Clinical effectiveness and cost-effectiveness of minimally invasive techniques to manage varicose veins: a systematic review and economic evaluation

Endovenous laser ablation

Endovenous laser ablation involves insertion and activation of a laser fibre into the refluxing vein. Wavelengths are used to target deoxygenated haemoglobin and/or water, which result in heating and thrombosis or occlusion of the vein. 53 Patients with either GSV or SSV incompetence might receive this intervention.

Radiofrequency ablation

Radiofrequency ablation involves insertion of a catheter into the varicose vein. Electrodes at the end of the catheter omit high radiofrequency energy, which heats tissue at the site, causing collagen shrinkage, denudation of endothelium (the cells that line the blood vessels) and obliteration of the venous lumen (space inside the vein). 38 This includes techniques such as VNUS Closure,^® VNUS ClosureFast^® (VNUS Medical Technologies, Inc., San Jose, CA)54 and Olympus RFiTT^® (Olympus Surgical Technologies Europe, Hamburg, Germany). 55 Patients with either GSV or SSV incompetence might receive this intervention.

Foam sclerotherapy

Foam sclerotherapy involves the mixing of air with liquid sclerosing solution to create foam. The foam is injected into the affected vein guided by ultrasound. 12 Patients with either GSV or SSV incompetence might receive this intervention.

Transilluminated phlebectomy

Transilluminated phlebectomy offers an alternative to multiple phlebectomies. It involves hydrodissection of the varicosities, transillumination facilitating direct visualisation of the varicosities, and varicosity removal using a powered endoscopic tissue dissector. 39 This includes techniques such as powered phlebectomy (TriVex™; InaVein, Lexington, MA). 56 Patients would only receive this intervention if there was no GSV incompetence.

Intervention

New minimally invasive methods of managing varicose veins: EVLA, ultrasound-guided foam sclerotherapy (just FS), RFA and TIPP.

Population and relevant subgroups

Adults aged ≥ 16 years who are being treated specifically for varicose veins.

Relevant comparators

Although any comparator was considered, the reviews focused principally on surgical treatment. Other comparators included LS, not non-FS, etc., and conservative management. Head-to-head trials comparing the minimally invasive techniques were also included.

Identification of studies

A comprehensive search was undertaken to identify systematically clinical effectiveness literature comparing different methods for the management of varicose veins. The search involved combining terms for the population (varicose veins) with terms for the interventions of interest (i.e. the minimally invasive techniques). This highly sensitive search strategy (i.e. not using terms for comparators, outcomes or study design) was possible because scoping searches retrieved relatively small and manageable numbers of citations. An example MEDLINE search strategy is reported in Appendix 1. The aim of the strategy was to identify all studies comparing the techniques of interest with each other, conventional surgery, LS or conservative management (no RCT filter was used). All searches were performed by an information specialist (AC) in July 2011.

The following electronic databases were searched from inception for published and unpublished research evidence:

• MEDLINE (Ovid) 1946–

• EMBASE (Ovid) 1980–

• Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCO) 1982–

• the Cochrane Library including the Cochrane Database of Systematic Reviews (CDSR), Cochrane Central Register of Controlled Trials (CENTRAL), Database of Abstracts of Reviews of Effects (DARE), Health Technology Assessment (HTA) database and NHS Economic Evaluation Database (NHS EED) 1991–

• Biological Abstracts (BIOSIS Previews) (via ISI Web of Science) 1969–

• Science Citation Index (SCI) (via ISI Web of Science) 1900–

• Social Science Citation Index (SSCI) (via ISI Web of Science) 1956–

• Conference Proceedings Citation Index-Science – (via ISI Web of Science) 1990–

• UK Clinical Trials Research Network

• Current Controlled Trials

• ClinicalTrials.gov.

All citations were imported into Reference Manager version 12 (Thomson ResearchSoft, San Francisco, CA, USA) and duplicates deleted. Titles and abstracts of all unique citations were then screened independently by two reviewers (JL, EEH) using the inclusion criteria outlined below after a test screen on a sample of citations. Disagreements or queries were resolved by consensus or with reference to a third team member (CC or JM) where necessary. The full papers of all potentially relevant citations were then retrieved so that an in-depth assessment concerning inclusion could be made. Reference tracking of all included studies and relevant reviews was also performed to identify additional, relevant studies not retrieved by the search of electronic databases. Clinical advisors were also contacted about relevant RCTs that might have been missed.

Inclusion and exclusion criteria

Data abstraction strategy

Data abstraction was performed by one reviewer into a standardised data extraction form (see Appendix 2) and independently checked for accuracy by a second. Discrepancies were resolved by discussion between the two reviewers and, if agreement could not be reached, a third reviewer was consulted.

Critical appraisal strategy

The quality assessment of included RCTs was performed by one reviewer, using appropriate quality assessment criteria adapted from a published checklist for surgical interventions (see Appendix 3), and independently checked for accuracy by a second. Discrepancies were resolved by discussion between the two reviewers and, if agreement could not be reached, a third reviewer was consulted. Blinding of patients and outcome assessors were not retained as criteria because the techniques generally did not permit such blinding, so the risk of detection bias was often inherently high. Other amendments to the tool are described in Appendix 3. The 5% level of attrition specified in the original tool was retained, as this proportion has been reported to be the level least likely to affect outcomes adversely. 61

Methods of data synthesis

Technical recurrence, VCSS and pain score data were tabulated, and included studies were combined in a formal network meta-analysis. A network meta-analysis allows a comprehensive comparison of all interventions that are linked with respect to at least one common intervention without breaking the randomisation within studies. A network meta-analysis makes the same assumptions as standard pairwise meta-analyses. In particular, that there is consistency of direct and indirect evidence about treatment effects across the network.

The summary statistics that were analysed were the number of patients who had an event for technical recurrence, and the mean VCSS and mean pain score. In each case, the data were analysed using a random effects model (to allow for heterogeneity in treatment effects across studies) using Markov chain Monte Carlo simulation implemented in the WinBUGS (MRC Biostatistics Unit, Cambridge, UK) and OpenBUGS (Members of the OpenBUGS Project Management Group) software packages. The analysis was conducted using a Bayesian framework in order to quantify the joint distribution about uncertain parameters as required for the economic model.

For technical recurrence, the statistical model accounted for the variation in the duration of follow-up between studies using a complimentary log–log link function assuming that the underlying survivor functions follow Weibull distributions with separate shape and scale parameters to allow for the possibility of non-proportional hazards (see Appendix 4). Results of the network meta-analyses are reported in terms of the hazard ratios and 95% credible intervals (CrIs) relative to the baseline intervention (i.e. stripping) at 6 months, 1 year and 2 years. The posterior medians of the between-study standard deviations (SDs) for the shape and scale parameters together with their 95% CrIs are also presented.

For VCSS and pain scores, the statistical model used an identity link by assuming a normal distribution for the observed sample means (see Appendix 5). Results of the network meta-analyses are reported in terms of the mean difference (MD) and 95% CrIs relative to the baseline intervention (i.e. stripping). The posterior median of the between-study SD together with the 95% CrI was also presented.

Convergence of the models to their posterior distributions was assessed using the Gelman–Rubin convergence statistic. 62 Convergence occurred after 200,000 iterations for technical recurrence, after 10,000 iterations for VCSS and after 30,000 iterations for pain. There was some suggestion of high autocorrelation between successive iterations of the Markov chains; to compensate for this the Markov chains were thinned every 25 iterations for technical recurrence, every 10 iterations for VCSS and every 20 iterations for pain. Parameter estimates were estimated based on 20,000 iterations of the Markov chains for technical recurrence, 20,000 iterations for VCSS and 30,000 for pain.

The total residual deviance was used to assess formally whether or not the statistical model provided a reasonable representation of the sample data. The total residual deviance is the mean of the deviance under the current model minus the deviance for the saturated model, so that each data point should contribute about to the deviance. 63

To enable the estimation of intervention-specific survivor functions for the technical recurrence data as required for the economic model, a separate random effects meta-analysis was conducted on the stripping intervention arms. Absolute estimates of survivor functions (no technical recurrence function) and population mean times to technical recurrence were estimated for each intervention by projecting the estimates of treatment effect from the network meta-analysis onto the baseline survivor function.

