were analysed. The total populations of the venography and ultrasound groups were 1733 and 721 patients respectively. Five studies were controlled trials designed to test the efficacy of medical thromboprophylaxis11, 12, 14, 15, 26. The age of the patients was available in 17 studies6–21, 27; the mean age was 66·4 years. The study populations ranged from 2514 to 30213 participants.

Of 22 studies, two specifically acknowledged financial support from a sponsoring medical company12, 26, whereas nine others stated that no support was received6–8, 10, 16–19, 25. Fourteen studies specified exclusion criteria that included antiplatelet and anticoagulant medications, bleeding disorders and pre-existing venous disease7–14, 17–19, 23, 26, 27; two studies specified no exclusion criteria6, 24 and the rest did not mention exclusion criteria. A study by Ko and colleagues18 was established to investigate low-risk patients; however, the exclusion criteria were similar to those in other studies, and so the trial was not excluded from this systemic review. Four studies permitted the use of elastic stocking for perioperative thromboprophylaxis6, 12, 21, 24, whereas nine provided no method of thromboprophylaxis7, 15–20, 22, 26.

Preoperative deep vein thrombosis

Preoperative screening was reported in four studies7, 14, 21, 24. Three studies (193 patients) used duplex ultrasonography, with a pooled rate of preoperative DVT of 1·5 (95 per cent confidence interval (c.i.) 0·2 to 3·9) per cent (I² = 8·6 per cent, n = 193)7, 14, 24. All were proximal vein thromboses. No preoperative DVT was identified in the single venographic study (58 patients)21.

Postoperative deep vein thrombosis

Of 13 studies that used venography to investigate for postoperative DVT, four used bilateral venography9, 10, 26, 27, four used venograms ipsilateral to the orthopaedic surgery6, 13, 20, 22, and two studies did bilateral venography in patients who had bilateral surgery and ipsilateral venography in those who had unilateral surgery16, 17. The other three studies12, 21, 25 did not specify the extent of the investigation. The results are summarized in Table1.

Table 1. Pooled rates of postoperative deep vein thrombosis from venographic studies

	Pooled rate (%)	Pooled 95% c.i. (%)	I^2 (%)	n	No. of studies	Venograms analysed
Values in parentheses are ranges. c.i., Confidence interval; DVT, deep vein thrombosis. * Two studies with bilateral operation and bilateral venography were excluded. † Only the side of the operated leg was examined.
DVT at all sites	31·7 (9·7–62·5)	22·4, 41·8	94·7	1733	136, 9, 10, 12, 13, 16, 17, 20–22, 25–27	All
	44·0 (10·6–62·5)	27·3, 61·4	90·6	344	49, 10, 26, 27	Bilateral*
	22·1 (4·0–53·1)	11·1, 35·5	91·9	628	66, 13, 16, 17, 20, 22	Ipsilateral†
Proximal DVT	8·9 (0–20·8)	6·4, 11·8	68·9	1524	136, 9, 10, 12, 13, 16, 17, 20–22, 25–27	All
	13·4 (6·0–20·8)	8·4, 19·3	54·9	334	49, 10, 26, 27	Bilateral*
	8·3 (0–11·9)	4·9, 12·4	58·9	628	66, 13, 16, 17, 20, 22	Ipsilateral†
Distal DVT	22·5 (4·0–50·0)	14·5, 31·6	92·4	1259	126, 10, 12, 13, 16, 17, 20–22, 25–27	All
	29·9 (10·0–50·0)	11·3, 53·0	92·4	234	310, 26, 27	Bilateral*
	18·0 (4·0–49·0)	9·7, 28·2	87·6	628	66, 13, 16, 17, 20, 22	Ipsilateral†
Isolated distal DVT	18·8 (2·0–50·0)	10·8, 28·4	93·7	1259	126, 10, 12, 13, 16, 17, 20–22, 25–27	All
	25·9 (10·0–50·0)	7·6, 50·5	93·6	234	310, 26, 27	Bilateral*
	14·3 (2·0–42·9)	6·3, 24·9	90·0	628	66, 13, 16, 17, 20, 22	Ipsilateral†

Of nine studies that used ultrasound imaging to look for postoperative DVT, six examined both proximal and calf veins7, 8, 11, 14, 15, 18, whereas the rest examined only the proximal veins. Seven studies examined both legs7, 11, 14, 15, 18, 19, 24 and the other two did not specify the side. The results are summarized in Table2.

