Participants

Trials were conducted in Belgium, Canada, Denmark, Egypt, France, Germany, the Netherlands, Norway, South Africa, Sweden, Switzerland, Turkey, the UK and the USA. All trials included men and women, with an age range from 18 to 75 years and with a preponderance of older adults. The participants had diverse underlying causes for developing DVT, and varying degrees of level and extent of occlusion. Both Elsharawy 2002 and CAVA 2020 included DVTs affecting femoral and iliofemoral veins; CAVENT included pelvic, femoral and iliofemoral veins; ATTRACT included proximal DVT; other trials included thrombosis affecting different combinations of levels, including popliteal. The only study to include calf vein thrombosis only was Schulman 1986. See Table 1,

Participant inclusion criteria

Participant inclusion criteria have become more restrictive over time. In the earliest study (Kakkar 1969), there were only four contra‐indications: surgery within three days, an unhealed wound, peptic ulcer and hypertension. By the time of Schweizer 2000, a more comprehensive list of contra‐indications had been developed including: surgery or head trauma within the previous three months, malignancy, renal and hepatic disorders, and bleeding disorders. In these later studies, the more comprehensive list reduces the proportion of eligible participants.

Interventions

Interventions included systemic, loco‐regional, catheter‐directed and pharmacomechanical thrombolysis. Systemic and loco‐regional techniques differ only in the veins used to deliver an infusion: the arm or foot respectively. Catheter‐directed thrombolysis (CDT) is a more invasive procedure in which a catheter is inserted percutaneously into a vein using imaging to guide to the clot location. The thrombolytic agent is infused through the catheter into the blood clot itself and the position of the catheter is altered according to the progress made in lysing the blood clot. The majority of trials assessed systemic thrombolysis, with streptokinase the most common agent used. The dose used varied: Schulman 1986 used a low‐dose regime of systemic streptokinase, Tsapogas 1973 used loco‐regional streptokinase and Elsharawy 2002 used catheter‐directed streptokinase with frequent radiological assessment, a technique used again in CAVENT. ATTRACT used mechanical therapy in addition to CDT (pharmacomechanical thrombolysis). These adjunctive therapies could include balloon maceration, catheter aspiration, thrombectomy, percutaneous transluminal balloon venoplasty, stent placement, or combination of the above (ATTRACT; Vedantham 2017). CAVA 2020 used ultrasound accelerated CDT to deliver urokinase.

Goldhaber 1990, Turpie 1990 and Verhaeghe 1989 used systemic tPA. While doses of tPA varied, there was no obvious cut‐off for high or low doses. Goldhaber 1996 randomised two regimes of tPA, with and without heparin, compared to heparin alone. The two treatment arms were combined for the purposes of this review. Schweizer 1998 had two treatment arms, loco‐regional tPA and urokinase; and Schweizer 2000 had four treatment arms: systemic streptokinase, systemic urokinase, loco‐regional urokinase and loco‐regional tPA. Kiil 1981 used low‐dose systemic urokinase.

Co‐treatments

Monitoring regimes for heparinisation varied, and length of anticoagulation after the initial phase may be limited to a few months or continued for over a year. In some trials, especially the more recent ones, the use of compression bandages and elevation were reported; and for longer follow‐up, some participants were required to use compression stockings with rigorous ascertainment of compliance with the continued treatment.

Size

Nine studies had fewer than 50 participants (Arnesen 1978; Elsharawy 2002; Goldhaber 1996; Kakkar 1969; Kiil 1981; Marder 1977; Schulman 1986; Tsapogas 1973; Verhaeghe 1989), and four studies had more than 100 participants (ATTRACT; CAVA 2020; CAVENT; Schweizer 2000). Most studies therefore lacked power to detect statistically significant effects. A power calculation was described in four studies (CAVA 2020; CAVENT; Elsharawy 2002; Schweizer 2000). ATTRACT was the largest trial, with 692 participants.

Outcomes

One trial (Verhaeghe 1989) reported results for randomised participants together with non‐randomised participants, but we have used only the data from randomised participants. Some studies reported outcomes using scales that could not be combined (Marder 1977). Removal of the clot was reported using various categorisations. In order to capture as much information as possible, we report in this review both complete clot dissolution or lysis, indicating that the venous patency was 100% restored, and any degree of venographic improvement in patency. Tsapogas 1973 reported partial or complete clearance (75% to 100%), a measure not used in any other study, and others reported partial clearance (50% to 100%). The more recent trials report primarily on PTS and QoL (ATTRACT; CAVA 2020; CAVENT), with CAVENT also reporting cost comparisons.