The method of analysis for technical recurrence differed from what was described in the protocol (i.e. an analysis if binary data with results presented as odds ratios) to enable an adjustment for variation between studies in the duration of follow-up.

Quantity and quality of research available

Characteristics of included studies

The searches identified 1453 unique citations. One hundred and twelve full papers were retrieved as being potentially relevant. Forty-five of these papers were excluded for at least one of the following reasons: not a RCT; correspondence relating to a relevant RCT; RCTs of comparator interventions only; RCT of co-interventions; duplicate publications' or not available (see Appendix 6). Eleven citations represented relevant ongoing trials64–74 and none of the available data were appropriate for analysis in one study. 75,76 Fifty-four citations, representing 34 different studies, therefore provided data for analysis (see the PRISMA flow chart, Figure 1).

FIGURE 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart.

There was a total of 3873 participants across all trials in which randomised numbers in each arm were reported. The number of randomised participants in a single trial ranged from 2845 to 710. 77 Where diagnostic information was reported, all participants received a diagnosis using duplex scanning. Only three papers failed to report this information; all were abstracts. 78–80

The mean age of participants ranged from 3381 to 54 years. 82,83 There was a majority of female participants in every trial; the percentage of female participants ranged from 54%81 to 95%. 84 In all trials participants were required to have varicose veins diagnosed by duplex scanning and categorised according to the CEAP score. The vast majority of participants in any trial were C2 on the CEAP score (varicose veins). This was not the case in only 3 of the 34 trials, in which the majority were C3,83 C478 or C5. 85 The UK was the single most frequent location (12 trials39,79,83,86–94); the remainder were conducted in centres across 14 other countries (USA, Brazil, China, Egypt, Austria, Denmark, Finland, France, Germany, Ireland, the Netherlands, Spain, Sweden and Switzerland) (see Tables 2–9 for a summary of the included studies' characteristics).

Fourteen trials53,55,84,87,88,95–103 evaluated EVLA, but Rasmussen et al. 95 was a multiarm trial with more than one comparator. Eight compared the intervention with a form of conventional surgery (Table 2):84,86,87,95–100 six with RFA (Table 3);53,55,88,95,101,102 Disselhoff et al. 103 with cryostripping; and Rasmussen et al. 95 with FS (Table 4).

TABLE 2 - Study characteristics by intervention and comparator: EVLA vs. conventional surgery

C, comparator; GA, general anaesthetic; I, intervention; I1, intervention 1; I2, intervention 2; LA, local anaesthetic; NR, not reported.

Study and location	Unit of randomisation	n	Age (mean in years)	Sex (female/male)	CEAP score (C2–C6)	GA/LA
Carradice 2011,86,96 UK	Patient	I: 139	I: 49	I: 85/54	I: C2 = 95; C3–6 = 43; unknown = 1	I: NR
		C: 137	C: 49	C: 90/47	C: C2 = 96; C3–6 = 41	C: NR
		Total: 276
Christenson 2010,97 Switzerland	Limb (mixed unilateral and bilateral)	I: 100	I: 45	I: 67/33	I: C2 = 34; C3 = 58; C4 = 7; C5 = 1; C6 = 0	I: NR
		C: 100	C: 47	C: 71/29	C: C2 = 26; C3 = 51; C4 = 18; C5 = 2; C6 = 3	C: NR
		Total: 200
Darwood 2008,87 UK	Patient (unilateral and bilateral)	I1: 47	I1: 42	I1: 22/16	I1: C2 = 37; C3 = 4; C4 = 2; C5 = 3; unknown = 1	I: LA
		I2: 33	I2: 52	I2: 16/11	I2: C2 = 24; C3 = 6; C4 = 1; C5 = 0; unknown = 2
		C: 34	C: 49	C: 16/14	C: C2 = 23; C3 = 9; C4 = 0; C5 = 1; unknown = 1	C: GA
		Total: 114
De Medeiros 2005,84 Brazil	Limb (bilateral)	Total: 20	46	19/1	I: C2 = 9; C3 = 2; C4 = 3; C5 = 4; C6 = 2	I: LA
					C: C2 = 11; C3 = 5; C4 = 3; C5 = 1; C6 = 0	C: LA
Kalteis 2008,98,104 Austria	Patient	I: 47	I: 46	I: 37/10	I: C2 = 74%; C3 = 19%; C4 = 7%	I: NR
		C: 48	C: 47	C: 34/14	C: C2 = 69%; C3 = 27%; C4 = 4%	C: NR
		Total: 95
Rasmussen 2007,44 2009,105 2010,99 Denmark	Patient	I: 69	I: 53	I: 41/21	I: C2 = 50; C3 = 3; C4 = 9	I: LA
		C: 68	C: 54	C: 43/16	C: C2 = 51; C3 = 5; C4 = 3	C: LA
		Total: 137
Rasmussen 2011,95 Denmark	Patient	I: 125	I: 52	I: 72% female	I: C2–3 = 95%; C4–6 = 5%	I: LA
		C: 124	C: 50	C: 77% female	C: C2–3 = 97%; C4–6 = 3%	C: LA
		Total: 249
Pronk 2010,100,106 Netherlands	Patient	I: 62	I: 49	I: 46/16	I: C2 = 29; C3 = 29; C4 = 4; C5 = 0	I: LA
		C: 68	C: 50	C: 53/15	C: C2 = 26; C3 = 36; C4 = 5; C5 = 1	C: LA
		Total: 130

TABLE 3 - Study characteristics by intervention and comparator: EVLA vs. RFA

C, comparator; GA, general anaesthetic; I, intervention; LA, local anaesthetic; NR, not reported.

Study and location	Unit of randomisation	n	Age (mean in years)	Sex (female/male)	CEAP score (C2–C6)	GA/LA
Nordon 2011,88 UK	Patient	I: 80	I: 47	I: 54/26	I: C2 = 68; C3 = 3; C4–6 = 9	I: GA
		C: 79	C: 47	C: 45/34	C: C2 = 68; C3 = 2; C4–6 = 9	C: GA
		Total: 159
Gale 2009,107 2010,53,108 USA	Patient	I: 48	I: 49	I: 36/12	NR	I: LA
		C: 46	C: 46	C: 29/17	NR	C: LA
		Total: 94
Goode 2008,92,109,110 2010,55 UK	Bilateral: limb Unilateral: patient	I: Bilateral 17; unilateral 22	I: Bilateral 47; unilateral: 48	I: Bilateral 15/2; unilateral 15/7	C2 only	I: GA
		C: Bilateral 17; unilateral 23	C: Bilateral 47; unilateral 45	C: Bilateral 15/2; unilateral:15/8	C2 only	C: GA
		Total: 79
Rasmussen 2011,95 Lawaetz 2010,111 Denmark	Patient	I: 125	I: 52	I: 72% female	I: C2–3 = 95%; C4–6 = 5%	I: LA
		C: 125	C: 51	C: 70% female	C: C2–3 = 92%; C4–6 = 8%	C: LA
		Total: 250
Morrison 2005,101 USA	Bilateral	Total: 50	NR	NR	NR	NR
Shepherd 2009,112 2010,93,102 UK	Limb	I: 64	I: 48	I: 42/22	I: C1–2 = 26; C3–4 = 36; C5–6 = 2	I: GA
		C: 67	C: 49	C: 47/20	C: C1–2 = 23; C3–4 = 39; C5–6 = 5	C: GA
		Total: 131

TABLE 4 - Study characteristics by intervention and comparator: EVLA vs. others

C, comparator; GA, general anaesthetic; I, intervention; LA, local anaesthetic; NR, not reported.

Study and location	Intervention	Control	Unit of randomisation	n	Age (mean in years)	Sex (female/male)	CEAP score (C2–C6)	GA/LA
Disselhoff 2008,103 2011,113 Netherlands	EVLA	Cryostripping	Patient	I: 60	I: 46	I: 41/19	C2 only	LA or GA
				C: 60	C: 49	C: 42/18
				Total: 120
Rasmussen 2011,95 Lawaetz 2010,111 Denmark	EVLA	FS	Patient	I: 125	I: 52	I: 72% female	I: C2–3 = 95%; C4–6 = 5%	I: LA
				C: 124	C: 51	C: 76% female	C: C2–3 = 96; C4–6 = 4	C: LA
				Total: 249

Thirteen trials42,45,53,55,80,81,83,88,89,95,101,102,114 evaluated RFA, one of which, Rasmussen et al. ,95 had more than one comparator. Six trials compared the intervention with a form of conventional surgery (Table 5),42,45,81,83,89,95 six with EVLA (see Table 3),53,55,88,95,101,102 Stötter et al. 114 with invagination cryostripping and Rasmussen et al. 95 with FS, and Lin et al. 80 compared RFA in combination with TriVex with conventional surgery and TriVex (Table 6).