Table 2. Pooled rates of postoperative deep vein thrombosis from ultrasound studies

	Pooled rate (%)	Pooled 95% c.i. (%)	I^2 (%)	n	No. of studies	Ultrasound examinations analysed
Values in parentheses are ranges. c.i., Confidence interval; DVT, deep vein thrombosis.
DVT at all sites	9·4 (0–38·2)	3·7, 17·2	90·1	721	97, 8, 11, 14, 15, 18, 19, 23, 24	All
	10·8 (0–38·2)	2·2, 24·6	93·1	422	67, 8, 11, 14, 15, 18	Proximal and distal vein
	14·6 (0–38·2)	4·4, 29·2	91·3	355	57, 11, 14, 15, 18	Bilateral, proximal and distal vein
Proximal DVT	5·9 (0–15·0)	3·2, 9·2	66·3	721	97, 8, 11, 14, 15, 18, 19, 23, 24	All
	5·3 (0–15·0)	1·7, 10·9	77·5	422	67, 8, 11, 14, 15, 18	Proximal and distal vein
	7·2 (0–15·0)	3·2, 12·6	65·7	355	57, 11, 14, 15, 18	Bilateral, proximal and distal vein
Distal DVT	5·9 (0–29·1)	0·7, 15·8	91·2	422	67, 8, 11, 14, 15, 18	Proximal and distal vein
	7·8 (0–29·1)	1·1, 19·9	91·3	355	57, 11, 14, 15, 18	Bilateral, proximal and distal vein
Isolated distal DVT	5·8 (0–27·3)	0·7, 15·3	90·7	422	67, 8, 11, 14, 15, 18	Proximal and distal vein
	7·6 (0–27·3)	1·1, 19·2	90·8	355	57, 11, 14, 15, 18	Bilateral, proximal and distal vein
Bilateral DVT	2·6 (0–10·9)	0·6, 6·0	65·9	417	67, 14, 15, 18, 19, 24	Bilateral
Isolated contralateral DVT	1·5 (0–9·1)	0·2, 4·0	67·7	484	77, 8, 14, 15, 18, 19, 24	Bilateral

There were 1126 patients from 14 studies who had THA or HFS6, 7, 12–16, 18–20, 22, 24, 26, 27, and 1328 patients from 11 studies who had TKA8–13, 17, 18, 21, 23, 25. The rates of postoperative DVT are summarized in Table3.

Table 3. Postoperative deep vein thrombosis (DVT) rates after total hip arthroplasty (THA), hip fracture surgery (HFS) and total knee arthroplasty (TKA)

	Imaging	Pooled rate (%)	Pooled 95% c.i. (%)	I^2 (%)	n	No. of studies
Values in parentheses are ranges. c.i., Confidence interval.
DVT at all sites, THA/HFS	Venography	25·8 (4·0–53·1)	15·7, 37·4	91·6	767	86, 12, 13, 16, 20, 22, 26, 27
	Ultrasound	10·8 (0–38·2)	3·4, 21·7	87·1	359	67, 14, 15, 18, 19, 23
DVT at all sites, TKA	Venography	42·5 (10·6–62·5)	27·8, 57·8	95·6	966	79, 10, 12, 13, 17, 21, 25
	Ultrasound	9·5 (0–31·0)	1·3, 24·0	93·0	362	48, 11, 18, 23
Proximal DVT, THA/HFS	Venography	9·6 (2·0–20·8)	6·1, 13·7	67·7	767	86, 12, 13, 16, 20, 22, 26, 27
	Ultrasound	7·2 (0–13·6)	4·3, 10·8	30·1	359	67, 14, 15, 18, 19, 24
Proximal DVT, TKA	Venography	8·7 (5·0–14·5)	5·5, 12·7	65·7	757	79, 10, 12, 13, 17, 21, 25
	Ultrasound	5·2 (0–15·5)	1·2, 11·8	80·9	362	48, 11, 18, 23
Isolated distal DVT, THA/HFS	Venography	13·6 (2·0–42·9)	7·4, 21·4	85·5	681	76, 13, 16, 20, 22, 26, 27
	Ultrasound	5·7 (0–27·2)	0, 23·2	92·0	197	47, 14, 15, 18
Isolated distal DVT, TKA	Venography	30·0 (10·3–50·0)	16·6, 45·4	92·8	578	510, 13, 17, 21, 25
	Ultrasound	6·7 (0–15·5)	0·4, 19·4	88·5	225	38, 11, 18

Mortality

No death from postoperative PE was reported in 17 studies (2064 patients)7, 8, 10, 11, 13, 15–25, 27. However, six studies (636 patients) included reporting deaths from other causes in their protocol7, 10, 11, 16, 19, 22. Three studies specifically reported no mortality from other causes11, 16, 22. The other three reported seven deaths: two from myocardial infarction19, and five unspecified but not related to PE or DVT7, 10. Seven late deaths were also reported in two studies (295 patients)7, 16. Of these, two patients who died at 7 and 11 months after surgery had had a negative lung scan and negative venogram respectively16, whereas the cause of the other five deaths recorded at 3-month follow-up was unknown7. None had an autopsy.