Length of follow‐up

All trials assessed outcomes in the period immediately after treatment. This was usually at one week, although the range was 36 hours to one month. We collectively grouped these as early outcomes. Intermediate outcomes have been classified as those determined after six months and under five years. No data were reported between this early and intermediate phase (i.e. after one month and before six months). Late outcomes were those reported five years or more after the intervention. PTS was assessed between one and six years. The longest follow‐up (six years) was in the Arnesen 1978 study.

Ongoing studies

For this update, we identified two new ongoing studies (ChiCTR‐INR‐16009090; NCT02959801). Two previously ongoing studies are now included (ATTRACT; CAVA 2020). We contacted the study investigators of the three ongoing studies, but no data were available (ChiCTR‐INR‐16009090; IRCT201108035625N3; NCT02959801). See Characteristics of ongoing studies for further details.

Studies awaiting classification

For this update, two potential new studies were identified and placed into Studies awaiting classification until a thorough assessment of the full text can be made (Gong 2018; Su 2017).

Complete clot lysis

Eight trials with 592 participants reported on the occurrence of early complete clot lysis (up to one month follow‐up) (Common 1976; Elliot 1979; Elsharawy 2002; Goldhaber 1990; Kakkar 1969; Schulman 1986; Schweizer 2000; Ugurlu 2002). In all trials this was more likely in the treatment group, although the extent of the effect varied and the results were statistically heterogeneous. A random‐effects model demonstrated a benefit from thrombolysis treatment (RR 4.75, 95% CI 1.83 to 12.33; P = 0.001; Analysis 1.1). No differences between thrombolysis strategy used were seen with subgroup analysis (test for subgroup differences: P = 0.41).

Seven trials with a total of 654 participants reported intermediate clot lysis (after six months) and in all cases this was more likely in the groups treated with thrombolysis (Common 1976; CAVENT; Elliot 1979; Elsharawy 2002; Schulman 1986; Schweizer 1998; Schweizer 2000). There was a benefit seen with thrombolysis treatment, (RR 2.42, 95% CI 1.42 to 4.12; P = 0.001 using a random‐effects model (moderate‐certainty evidence; Analysis 1.2). No differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.37). We downgraded the certainty of the evidence from high to moderate due to imprecision.

Two trials with a total of 206 participants reported clot lysis at five years and over (Arnesen 1978; CAVENT). There was no clear benefit of clot lysis with thrombolysis seen at this time point (RR 3.25, 95% CI 0.17 to 62.63; P = 0.44; Analysis 1.3). No differences between thrombolysis strategies used were seen with subgroup analysis (test for subgroup differences: P = 0.06).

CAVA 2020 data on clot lysis are to be published in subsequent papers.

Bleeding

This category excluded cerebral bleeding and minor bleeds, for example, oozing from venepuncture sites and superficial haematomas. All 19 trials reported on the occurrence of bleeding episodes (Arnesen 1978; ATTRACT; CAVA 2020; CAVENT; Common 1976; Elliot 1979; Elsharawy 2002; Goldhaber 1990; Goldhaber 1996; Kakkar 1969; Kiil 1981; Marder 1977; Schulman 1986; Schweizer 1998; Schweizer 2000; Tsapogas 1973; Turpie 1990; Ugurlu 2002; Verhaeghe 1989). Overall, 6.7% (72/1073) of participants in the thrombolysis group experienced a bleeding complication compared to 2.2% (20/870) of participants in the standard anticoagulation group (RR 2.45, 95% CI 1.58 to 3.78; 1943 participants; 19 studies; P < 0.0001; moderate‐certainty evidence; Analysis 1.4). No differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.25). We downgraded the certainty of the evidence from high to moderate due to imprecision. We detected no indication of reporting bias (Figure 4).

Figure 4

---

---

Funnel plot of comparison: 1 Thrombolysis versus standard anticoagulation, outcome: 1.4 Bleeding (early, subgrouped by thrombolysis strategy).