TABLE 5 - Study characteristics by intervention and comparator: RFA vs. conventional surgery

One participant was excluded post randomisation as it was found they did not satisfy inclusion criteria.

C, comparator; GA, general anaesthetic; I, intervention; LA, local anaesthetic; NR, not reported.

Study and location	Unit of randomisation	n	Age (mean in years)	Sex (female/male)	CEAP score (C2–C6)	GA/LA
ElKaffas 2011,81 Egypt	Patient	I: 90	I: 33	I: 48/42	I: C2 = 51; C3 = 27; C4 = 9; C5 = 3	I: LA
		C: 90	C: 35	C: 45/45	C: C2 = 45; C3 = 27; C4 = 12; C5 = 6	C: GA
		Total: 180
Hinchliffe 2006,83 UK	Bilateral: limb	Total: 16	54	12/4	C2 = 1; C3 = 14; C4 = 1	I: GA
						C: GA
Rasmussen 2011,95 Lawaetz 2010,111 Denmark	Patient	I: 125	I: 51	I: 76% female	I: C2–3 = 92%; C4–6 = 8%	I: LA
		C: 124	C: 50	C: 77% female	C: C2–3 = 97%; C4–6 = 3%	C: LA
		Total: 249
Rautio 2002,45 Perala 2005,43 Finland	Patient	I: 15	I: 33	I: 14/1	I: NR	I: NR
		C: 13	C: 38	C: 12/1	C: NR	C: NR
		Total: 28
Lurie 2003,42 2005,115 Multicentre	Patient	I: 44a	I: 49	I: 32/13	I: C2 = 36; C3 = 4; C4 = 4	LA or GA
		C: 36	C: 47	C: 26/10	C: C2 = 28; C3 = 4; C4 = 4
		Total: 81
Subramonia 2010,89 Balakrishnan 2008,116 UK	Patient	I: 47	I: 47	I: 34/13	I: C2 = 37; C3 = 9; C4–6 = 1	I: GA
		C: 41	C: 45	C: 27/14	C: C2 = 33; C3 = 7; C4–6 = 1	C: GA
		Total: 88

TABLE 6 - Study characteristics by intervention and comparator: RFA vs. various comparators

C, comparator; C1, comparator 1; C2, comparator 2; GA, general anaesthetic; I, intervention; LA, local anaesthetic; NR, not reported.

Study and location	Intervention	Control	Unit of randomisation	n	Age (mean in years)	Sex (male/female)	CEAP score	GA/LA
Lin 2009,80 China	RFA and TriVex	Stripping and TriVex	Patient	I: 75	I: NR	I: NR	I: NR	I: NR
				C: 75	C: NR	C: NR	C: NR	C: NR
				Total: 150
Stötter 2005,114 Germany	RFA	Invagination or cryostripping	Patient	I: 20	I: 43	I: 14/6	I: NR	I: NR
				C1: 20	C1: 53	C1: 15/5	C1: NR	C1: NR
				C2: 20	C2: 42	C2: 14/6	C2: NR	C2: NR
				Total: 60
Rasmussen 2011,95 Laaetz 2010,111 Denmark	RFA	FS	Patient	I: 125	I: 51	I: 70% female	I: C2–3 = 92%; C4–6 = 8%	I: LA
				C: 124	C: 51	C: 76% female	C: C2–3 = 96%; C4–6 = 4%	C: LA
				Total: 249

Thirteen trials77–79,85,90,91,95,117–122 evaluated FS, one of which, Rasmussen et al. ,95 had more than one comparator. Ten trials77–79,85,90,91,95,117,118,122 compared the intervention with a form of conventional surgery (Table 7), one of which, Liamis et al. ,79 used ‘reverse’ FS. Three trials119–121 compared the intervention with LS (Table 8), while Rasmussen et al. 95 compared it with both EVLA and RFA.

TABLE 7 - Study characteristics by intervention and comparator: FS vs. conventional surgery

C, comparator; GA, general anaesthetic; I, intervention; LA, local anaesthetic; NR, not reported; SFJ, saphenofemoral junction.

Study and location	Intervention	Control	Unit of randomisation	n	Age (mean in years)	Sex (female/male)	CEAP score	GA/LA
Abela 2008,91 UK	‘Reverse’ FS	Stripping	NR	I: 30	I: 45	I: 22/8	All C2 and C3	I: NR
				C1: 30	C1: 46	C1: 17/13		C1: NR
				C2: 30	C2: 47	C2: 15/15		C2: NR
				Total: 30
Bountouroglou 2004,123 2006,90 UK	FS and SFJ ligation	Stripping	Patient	I: 30	I: 42	I: 14/16	I: C2 = 11; C3 = 8; C4 = 7; C5 = 3; C6 = 1	I: NR
				C: 30	C: 43	C: 18/12	C: C2 = 8; C3 = 14; C4 = 6; C5 = 1; C6 = 1	C: NR
				Total: 58
Figuereido 2010,85 Brazil	FS	Stripping	NR	I: 27	I: 53	I: 23/4	C5 only	I: NR
				C: 29	C: 49	C: 23/6		C: NR
				Total: 56
Jia 2010,78 China	FS and SFJ ligation	Stripping	Patient	I: NR	I: NR	I: NR	Median C4 in both groups	I: NR
				C: NR	C: NR	C: NR		C: NR
				Total: 60
Kalodiki 2008,94 2011,117 UK	FS and SFJ ligation	Stripping	NR	I: 43	I: 49	I: 32/11	All C2–C6, similar between groups	I: LA
				C: 39	C: 47	C: 23/16		C: GA
				Total: 82
Liamis 2005,79 UK	‘Reverse’ FS and SFJ ligation	Stripping	Limb	I: 30	I: NR	I: NR	I: NR	I: NR
				C: 30	C: NR	C: NR	C: NR	C: NR
				Total: 60
Rasmussen 2011,95 Lawaetz 2010,111 Denmark	FS	Stripping	Patient	I: 124	I: 51	I: 76% female	I: C2–3 = 96%; C4–6 = 4%	I: LA
				C: 124	C: 50	C: 77% female	C: C2–3 = 97%; C4–6 = 3%	C: LA
				Total: 248
Shadid 2010,122 Netherlands	FS	Stripping	Patient	I: 227	I: NR	I: NR	I: NR	I: NR
				C: 198	C: NR	C: NR	C: NR	C: NR
				Total: 425
Wright 2006,77 Europe	FS	Stripping	Patient	I: 178	I: 50	I: 112/66	I: C2 = 144; C3 = 14; C4 = 20	I: NR
				C: 94	C: 49	C: 60/34	C: C2 = 73; C3 = 11; C4 = 10	C: NR
				Total: 272

TABLE 8 - Study characteristics by intervention and comparator: FS vs. LS

C, comparator; GA, general anaesthetic; I, intervention; LA, local anaesthetic; NR, not reported.

Study and location	Intervention	Control	Unit of randomisation	n	Age (mean in years)	Sex (female/male)	CEAP score (C2–C6)	GA/LA
Alos 2006,119 Spain	FS	LS	Limb/region – bilateral	Total: 75	59	Total: 69/6	Total: NR	Total: NR
Hamel-Desnos 2006,120 Ouvry 2008,82 France	FS	LS	Patient	I: 45	I: NR	I: NR	I: NR	I: NR
				C: 43	C: NR	C: NR	C: NR	C: NR
				Total: 88
Rabe 2008,121 Germany	FS	LS	Patient	I: 54	I: 51	I: 35/19	I: C2 = 26; C3 = 15; C4 = 12; C5 = 1	I: NR
				C: 52	C: 50	C: 39/13	C: C2 = 26; C3 = 14; C4 = 8; C5 = 4	C: NR
				Total: 106

Finally, Aremu et al. 124 compared TIPP with conventional surgery and Chetter et al. 39 compared TIPP with standard multistab incision phlebectomy (MSIP) (Table 9).