Symptomatic postoperative venous thromboembolism

The rates of symptomatic postoperative DVT and PE are summarized in Table4. Three studies included routine postoperative PE screening. Pookarnjanamorakot and colleagues8 arranged a routine lung scan in 67 patients, of whom five (7 per cent) had a high probability of PE. However, all of these had negative ultrasound vein scans and none was symptomatic. Kim and co-workers performed a lung perfusion scan in every patient before and after surgery in two studies (200 and 264 patients)16, 17; no patient had the combination of a positive lung scan and symptoms, but the number of positive lung scans was not recorded.

Table 4. Pooled rates of all symptomatic venous thromboembolism

	Pooled rate (%)	Pooled 95% c.i. (%)	I^2 (%)	n	No. of studies
Values in parentheses are ranges. c.i., Confidence interval; DVT, deep vein thrombosis; THA, total hip arthroplasty; HFS, hip fracture surgery; TKA, total knee arthroplasty; PE, pulmonary embolism.
Symptomatic DVT, all studies	4·5 (0–30·6)	1·9, 8·1	88·0	1506	156–8, 10, 11, 14, 17–23, 26, 27
Symptomatic DVT, DVT screening protocol studies	4·3 (0–30·6)	0·9, 10·2	89·8	713	106, 7, 10, 11, 14, 18–20, 22, 26
Symptomatic DVT, THA/HFS	3·9 (0–30·6)	0·7, 9·7	86·8	541	86, 7, 14, 19, 20, 22, 26, 27
Symptomatic DVT, TKA	2·7 (0–21·6)	0·1, 8·1	90·7	714	68, 10, 11, 17, 21, 23
Symptomatic PE, all studies	0·6 (0–2·5)	0·3, 1·0	7·4	2233	187–13, 15–25
Symptomatic PE, PE screening protocol studies	0·6 (0–2·5)	0·2, 1·2	36·1	1456	1010, 13, 16–20, 22, 24, 25
Symptomatic PE, THA/HFS	0·3 (0–1·7)	0, 1·0	0	633	77, 15, 16, 19, 20, 22, 24
Symptomatic PE, TKA	0·5 (0–1·7)	0·1, 1·1	17·5	1053	88–11, 17, 21, 23, 25
Death from PE	0·2	0, 0·4	0	2064	177, 8, 10, 11, 13, 15–25, 27
Death from PE, PE screening protocol studies	0·2	0, 0·4	0	1506	1110, 13, 16–20, 22, 24–26

Medium-term follow-up was reported in three studies7, 16, 18. Chan et al.7 reported 90 patients without DVT during the hospital stay. At 3 months, 73 patients were interviewed and none reported symptoms of VTE; 51 patients had follow-up ultrasonography, where one asymptomatic DVT was found. Ko and colleagues18 followed up all 80 patients for at least 6 months; no patient who had a negative venous duplex ultrasound examination developed symptomatic VTE.

Rate of deep vein thrombosis over time

Reported DVT rates over time were illustrated by a forest plot of proximal DVT rates on routine venography (Fig.2). The proximal DVT rate in the studies between 1979 and 1987 (earliest third) were compared with those from 1995 to 2003 (last third). In the first group, the direct calculated rate of postoperative DVT was 4·7 per cent (3 studies6, 20, 22; pooled rate 5·1 per cent; I² = 44·8 per cent) compared with 10·3 per cent in the second group (8 studies9, 12, 13, 16, 17, 21, 25, 27; pooled rate 10·0 per cent; I² = 74·5 per cent). There was a significant increase between the two intervals (P = 0·020).

Figure 2. Forest plot of rate of proximal deep vein thrombosis after orthopaedic surgery from the venographic studies, sorted by time. *Mid-year of patient recruitment. Values in parentheses are confidence intervals

Rate of deep vein thrombosis among ethnic groups

Proximal DVT rates were compared among the various ethnic groups: South-East Asian (pooled rate 11·8 (95 per cent c.i. 5·6 to 19·9) per cent; I² = 76·4 per cent), Japanese (pooled rate 11·0 (8·4 to 14·0) per cent; I² not computable as only 2 studies), Korean (pooled rate 7·5 (3·4 to 13·2) per cent; I² = 64·0 per cent) and Chinese (pooled rate 5·6 (2·9 to 9·1) per cent; I² = 30·0 per cent). The pooled rate was highest in South-East Asia and lowest in China. There were significant differences between Chinese and Korean patients (P = 0·032), between Chinese and Japanese patients (P = 0·002), and between Chinese and South-East Asian patients (P = 0·007). The other comparisons were not significantly different (Korean versus Japanese, P = 0·332; Korean versus South-East Asian, P = 0·202; Japanese versus South-East Asian, P = 0·708).