PTS

Six studies reported clinically assessed PTS at six months to 5 years (intermediate) (ATTRACT; CAVA 2020; CAVENT; Elliot 1979; Schweizer 1998; Schweizer 2000), in a format that could be combined, with a total of 1393 participants. Fewer cases of PTS were reported in those participants receiving thrombolysis compared to those in the control group (RR 0.78, 95% CI 0.66 to 0.93; 1393 participants; six studies; P = 0.006; moderate‐certainty evidence; Analysis 1.5). We downgraded the certainty of the evidence from high to moderate due to imprecision and small number of participants in the majority of the included studies. The incidence in the thrombolysis group compared with the control group was 383/771 (49.6%) versus 329/622 (52.8%, ranging from 35% to 96% in different trials, which may reflect definitions, treatment doses and adjunctive treatments). We detected heterogeneity (P = 0.01), so we used a random‐effects model. No differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.22; Analysis 1.5). Two studies provided data for systemic thrombolysis and these showed a reduction in the incidence of PTS (RR 0.54, 95% CI 0.31 to 0.92; 170 participants; P = 0.02). It is of note that both studies used a much higher dose of thrombolytic than those used in other studies (Elliot 1979; Schweizer 2000; see Characteristics of included studies). No clear benefit for CDT was shown by pooling results from the ATTRACT, CAVA 2020 and CAVENT trials (RR 0.89, 95% CI 0.74 to 1.05; 1032 participants; P = 0.17).

We also carried out subgroup analysis by DVT level (iliofemoral, femoropopliteal or non‐specified). We have included all participants of CAVENT in the iliofemoral group as this population was similar to the ATTRACT iliofemoral group (personal communication with the study authors). In keeping with reports above, overall the thrombolysis group experienced less PTS (RR 0.82, 95% CI 0.71 to 0.94; 1393 participants; six studies, P = 0.006). No differences in PTS incidence between level of DVT were detected between subgroups (test for subgroup differences: P = 0.29; Analysis 1.6).

Two studies with 211 participants (Arnesen 1978; CAVENT), reported reduced incidence of clinically assessed PTS at over five years (late follow‐up) following thrombolysis (RR 0.56, 95% CI 0.43 to 0.73; 211 participants; two studies; P < 0.0001; moderate‐certainty evidence; Analysis 1.7). We downgraded the certainty of the evidence from high to moderate due to the small number of studies and participants. In the control group, the incidence of PTS was 75/107 (70%) and in the thrombolysis group 41/104 (39%). No differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.28).

Any improvement in venous patency

Nine trials reported on improvements in venous patency defined by a change in occlusion of the affected segment after treatment (Arnesen 1978; Common 1976; Elsharawy 2002; Goldhaber 1990; Goldhaber 1996; Kakkar 1969; Kiil 1981; Turpie 1990; Ugurlu 2002). All of these studies used systemic thrombolysis strategies except for Elsharawy 2002. Out of a total of 421 participants, improvement was more likely in those receiving thrombolysis (RR 2.48; 95% CI 1.35 to 4.57, P = 0.004; Analysis 1.8). Statistical heterogeneity was noted (P < 0.0001), and a random‐effects model was used. The study by Marder 1977, which showed a difference in mean change from venograms, could not be included due to the reporting format used. A greater improvement was noted but for randomised participants this was not reported to be significantly different. Similarly the Verhaeghe 1989 data could not be included in the meta‐analysis. No clear differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.05).

Stroke or intracerebral haemorrhage

Three trials reported the occurrence of stroke or intracerebral haemorrhage (Common 1976; Goldhaber 1990; Marder 1977). All trials described bleeding complications, therefore the absence of mention of any serious neurological complications or cerebral bleeds was taken to indicate that none were detected. Out of a total of 1943 participants three events occurred in the treatment group (3/1943) and none in the control group. The pooled RR was 1.92 (95% CI 0.34 to 10.86; P = 0.46; 19 studies; Analysis 1.9). All studies where stroke occurred were from before 1990.

Leg ulceration

Five studies described ulceration of the leg occurring more than six months from trial entry (ATTRACT; CAVENT; Elliot 1979; Schulman 1986; Schweizer 1998). Fifteen events occurred in the treatment group (15/513) and 19/520 in the control group (RR 0.76, 95% CI 0.39 to 1.49; 1033 participants; five studies; P = 0.43; Analysis 1.10). No differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.72).

Arnesen 1978 reported at a mean of 6.5 years and so fell within the definition of late ulceration. This study involved a small number of participants, with 3/18 control participants experiencing ulceration after six years compared to 0/17 in the thrombolysis participants (RR 0.15, 95% CI 0.01 to 2.72; P = 0.20; Analysis 1.11).