TABLE 9 - Study characteristics by intervention and comparator: TIPP vs. various comparators

C, comparator; GA, general anaesthetic; I, intervention; LA, local anaesthetic; NR, not reported.

Study and location	Intervention	Control	Unit of randomisation	n	Age (mean in years)	Sex (male/female)	CEAP score	GA/LA
Aremu 2004,124 Ireland	TIPP	Stripping	Limb (unilateral and bilateral)	I: NR	I: NR	I: NR	I: C2 = 53%; C3 = 47%	I: NR
				C: NR	C: NR	C: NR	C: C2 = 61%; C3 = 39%	C: NR
				Total: 141
Chetter 2005,39 UK	TIPP	MSIP	Patient	I: 29	I: 48	I: 19/10	I: C2 = 27; C5 = 2	I: NR
				C: 33	C: 50	C: 24/9	C: C2 = 29; C4 = 4;	C: NR
				Total: 62

No trial included conservative management as a comparator. Only three trials119–121 included LS as a comparator (with FS) (see Table 8). The principal common comparator was therefore surgery (i.e. ligation and stripping).

Quality of included studies

The methodological quality assessment of each included study is summarised in Figure 2 and Table 10. Only the findings affecting the most pertinent criteria are outlined here. Twelve of the 34 included trials failed adequately to report methods of either randomisation or allocation concealment, or reported inadequate methods. 77–80,83,98,101,114,117,120–122 Two further trials reported adequate randomisation but inadequate allocation concealment. 97,100 These studies, as reported, therefore had a high risk of selection bias. Thirteen studies either clearly failed to conduct an intention-to-treat analysis or left it unclear whether or not they had done so, and so were categorised as not clearly conducting such an analysis;42,53,55,77,79,80,88,91,102,114,117,122,124 six of these also failed to report adequate methods of randomisation and allocation concealment,77,79,80,114,117,122 all but one of which were abstracts only. 114

FIGURE 2.

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

TABLE 10 - Methodological quality summary: review authors' judgements about each methodological quality item for each included study

N, no; NA, not applicable; Y, yes.

Abstract only.

Article claims that the population was ‘homogenous’, therefore ‘Y’, but this could not be verified.

Study	Randomisation?	Allocation concealment?	Inclusion/exclusion criteria defined?	Intention-to-treat analysis performed?	Matched groups at baseline (or adjustment of results if not)?	Surgeons sufficiently experienced at each procedure?	Care programmes identical other than surgical intervention?	Outcomes clearly defined?	Appropriate length of follow-up to measure recurrence?	Attrition: ≤5% lost to follow-up?
Abela 200891	Y	Y	Y	N	Y	N	Y	Y	NA	N
Alos 2006119	Y	Y	Y	Y	Y	N	Y	Y	Y	Y
Aremu 2004124	Y	Y	N	N	N	N	Y	Y	Y	N
Bountouroglou 200690	Y	Y	Y	Y	Y	N	Y	Y	N	N
Carradice 201186,96	Y	Y	Y	Y	Y	N	Y	Y	Y	N
Chetter 200639	Y	Y	Y	Y	N	N	Y	Y	NA	Y
Christenson 201097	Y	N	Y	Y	N	N	Y	Y	Y	Y
Darwood 200887	Y	Y	Y	Y	N	N	N	Y	Y	N
De Medeiros 200584	Y	Y	Y	Y	Y	N	Y	Y	NA	Y
Disselhoff 2008,103 2011113	Y	Y	Y	Y	Y	Y	Y	Y	Y	N
ElKaffas 201181	Y	Y	Y	Y	N	N	N	Y	Y	N
Figueiredo 200985	Y	Y	Y	Y	N	N	N	Y	NA	N
Gale 201053	Y	Y	Y	N	Y	N	Y	Y	Y	N
Goode 201055	Y	Y	Y	N	N	N	Y	Y	N	Y
Hamel-Desnos 2003120	N	N	Y	Y	Yb	N	Y	Y	Y	N*
Hinchliffe 200683	N	N	Y	Y	N	N	Y	Y	NA	Y
aJia 201078	N	N	N	Y	N	N	N	N	N	N
aKalodiki 2008117	N	N	N	N	Y	N	N	Y	Y	N
Kalteis 200898	N	N	Y	Y	Y	N	Y	Y	NA	Y
aLiamis 200579	N	N	N	N	N	N	N	N	N	N
aLin 200980	N	N	N	N	N	N	N	N	N	N
Lurie 200342	Y	Y	Y	N	Y	N	Y	Y	Yb	Nb
aMorrison 2005101	N	N	N	Y	N	N	N	Y	Y	Y
aNordon 201188	Y	Y	Y	N	Y	N	Y	Y	N	Y
Perala 2005,43 Rautio 200245	Y	Y	Y	Y	N	N	Y	Y	NA	Y
Pronk 2010100	Y	N	Y	Y	Y	Y	Y	Y	Y	Y
Rabe 2007121	N	N	Y	Y	Y	N	Y	Y	N	Y
Rasmussen 2007, 201044,99	Y	Y	Y	Y	Y	Y?	Y	Y	Y*	Y
Rasmussen 201195	Y	Y	Y	Y	Y	N	N	Y	Y	N
aShadid 2010122	N	N	Y	N	N	N	N	Y	Y	Y
Shepherd 2010102	Y	Y	Y	N	Y	Y	Y	Y	NA	N
Stötter 2006114	N	N	Y	N	N	N	N	Y	Y	Y
Subramonia 201089	Y	Y	Y	Y	Y	N	Y	Y	N	Y
Wright 200677	N	N	Y	N	N	N	N	Y	Y	N

Two further key criteria were assessed that had the greatest potential to confound the results of this report: the non-comparability of groups at baseline and non-identical care programmes post intervention. Sixteen of the trials reported statistically significant or substantial differences between within-study groups in terms of potential confounders (such as age or CEAP score),39,45,55,77–81,83,85,87,97,101,114,122,124 while 12 of 34 trials either reported non-identical care programmes post intervention or did not make it clear what occurred. 77–81,85,87,95,101,114,117,122 An assessment of reporting bias was deemed not to be possible because no published protocol could be identified for any included trial. This criterion is therefore omitted from the tables.

Studies published as abstracts only would obviously potentially be considered to be at higher risk of bias than studies published as full papers because details of study conduct were often not reported in that abbreviated format. Twelve studies reported in full papers were at risk of two or more of these forms of bias,55,77,81,83,85,87,97,98,114,120,121,124 as were six studies reported as abstracts only. 78–80,101,117,122 The remaining 16 trials were at risk of either one or none of these forms of bias. 39,42,45,53,84,86,88–91,95,99,100,102,113,119

The majority of the trials used in the network meta-analyses (e.g. those reporting technical recurrence data for EVLA vs. stripping or EVLA vs. RFA, etc.) were at risk of either selection or attrition bias due to inadequate randomisation, allocation concealment or intention-to-treat analysis. Given that these types of bias potentially affected almost all studies included in the analysis, no sensitivity analysis was performed based on these quality appraisal criteria.

Assessment of effectiveness

First, a narrative synthesis is provided for all outcomes: failure of procedure, recurrence, VCSS, pain score, return to work or normal activity, and adverse events. Second, Quantitative analysis will present a quantitative synthesis using a formal network meta-analysis approach for those outcomes for which sufficient and appropriate data were available: technical recurrence, VCSS and pain score. Randomisation and analysis in the primary studies was described as being by patient or by limb when patients were unilateral; when patients were bilateral, randomisation was by limb. Data were therefore all per limb or per patient; no data were per procedure (i.e. there were no data where multiple procedures were conducted on the same limb).