Discussion

This systematic review and meta-analysis of over 2000 orthopaedic procedures in Asia found no postoperative deaths, despite omission of thromboprophylaxis. The only other similar review by Liew and colleagues40 from Malaysia comprised 12 studies with a total population of 1495. The lack of deaths from PE was also reported in an Asian multinational study of 407 patients by Piovella and co-workers34, where the single death (0·2 per cent) was from cerebral embolism. Another Asian multinational study by Leizorovicz et al.5 reported a 0·2 per cent rate of sudden death (4 of 2402); however, no specific rates of VTE were described. The rates of symptomatic PE were low (0·6 per cent; I² = 7·4 per cent) in the present meta-analysis, and comparable between the venographic and ultrasonographic studies, including those with a postoperative PE screening protocol10, 13, 16–20, 22, 24, 25.

The overall rate of symptomatic DVT was 4·5 per cent. However, heterogeneity among the studies of symptomatic DVT was high. Symptoms of DVT can be subtle and non-specific. Ko and colleagues18, who used duplex Doppler ultrasound imaging for screening, reported a 27·2 per cent sensitivity and 31·6 per cent positive predictive value for clinical symptoms of DVT. In a study by Jain et al.14, none of ten with patients with DVT symptoms had this confirmed on duplex imaging. The rate of symptomatic DVT was higher in venographic than in ultrasonographic studies, presumably because the sensitivity of venography is higher than that of ultrasound imaging.

The rate of DVT varied after the different orthopaedic procedures. The DVT rate is probably best approximated from pooling of bilateral venography studies, although these were relatively few (2 each for the THA and HFS26, 27, and TKA9, 10 groups). Pooling all available venographic studies, the postoperative DVT rates after THA and HFS, and TKA were 25·8 and 42·5 per cent respectively. However, proximal DVT rates may have more clinical relevance as isolated calf DVT has a very low risk of PE (0–13 per cent)41–44 and no report of fatal PE43. Complications occasionally materialize if the DVT extends proximally45, 46. The venographic proximal DVT rates in the THA and HFS, and TKA groups were 9·6 and 8·7 per cent respectively, and showed acceptable heterogeneity.

When comparing rates detected by ultrasound imaging and venography, the difference in the rate of distal DVT was more pronounced than in the rate of proximal DVT. Duplex imaging is less accurate in the diagnosis of calf vein thrombosis47. The similarity between rates of proximal DVT diagnosed by venography and ultrasonography allowed their combination in the comparison of ethnic groups and changes in DVT rates over time. There appeared to be variation among ethnic groups and geographical locations in Asia, perhaps partly explaining the heterogeneity among studies. Some authors believe that the prevalence of VTE in Asia is increasing as a result of environmental change and Westernization of food8, 10, 13, 24, 29. The present systematic review supports this view. However, this analysis may be limited by the small number of the studies in the early group, and the possibility of improved diagnosis over time. These findings also aroused comment from the Asia–Pacific Thrombosis Advisory Board48. Although it still recommends routine prophylaxis, it admits flexibility, and local guidelines based on prevalence may be appropriate.

The limitations of this systematic review included the small population in many studies. Heterogeneity between studies was high in some of the analyses, including the rate of symptomatic DVT and variations in the rates according to the method of diagnosis. Currently the standard method for diagnosis of PE is computed tomography pulmonary angiography, which is more accurate than lung perfusion scan. The number of symptomatic PEs should be viewed in the light of the relatively low specificity of lung scans49 which were used in the majority of patients in this meta-analysis. As the available papers came mainly from East Asian and South-East Asian countries, this review may not represent South Asian countries, which include a substantial proportion of Asia. Data on Chinese patients were taken from Hong Kong and Taiwanese papers; reports directly from the People's Republic of China, which includes the largest population in Asia, were not available.

Summaries of similar patients in Western studies suggest proximal DVT rates of over 20 per cent after orthopaedic procedures50, 51, compared with 8·7 per cent (TKA) and 9·6 per cent (THA, HFS) in the present review. The rate of all-site DVT on venography (31·7 per cent) may approach Western values, but most thromboses were distal and therefore of less significance. Although the possible trend towards increasing incidence, and the ethnic variation, require further consideration48, the lack of any reported death from VTE questions the potential benefit of routine thromboprophylaxis in these orthopaedic patients.

Acknowledgements

The Bangkok Metropolitan Administration Research Committee funded this project. The authors declare no conflict of interest.