Mortality

Ten trials reported on deaths occurring up to one month after treatment (ATTRACT; Arnesen 1978; Common 1976; Elliot 1979; Elsharawy 2002; Kakkar 1969; Kiil 1981; Marder 1977; Schulman 1986; Schweizer 2000); three trials reported that no deaths occurred in this period (ATTRACT; Elsharawy 2002; Schweizer 2000). A total of five events occurred in the treatment group (5/677) and seven in the control group (7/543), out of a total of 1220 participants. There were relatively few events and no clear difference between the groups (RR 0.76; 95% CI 0.31 to 1.89; P = 0.56; Analysis 1.12). All the recorded events occurred within the systemic thrombolysis subgroup, a test for subgroup differences was not estimable. All trials where deaths occurred at this time point were from before 1987.

Four trials with a total of 1144 participants reported mortality occurring up to five years after treatment (ATTRACT; CAVA 2020; Elliot 1979; Schweizer 2000). No deaths were reported in either group in Schweizer 2000; CAVA 2020, ATTRACT and Elliot 1979 reported similar numbers of deaths in each group (RR 0.81, 95% CI 0.39 to 1.69; 1144 participants; four studies; P = 0.58; Analysis 1.13). Most deaths were unrelated to the clot or treatment but rather to underlying conditions. ATTRACT reported one death due to PE in the thrombolysis group. No differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.76).

Two trials with a total of 230 participants reported mortality after five years follow‐up (Arnesen 1978; CAVENT). Seven deaths occurred in the thrombolysis group (7/111) and 12/119 in the control group with a RR of 0.61 (95% CI 0.25 to 1.50; P = 0.28; Analysis 1.14). No differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.17).

Recurrent venous thromboembolism (DVT/VTE)

At intermediate time points, one trial reported on recurrent DVT (Arnesen 1978), while ATTRACT, CAVA 2020 and CAVENT reported recurrent VTE. Seventy‐three events (including one fatal PE event in the ATTRACT study) occurred in the treatment group (73/520), compared to 58/547 in the control group. The RR was 1.32, 95% CI 0.96 to 1.83; 1067 participants; four studies; P = 0.09); Analysis 1.15. No differences were seen with subgroup analysis by thrombolysis strategy (test for subgroup differences: P = 0.92).

At five year follow‐up, CAVENT reported 13/87 and 21/89 VTE events in the thrombolysis and anticoagulation groups respectively (RR 0.63, 95% CI 0.34 to 1.18; Analysis 1.16). Thirteen events were in the ipsilateral leg, 10 in the contralateral leg, nine were PE and two were unknown. Six patients with chronic iliac vein occlusions (one in the CDT group and five in the control group), were referred and had endovascular recanalisation with stenting. Although randomised to the treatment group, the CDT patient had not received CDT as planned, due to technical failure (Haig 2016).

PE

Six trials reported the occurrence of a PE in the early phase (Arnesen 1978; Elliot 1979; Elsharawy 2002; Kakkar 1969; Schulman 1986; Schweizer 2000). One study noted the absence of any PE (Schulman 1986). The diagnostic criteria used were variable. With the exception of participants who died from PE (one in the treatment group, two in the control group), transient clinical symptoms often occurred but with no objective diagnostic confirmation described. Where deaths were attributed to PE, postmortem examinations were not mentioned. For this reason, the results should be interpreted with caution. The RR was 1.01 (95% CI 0.33 to 3.05; 443 participants; six studies; P = 0.98; Analysis 1.17). No differences were detected between subgroups (P = 0.43). CAVENT did not measure this outcome at this time point. ATTRACT reported that there were 6/336 recurrent VTE events in the thrombolysis group compared to 4/355 (P = 0.5) within the first 10 days; it is not specified if these were PE or DVT.

Venous function (intermediate)

Three trials reported on presence of normal venous function (CAVENT; Elsharawy 2002; Schulman 1986). Overall, no clear benefit to venous function with thrombolysis was shown, (RR 2.18; 95% CI 0.86 to 5.54; 255 participants, 3 studies, P = 0.10, Analysis 1.18). Heterogeneity was detected (P = 0.009) so a random‐effects model was used. Subgroup analysis suggests a difference between use of systemic and CDT strategies (P = 0.03) with increased normal venous function being seen in the CDT group (RR 3.18, 95% CI 1.41 to 7.19; 224 participants; two studies; P = 0.005).