Quantitative analysis

Technical recurrence

A network meta-analysis was used to compare the hazard of having technical recurrence when treating with EVLA, RFA and FS compared with stripping for 6 months, 1 year and 2 years. These durations were a pragmatic decision in that we were looking at short- and medium-term results and there was not expected to be a great difference between the data for 6 months and 2 years. It was the only viable approach for generating a network because different trials used different lengths of follow-up. A total of 23 studies42,45,53,55,71,72,75,77,78,80–86,88,95,96,99,100,110,114 comparing pairs or quadruplets of interventions provided information at various follow-up times. Bountouroglou et al. 90 and Kalteis et al. 98 were excluded from the analysis because there were no events in either intervention arm and, as a consequence, these studies provided no information about the intervention effects. 129

Figure 3 presents the network of evidence. A summary of all the trials (data) included in the base-case network meta-analysis for technical recurrence is presented in Appendix 7.

FIGURE 3.

Technical recurrence: network diagram of different interventions vs. stripping. The nodes represent the interventions. Lines between nodes indicate when interventions have been compared. The numbers against each line represent the number of times that each pair of interventions has been compared.

The network meta-analysis model fitted the data reasonably well, with a total residual deviance close to the total number of data points included in the analysis. The total residual deviance was 64.11, which compared favourably with the 60 non-zero data points being analysed. Figure 4 presents the fitted survivor function (i.e. probability of no technical recurrence) for each intervention.

FIGURE 4.

Technical recurrence: survivor function (no technical recurrence function) of each intervention (median).

The results suggested that there was mild heterogeneity between studies in the shape parameter, but that there was mild to potentially moderate heterogeneity between studies in the scale parameter (Table 19).

TABLE 19 - Technical recurrence: posterior distribution of the between-study SD for the Weibull shape and scale parameter (random effects)

Parameter	Median (95% CrI)
Between-study SD (Weibull shape parameter – natural scale)	0.17 (0.01 to 0.45)
Between-study SD (Weibull scale parameter – log scale)	0.26 (0.02 to 0.91)

Endovenous laser ablation exhibited the greatest effect on technical recurrence relative to stripping, although there was some evidence that the benefit decreases over time (2-year HR 0.84, 95% CrI 0.44 to 1.81) (Table 20). RFA was associated with a small and relatively constant effect on technical recurrence over time relative to stripping (2-year HR 0.94, 95% CrI 0.42 to 2.51). FS was worse than stripping over the first year, although there was a small benefit after 2 years (2-year HR 0.92, 95% CrI 0.43 to 1.60). In each case there was considerable uncertainty about which intervention was the most beneficial.

TABLE 20 - Technical recurrence: posterior distribution for the hazard ratios relative to stripping at 6 months, 1 year and 2 years (random effects)

Comparison	Median (95% CrI) [probability hazard ratio > 1]
	6 months	1 year	2 years
EVLA vs. stripping	0.70 (0.27 to 1.45) [0.150]	0.77 (0.37 to 1.54) [0.182]	0.84 (0.44 to 1.81) [0.257]
RFA vs. stripping	0.92 (0.39 to 2.11) [0.409]	0.93 (0.42 to 2.22) [0.411]	0.94 (0.42 to 2.51) [0.421]
FS vs. stripping	1.12 (0.53 to 2.27) [0.659]	1.02 (0.49 to 1.84) [0.524]	0.92 (0.43 to 1.60) [0.359]

Venous Clinical Severity Score

Venous Clinical Severity Score was analysed based on the data available at 1 year. However, for studies that did not provide 1-year data, the 6-month data were used, or the first available value after 1 year. A total of six studies43,53,86,97,102,117 were selected for the analysis. All studies were two-arm trials. Among all the studies, Carradice et al. 86,96 and Christenson et al. 97 compared EVLA with stripping; Perala et al. 43 compared RFA with stripping; Kalodiki et al. 94,117 compared FS with stripping; Gale et al. 53 and Shepherd et al. 102 compared RFA with EVLA. Four out of six studies reported one of median, interquartile range (IQR) or range and it was not possible to estimate the sample mean and sample SD as required for the network meta-analysis. 44,87,90,113 The authors were contacted and asked to provide the sample means and sample SDs from their studies; three authors provided the required sample means and sample SDs (see Table 15). 53,94,96 The missing SD from the Perala et al. study43 was treated as another uncertain parameter in the analysis. 130

Figure 5 presents the network of evidence. A summary of all the trials (data) included in the base-case network meta-analysis is presented in Appendix 8.

FIGURE 5.

Venous Clinical Severity Score: network diagram of different interventions vs. stripping. The nodes represent the interventions. Lines between nodes indicate when interventions have been compared. The numbers against each line represent the number of times that each pair of interventions has been compared.

The network meta-analysis model fitted the data well, with the residual deviance close to the total number of data points included in the analysis. The total residual deviance was 11.47, which compared favourably with the 12 data points being analysed. The between-study SD was estimated to be 0.22 (95% CrI 0.01 to 1.79) (Table 21). The intervention that exhibited the greatest effect relative to stripping was FS (MD –1.63, 95% CrI –2.90 to –0.42).

TABLE 21 - Venous Clinical Severity Score: posterior distribution for the MD compared with stripping (random effects)

Comparison and parameter	Median (95% CrI)	Probability of MD > 0
EVLA vs. stripping	–0.10 (–0.94 to 0.73)	0.324
RFA vs. stripping	0.15 (–0.50 to 0.95)	0.739
FS vs. stripping	–1.63 (–2.90 to –0.42)	0.015
Between-study SD	0.22 (0.01 to 1.79)

Pain

The effect of interventions on pain was assessed using a VAS based on data available within 7–14 days of treatment; all but two of the studies (see Table 16) measured pain scores at either 7 or 10 days (Christenson et al. 97 measured the pain score at 12 days and Rautio et al. 45 measured it at 14 days). Data were available from nine studies83,87–89,95,98,100,102,103 comparing pairs or quadruplets of interventions.

Among all the studies, four studies compared EVLA with stripping;87,95,97,98 three studies compared RFA with stripping;45,83,89 two studies compared RFA with EVLA;88,102 and one study had all four intervention arms. 95 Four studies reported a median, lower quartile and upper quartile. 83,87,89,98 To estimate the sample mean and sample SD from these studies, while acknowledging the skewness in the distribution, a gamma distribution was fitted to the median and interquartiles using least squares. Christenson et al. 97 reported only the mean and Nordon et al. 88 reported only the median, which we take as an estimate of the mean assuming that the data are normally distributed. The missing SDs were treated as additional uncertain parameters. 130

Figure 6 presents the network of evidence. A summary of all the trials (data) included in the base-case network meta-analysis is presented in Appendix 9.

FIGURE 6.

Pain score: network diagram of different interventions vs. stripping. The nodes represent the interventions. Lines between nodes indicate when interventions have been compared. The numbers against each line represent the number of times that each pair of interventions has been compared.

The network meta-analysis model represented the data well, with the residual deviance close to the total number of data points in the analysis. The total residual deviance was 22.29, which compares favourably with the 22 data points being analysed. The between-study SD was estimated to be 0.48 (95% CrI 0.06 to 1.12) (Table 22). The interventions that exhibited the greatest effects on pain relative to stripping were RFA (MD –1.26, 95% CrI –1.95 to –0.61) and FS (MD –0.80, 95% CrI –1.93 to 0.30).

TABLE 22 - Pain scores: posterior distribution for the MD comparing to stripping (random effects)

Comparison and parameter	Median (95% CrI)	Probability of MD > 0
EVLA vs. stripping	0.10 (–0.49 to 0.64)	0.653
RFA vs. stripping	–1.26 (–1.95 to –0.61)	0.001
FS vs. stripping	–0.80 (–1.93 to 0.30)	0.062
Between-study SD	0.48 (0.06 to 1.12)

Discussion

The analysis of the technical recurrence data suggested that the treatment effects for EVLA and FS were not constant over time. In particular, the early benefit associated with EVLA was much less, relative to stripping, after 2 years (15% reduction) than it was at 6 months (30% reduction). However, the results were inconclusive in determining which intervention was the most effective.

A benefit of the analysis of the technical recurrence data was that it did not assume proportional hazards; this is particularly important in terms of the assessment of cost-effectiveness as it does not assume that any treatment effect continues indefinitely. There were several limitations associated with the analysis of the technical recurrence data. In general, studies did not account for all patients at each follow-up time so that the technical recurrence response rates did not increase monotonically. Although authors often reported that theirs was an intention-to-treat analysis, some authors reported results as the number of events out of the number of patients randomised, whereas others reported results as the number of events out of the number of patients for which there were data. Some patients were assessed for their varicose veins in both limbs and results were often reported by limb rather than by patient; results will be correlated with patients and the analysis of the patient-level data should allow for variation within patients, which we were unable to do at the aggregate level. We fitted a Weibull model to the data, which effectively assumes that all patients will have a technical recurrence at some stage in the future; in practice, it is likely that a proportion of patients would never have a technical recurrence, and that a more appropriate model would be a ‘cure’ model in which the time to recurrence is conditional on not being ‘cured’, although it was not possible to do this with the data that were available. Some studies presented response rates at more than one time, which meant that we could estimate more than one parameter (i.e. the shape and scale parameter in the Weibull distribution). However, our model assumed that the observations were independent, which may have led to an overestimation of uncertainty.

The analysis of the VCSS data suggested that FS was the most effective intervention and was more effective than stripping. The analysis of the pain score data suggested that RFA was the most effective treatment and was more effective than stripping.

Identification of studies

A comprehensive search was undertaken to identify systematically cost-effectiveness literature comparing the different interventions for managing varicose veins. The search involved combining terms for the population (varicose veins) with terms for the interventions of interest (i.e. the minimally invasive techniques) and a filter designed to retrieve cost-effectiveness studies. An example MEDLINE search strategy is reported in Appendix 1. The aim of the strategy was to identify all studies that reported on costs and related analyses associated with the techniques of interest. Initial cost-effectiveness searches were performed by an information specialist (AC) in July 2011. Additional cost-effectiveness searches were completed in October 2011 and September 2012 to identify studies with costs associated with varicose veins. An example search strategy is reported in Appendix 1.

The following electronic databases were searched from inception for published and unpublished research evidence:

• MEDLINE (Ovid) 1950–

• EMBASE (Ovid) 1980–

• CINAHL (EBSCO) 1982–

• the Cochrane Library limited to HTA and NHS EED databases 1991–

• BIOSIS Previews (via ISI Web of Science) 1969–

• SCI (via ISI Web of Science) 1900–

• SSCI (via ISI Web of Science) 1956–

• EconLit (Ovid) 1961–.

Searches for utilities associated with the interventions of interest for treating varicose veins were performed by an information specialist (AC) in July 2011. An example search strategy is provided in Appendix 1.

The following databases were searched from inception for published and unpublished research evidence:

• MEDLINE (Ovid) 1950–

• EMBASE (Ovid) 1980–

• CINAHL (EBSCO) 1982–

• the Cochrane Library including the CDSR, CENTRAL, DARE, HTA and NHS EED databases 1991–

• Biological Abstracts (BIOSIS Previews) (via ISI Web of Science) 1969–

• SCI (via ISI Web of Science) 1900–

• SSCI (via ISI Web of Science) 1956–

• EconLit (Ovid) 1961–

• the Cost-Effectiveness Analysis Registry 1976–.

All citations were imported into Reference Manager Version 12 and duplicates deleted. Titles and abstracts of all unique citations were then screened independently by the cost-effectiveness reviewer (SH) using the inclusion criteria outlined below after a test screen on a sample of citations. Disagreements or queries were resolved by consensus or with reference to a second team member (CC or JM) where necessary. The full papers of all potentially relevant citations were then retrieved so that an in-depth assessment concerning inclusion could be made.

Methods

Study inclusion criteria were the same as for the clinical effectiveness review in terms of treatments and populations, but no limitation was put on study design. Additionally, studies had to report economic outcomes in terms of cost-effectiveness, cost–utility or cost–benefit. Search results were sifted, with the number of studies retained at each stage shown in Figure 7. Included studies were quality assessed according to study design. Primary economic analyses conducted alongside clinical trials were assessed using the checklist by Drummond et al. ;131 modelling studies using a checklist modified from Eddy. 132 These evaluations are reported in full in Appendix 10.

FIGURE 7.

Cost-effectiveness PRISMA flow chart.

Results

In total, four relevant economic studies were identified, two economic analyses conducted alongside RCTs133,134 and two modelling analyses. 135,136

The RCT compared cryostripping with EVLA in patients with primary symptomatic varicose veins, CEAP clinical class 2 (Disselhoff et al. 133). The second RCT compared RFA with surgery in primary or recurrent lower limb varicose veins, CEAP classes 2–6, though approximately 80% were C2 (Subramonia and Lees134).

The study by Disselhoff et al. 133 is of poor quality (see Appendix 5) and with a major flaw: the incremental cost-effectiveness ratios (ICERs) have been calculated incorrectly, calculating the cost per quality-adjusted life-year (QALY) for individual treatments before subtracting one from the other, instead of calculating the incremental cost per incremental QALY. However, given the limited evidence and sufficient data for recalculation of the ICERs the study is included in the economic review.

Details of disease recurrence, further treatments and utility [measured by Short Form questionnaire-6 Dimensions (SF-6D)] were recorded over 2 years following the initial intervention. The costs of lost productivity were also included in the analysis, but sufficient data are presented to allow recalculation of ICERs excluding these, for a health costs perspective. Given the very small differences in total costs and QALYs between the two treatments, the assumption that treatment costs for cryostripping and EVLA are the same apart from equipment costs is a limitation of this study. A further limitation is that patients chose whether or not to be treated as a day case or outpatient, with their costs of treatment inferred from this choice. Eighty-two per cent of cryostripping patients and 66% of EVLA patients had a day-case procedure, which was assumed to be more costly than an outpatient procedure.

Table 23 shows costs, QALYs and ICERs for EVLA in comparison with cryostripping, with the costs in the original units (it is assumed that costs are for 2003, but this is unclear). When costs were converted to pounds sterling (assuming the current exchange rate of €0.787 to £1) and inflated to 2011/12 prices using Health and Community Health Services inflation indices (1.27), the conversion factor was equal to 1.000. Therefore, the costs and ICERs shown in Table 23 can also be interpreted as £2011/12. The source data are from Disselhoff et al. ,133 but with the following adjustments:

• Costs of sick leave are excluded, to give the analysis a health cost perspective.

• An all-patient cost, taking into account the proportion of patients in each trial arm having day or outpatient care, was not calculated (see above).

• The QALYs shown in Table 23 are for all patients in each arm. Given the small patient numbers (n = 120) and uncertainty in the mean values reported by Disselhoff et al. ,133 distinguishing between those for outpatient and day care is not useful.

These results show EVLA to be both marginally more effective and more expensive than cryostripping over 2 years. The ICER for EVLA in comparison with cryostripping is €32,265, but the costs of EVLA fell during the course of the study, and if the lower cost is used the ICER falls to €15,365. However, the differences in costs and QALYs are very small relative to their uncertainties, so effectively the conclusion of this study is that the treatments are similar in terms of both costs and outcomes.

In the study by Subramonia and Lees134 resource use was collected alongside a clinical trial comparing RFA (n = 47) with surgery (n = 41). Costs include hospital costs (theatre and staff time, overheads, day ward, scans, non-protocol outpatient visits), primary health care and costs to patients. Protocol-driven costs that were the same for both treatments were excluded. These comprised hospital visits, scans and district nurse visits. The number of days of work lost by patients following the procedures was also recorded. A summary of the costs and outcomes for RFA and surgery is shown in Table 24.

Although the study generally appears to be of good quality (see Appendix 5), the incremental cost-effectiveness result quoted for RFA in comparison with surgery (incremental societal cost per work hour gained of £8.14) is clearly incorrect, including a valuation of the work hours gained in the numerator, and actual hours in the denominator. In fact, with an incremental health-care cost of £706, the incremental health costs per work hour gained are £18.11. If the work hours are valued, the total additional cost to society of RFA in comparison with surgery is £318. From a health-care perspective there was an incremental cost of £706 for an incremental improvement in the Aberdeen Varicose Veins Questionnaire (AVVQ) score at a median of 37 days of 0.88, which gives a cost of £806 per incremental unit improvement in AVVQ score. The health-care costs for RFA and surgery inflated to £2011/12 are shown in Table 25 and Costs. The cost of RFA is £1525 and surgery £687, giving a cost differential of £838.