QoL

Only ATTRACT, CAVA 2020 and CAVENT measured QoL. All three studies used the Venous Insufficiency Epidemological and Economic Study Quality of Life (VEINES‐QOL) measure, which includes both an overall score and a symptom score. In addition, CAVENT and CAVA 2020 used the generic instrument EQ‐5D; and ATTRACT and CAVA 2020 used the Medical Outcomes Study 36‐Item Short Form Health Survey (SF‐36). This includes both a physical component score (PCS) and mental component score (MCS). CAVA 2020 also used the Pain Disability Index (PDI), which reports limitations in daily activities due to pain (scored from 0 to 10, 0 no limitations, 10 fully disabled). VEINES‐QoL, EQ‐5D and SF‐36 scores range from 0 to 100, higher scores indicating better QoL or health perception. A difference of 3 to 4 points is considered clinically meaningful.

After 12 months, CAVA 2020 reported that there were no differences between the CDT and standard treatment group for any health‐related and disease‐specific QoL assessment (mean (SD) VEINES‐Sym score for CDT was 50.1 (11.1) compared to 49.7 (9) in the standard treatment group; mean (SD) SF36 general scores were 65.6 (17.8) in the CDT group compared to 64.9 (22.8) in the standard treatment group; mean (SD) EQ5D score for the CDT group was 85.7 (15.0) compared to 82.3 (21.0) in the standard treatment group; and using the PDI, the mean (SD) CDT group score was 8.7 (12.4) compared to 13.1 (16.3) in the standard treatment group.

After 24 months CAVENT reported there were no differences in QoL between the additional CDT and standard treatment arms; mean difference for the EQ‐5D index was 0.04 (95% CI ‐0.10 to 0.17), for the VEINES‐QOL score 0.2 (95% CI ‐2.8 to 3.0) and for the VEINES‐Sym score 0.5 (95% CI ‐2.4 to 3.4; P > 0.37). After five years, CAVENT reported no difference in mean generic QoL scores, disease specific QoL scores, or symptom severity score between the groups (Enden 2012; Enden 2013a).

Independent of treatment arms, patients with PTS had poorer outcomes than patients without PTS after 24 months; mean difference for EQ‐5D was 0.09 (95% CI 0.03 to 0.15), for VEINES‐QOL score 8.6 (95% CI 5.9 to 11.2) and for VEINES‐Sym score 9.8 (95% CI 7.3 to 12.3; P < 0.001). After five years the EQ‐5D, VEINES‐QOL and VEINES‐Sym scores for patients with PTS were lower than for those without PTS (Enden 2012; Enden 2013a).

After 24 months, ATTRACT reported no difference between the groups with the SF‐36 form (mean (SE) 11.18 (0.91) in the thrombolysis group compared to 10.06 (0.97) in the control group (P = 0.37)); with the VEINES‐QOL score (mean (SE) 27.67 (1.71) in the thrombolysis group compared to 23.47 (1.83) in the control group (P = 0.37)); or the VEINES‐Sym score (mean (SE) 20.58 (1.70) in the thrombolysis group compared to 17.31 (1.81) in the control group (P = 0.17)). A report of a secondary analysis of the ATTRACT study by Kahn 2018 reported VEINES‐QOL scores were better in the thrombolysis group compared to the control group at 30 days (mean (SE) 64.9 (1.4) versus 60.3 (1.4); P = 0.018) and six months (77.0 (1.4) versus 73.1 (1.4); P = 0.044), respectively. This improvement was also detected in the iliofemoral subgroup but not in the femoropopliteal subgroup (Kahn 2018).

Cost comparisons

Only CAVENT has reported on this outcome (Enden 2013b). Additional CDT accumulated 32.31 quality‐adjusted life years (QALYs) compared with 31.68 QALYs after standard treatment. The lifetime cost of CDT was USD 64,709 compared to USD 51,866 with standard treatment. The incremental cost‐effectiveness ratio was USD 20,429 per QALY gained, and the study authors concluded that the probability that CDT was cost‐effective was 82% at a willingness to pay threshold of USD 50,000 per QALY gained (Enden 2013b). CDT may have additional costs compared to systemic administration.

Sensitivity analyses

We carried out sensitivity analyses for all outcomes where the meta‐analysis included trials judged to have any domain at high risk of bias. To determine if results were robust, meta‐analyses were repeated excluding the following studies: Kakkar 1969; Marder 1977; Tsapogas 1973. Forest plots and summary figures were visually assessed and for all outcomes the results remained consistent.