Two modelling studies were identified (Adi et al. 135 and Gohel et al. 136). That by Adi et al. 135 is a very simple attempt to estimate cost-effectiveness of RFA compared with surgery, based on the results of a single poor-quality RCT (see Chapter 3, Quality of included studies, Rautio et al. 45) with very limited follow-up (2 weeks), so no long-term outcomes were included. Utilities were estimated from mean pain VAS scores for the two treatment arms, using a relationship between pain VAS and EQ-5D utility sourced from a study of back pain. 137 The latter gave a much steeper gradient in utility loss with increasing pain than that derived from a study of varicose veins (a utility decrement of 0.035 per 1% increase in pain VAS compared with 0.0026;86 see Utility values), suggesting that the QALY benefit of RFA compared with surgery reported by Adi et al. 135 may be overestimated by over 10 times. Costs were taken from the Finnish trial. 45 The difference in costs was £380 (year 2000 prices) and QALYs 0.016. The estimated ICER of RFA compared with surgery was £23,750, but with enormous uncertainty for the reasons described (see Appendix 5).

Gohel et al. 136 developed a Markov model to compare endovenous treatments for varicose veins (just FS, not UGFS, RFA and EVLA) with surgery over a 5-year time horizon. Model states include treatment success, residual varicosity, incomplete occlusion, recurrence of reflux and residual reflux. The study generally follows good modelling and reporting guidelines (see Appendix 5). Clinical effectiveness is based on two systematic literature reviews,29,48 and cost and utility data are adequately sourced, although the cost estimates are relatively simple. However, no information is reported (neither values nor source) regarding the likely important recurrence rates of reflux for treatments other than surgery. If it is assumed that the rate is the same for all treatments this is not made clear, nor the assumption discussed. Also the odds ratios for incomplete occlusion following treatment were from studies with different length of follow-up. The authors state that the ‘data suggested that the odds ratios for incomplete occlusion did not differ during follow-up’. However, the review (which does include non-randomised comparative studies) by van den Bos et al. ,49 which included a meta-regression of treatment success with time, indicates that relative success rates do vary with time from treatment. A further questionable assumption, which is not tested in sensitivity analysis, is that patients with residual varicosities and incomplete occlusion have the same utility value as untreated patients. In fact, the literature shows limited correlation between technical treatment success and patient symptoms. For example Merchant et al. 138 presents a chart indicating that 78% of limbs with anatomic failure following RFA were asymptomatic, compared with 90% of limbs which were classed as anatomic successes.

Costs of treatment (£2008/9) over 5 years varied from £0 for conservative treatment and £2000 for inpatient surgery. For active treatments the 5-year QALYs were fairly similar, ranging from 3.836 for UGFS performed under local anaesthetic (LA) to 3.958 for RFA with general anaesthetic (GA). Various ICERs are presented for different comparisons, but are not particularly informative. The probabilities of the treatments being cost-effective at a willingness-to-pay threshold of £20,000 are also reported. Those with the highest probabilities were EVLA – LA (35%), day-case surgery (29%) and RFA – LA (24%). The authors conclude that these are all likely to be cost-effective strategies for the treatment of varicose veins, although acknowledging considerable uncertainty in the results.

The uncertainty can be demonstrated by the monetary net benefits for each procedure. The results reported by Gohel et al. 136 were used to calculate these at a threshold of £20,000 (Table 26). This demonstrates very clearly that there is little difference in expected benefits between treatments, especially EVLA and RFA conducted under LA and day surgery.

Although sensitivity analysis was undertaken by Gohel et al. 136 some key assumptions were not tested (same recurrence rate for all treatments, utility value for clinically failed treatments). However, those that were undertaken illustrate the sensitivity of the results to model assumptions and uncertainty in model parameters.

Conclusions of economic review

The two economic studies carried out alongside clinical trials both had seriously flawed economic analyses, limiting the value of their conclusions. 133,134 Two studies compared RFA with surgery, both with short-term outcomes (< 40 days) (Subramonia and Lees134 and Adi et al. 135). The costs for both were based on resource use in a clinical trial, one in the UK134 and one in Finland. 45 Both showed the costs of RFA to be greater than surgery, but the cost difference was much greater in the UK study (£838, 2011/12) than that derived from the Finnish study45 by Adi et al. (£451, 2011/12). 135 Adi et al. 135 estimated a QALY gain derived from differences in mean pain VAS scores at 2 weeks after the procedures, giving an ICER for RFA compared with surgery of £23,750, although this result was highly uncertain. From a societal perspective, the additional costs of RFA compared with surgery were reduced (Subramonia and Lees134) or eliminated (Rautio et al. ,45 Adi et al. 135) by taking into account lost work days.

Results derived from Disselhoff et al. 133 show EVLA to be both marginally more effective and more expensive than cryostripping over 2 years. The ICER for EVLA in comparison with cryostripping is €32,265, but the differences in costs and QALYs are very small relative to their uncertainties, so effectively the conclusion of this study is that the treatments are similar in terms of both costs and outcomes.

One modelling study compared the principal endovenous treatments (FS, EVLA and RFA) with surgery. 136 Some key assumptions were not justified, or tested in sensitivity analysis. The modelled costs and benefits were very similar for all treatments, and demonstrate the sensitivity of the results to all assumptions.

Overall, the economic analyses of endovenous treatments in comparison with conventional treatment for varicose veins have been of limited scope and quality. They do demonstrate, however, that the differences in costs and benefits between treatments are small and sensitive to assumptions, and therefore the cost-effectiveness of the different procedures in relation to each other is likely to be uncertain, and vary by local costs.

Methods

The model structure

Outcomes of varicose vein procedures are complex. Several disease-specific quality-of-life measures have been developed for varicose veins in recognition of the fact that, although symptom relief is associated with clinical or anatomical outcomes, these are poor predictors of operative success from the patient's perspective. 139,140 In the model it is therefore not assumed that technical success equates to the patient being asymptomatic. Instead patients with technically successful and technically failed procedures have differing probabilities of being asymptomatic, with differing utility values attached to symptomatic and asymptomatic states. Ideally, direct HRQoL data would be used to model the outcomes of treatment with time, but there were insufficient data from the effectiveness literature with which to do this (see Chapter 3,Venous Clinical Severity Score).

The model structure is illustrated in Figure 8. Ovals represent events (numbered 1–3) and oblongs health states (A–D). Treatments included in the model are surgical stripping, FS, EVLA and RFA (Event 1). Costs and a loss of utility from the short-term adverse effects of treatment are assigned according to the treatment. Treatment may result in technical immediate (anatomical) success (states A and B) or failure (states C and D). If a failure, it is assumed that all patients will have further treatment with foam until technical success is achieved (Event 2; see Additional treatment to achieve a successful outcome). Patients with a successful clinical outcome nevertheless still have a probability of remaining symptomatic (state B). Thus, initial treatment may result in one of two health states (see Figure 8).

FIGURE 8.

Diagram of the model structure.

Patients whose treatment is successful nevertheless have a risk of failure over time (technical recurrence, Event 3). The proportion asymptomatic after recurrence is lower than for those whose treatment remains successful. It is assumed that only symptomatic patients with technical recurrence are retreated. It is assumed there is a delay of 6 months between treatment recurrence and retreatment as the development of symptoms may be gradual (‘time to retreatment’). The model was developed as a discrete event simulation (DES) model in Simul8^® to simulate the experience of patients undergoing treatment for varicose veins. The baseline model has a perspective of 10 years. This was chosen as a reasonable time over which to extrapolate the time to failure data. Of the studies included in the analysis most had a follow-up of a few months, with the longest at 2–3 years. Some cohort studies not included in the systematic review have followed patients for up to 10 years. The treatments for varicose veins considered in the model are for symptom relief and are assumed not to affect mortality, and therefore a lifetime time horizon is not necessarily the most appropriate. A DES model was chosen to allow non-constant hazard in the time to treatment failure/technical recurrence (i.e. not necessarily assume these are exponentially distributed). This method also obviates the need for arbitrary time cycles.

Patients may die at any time for any reason (all-cause mortality). 141 When patients enter the model the age-specific life expectancy distribution is sampled to determine their time to death. Patients with asymptomatic technical recurrence stay in the same state until their date of death. Patients whose treatment is successful might experience later disease recurrence if their sampled time to failure is less than their time to death. Symptomatic recurrences are retreated, after which they may be symptomatic or asymptomatic.

Treatments included are stripping, FS, EVLA and RFA. LS was omitted from the model because trials comparing LS and surgery were not reviewed (this was not a comparison of interest for this report). Consequently, the clinical effectiveness review only considered a part of the published LS data (from studies comparing LS with FS), which represents only a small amount of the potentially relevant data on LS. The data on LS reviewed in this report were therefore extremely limited and not included in the model.

The model can be run for cohorts of different ages entering the model, from age 40 to 80 years, in increasing decades. The methods for sourcing and deriving model parameters are described in the following section. The economic analysis is from the perspective of the UK NHS. All costs and benefits are discounted at a rate of 3.5%. A complete list of parameter values and the distributions used in the probabilistic sensitivity analysis are shown in Appendix 6.

Model data

Treatment effectiveness and adverse effects of treatment

Time to treatment failure/technical recurrence

Disease recurrence is labelled as Event 3 in Figure 8. Disease recurrence data were sourced from the meta-analysis of the effectiveness literature (see Chapter 3, Assessment of effectiveness). Uncertainty about the true time to technical failure was represented as probability distributions. These were computed by taking samples for the shape and scale parameters from the Weilbull distribution at each iteration of the Markov chain (see Chapter 3, Assessment of effectiveness) and using these as inputs.to the economic model. Correlated samples of alpha and beta parameters of the Weibull curves used to describe time to recurrence were generated for individuals from the mixed-treatment comparison of the failure data sourced from the effectiveness review. Treatment failure (technical recurrence) is defined as before. Initial treatment failure was defined as treatment failure within 1 month of the procedure, as determined by the failure curves (Event 2; see Figure 8). Figure 9 shows the survivor function for the different treatments (note it differs from Figure 4 in that it presents the plot of the mean of the individual Weibull parameters, rather than the median).

FIGURE 9.

Technical recurrence: survivor function (no technical recurrence function) of each intervention (mean).

The proportion of patients that is asymptomatic by technical success/failure

These proportions determine whether or not patients are in state A or B (see Figure 8) following treatment, or states C or D following disease recurrence. Very few included studies report symptomatic recurrence and none report asymptomatic recurrence (see Table 13); in fact, since the majority of procedures are technically successful it requires large cohorts, more likely achieved in observational studies, to identify the proportion asymptomatic in technically failed procedures. Two papers were identified with relevant data (Merchant and Pichot140 and Darvall et al. 142). The study by Merchant and Pichot140 is based on a multicentre prospective registry of patients treated with RFA, with up to 5 years of follow-up. A total of 1222 limbs were treated, of which 518 were available for follow-up at 6 months. Anatomical failure was identified using duplex imaging and defined as flow in a segment or all of treated vein, or groin reflux despite a completely occluded GSV trunk. The authors report: ‘clinical symptom improvement was seen in 70% to 80% of limbs with anatomical failures and in 85% to 94% of limbs with anatomical success from 6 months to 5 years after the radiofrequency obliteration’. From a figure showing the proportion of technical successes and failures that are symptomatic or not over the 5 years, the average proportion asymptomatic for the two groups was estimated. There is no indication that the proportion changes with time. The average proportion of technical successes with asymptomatic limbs is 89.3% and for technical failures 78.7%.

Darvall et al. 142 carried out a prospective cohort study on 246 patients having FS. 142 They examined the association between normalisation of venous refill time (a measure of technical success) and patient symptoms on the Aberdeen Varicose Vein Symptom Score (AVSS) scale 6 months after treatment. The proportion of patients with each symptom according to success or failure varies between symptoms and, in fact, only itching and swelling were significantly more common in patients with abnormal venous refill time. Overall, however, ‘relief of all symptoms was significantly more likely when there was normalisation of previously abnormal venous refill time (80% vs. 65%)’. 142

The somewhat lower rates of asymptomatic patients reported by Darvall et al. 142 than Merchant and Pichot140 may be due to consideration of more symptoms. Merchant and Pichot140 consider pain, fatigue and swelling, whereas Darvall et al. 142 also include itching, tingling, cramp, restless legs and heaviness, and are therefore likely to give a more sensitive measure of the patient being asymptomatic. However, duplex imaging, as used by Merchant and Pichot,140 to identify technical success is considered the gold standard. Venous refilling time was clearly somewhat problematical in the Darvall et al. 142 analysis as not all treated patients had abnormal venous refilling time prior to treatment, and therefore these patients had no potential for improvement on this measure. Although the two studies are measuring slightly different things, there is no clear reason for choosing one over the other. Moreover, their results are reasonably similar, so the results of both studies were combined by adding the number of patients in the two categories from both studies to give a proportion of asymptomatic for technical successes [0.88, standard error (SE) 0.014] and a relative risk of being asymptomatic if a technical failure (0.84, SE 0.048), resulting in the proportion asymptomatic for technical failures of 0.74. The former was characterised by a beta distribution based on the estimated numbers in each category, the latter with a gamma distribution based on the calculated log-normal distribution, but which prevented the implausible situation arising in probabilistic sensitivity analysis (PSA) runs of more failures being asymptomatic than successes. Note that there was no a priori reason or data to support different assumptions regarding the proportion of asymptomatic patients associated with technical treatment success or failure, so the same proportions were applied to all patients regardless of treatment. Technical failures are reassigned to asymptomatic or symptomatic at the time of failure independently of their originating state.

Adverse effects of treatment

Adverse events of treatment are presented in Chapter 3, Characteristics of included studies. Most adverse events of treatment are relatively mild and of short duration and require no treatment. They are potentially considered in the model in terms of the loss of utility (HRQoL) associated with each treatment (see Utility values). An exception is DVT, which can occasionally lead to death. DVT was therefore considered for inclusion in the model. However, the effectiveness review shows that DVT following treatment for varicose veins is very rare (see Chapter 3, Characteristics of included studies). There were insufficient data for meta-analysis. The lifetime discounted costs and QALYs associated with DVT can be estimated from an economic analysis of diagnostic tests for DVT. 143 It reports the total lifetime QALYs and costs for persons suffering from DVT with different diagnostic testing scenarios (see Appendix 5 and table 44 in Goodacre et al. 143). Total QALYs for a person with no DVT were 11.580, and with DVT 11.558 (scenarios 9 and 10, recommended in report as maximising net financial benefit, i.e. a QALY difference of 0.022). The total costs of DVT for scenarios 9 and 10 (including all subsequent treatment) were £248 and £245 respectively. So, for example, an absolute difference of 0.5% in DVT rate following different procedures would result in a QALY difference of 0.0001 and cost difference of £1.24. Thus, the effects of DVT on the cost-effectiveness of treatments is negligible and omitted from the economic model.

Costs

Initial procedure costs

National reference costs

National reference costs are available for varicose vein procedures, but they do not differentiate between different treatment methods. The latest available NHS activity data (2010/11) in Table 1 show that surgery is still the most common procedure (52%), although the others (EVLA, RFA and sclerotherapy) are also commonly provided. Of the total 36,000 varicose vein procedures captured in the 2009/10 national reference costs, almost 80% were conducted as day cases, the majority of which were a primary unilateral procedure without complications (84%). 52 The average cost of such procedures, inflated to 2011/12 with an inflation factor of 1.064,144 is £1155 (IQR £765–1355). 52

The literature

A total of 15 studies2,45,81,90,95,122,126,134,135,145–149 reporting costs of treatments for varicose veins were identified from the economic searches, with one further study identified in the evidence review. Several of the studies, including one from the UK (Gohel et al. 136), estimated the costs of treatment from standard charges/reimbursement costs with additional costs added for equipment and consumables particular to each treatment mode. These studies were not considered useful in informing the costs of treatment. All the 16 studies2,45,81,90,95,122,126,134–136,145–149 identified are listed in Appendix 11, together with a brief comment on the costing approach.

Two studies are primary studies which collected resource and cost data alongside RCTs in the UK (Bountouroglou et al. 90 and Subramonia and Lees134). The costs they report are presented below, together with those reported by Michaels et al. 2,150–152 from a RCT comparing traditional treatments for varicose veins in the UK. Patient characteristics and summary treatment details are show in Table 25. Note that Table 25 includes one further UK cost study which was identified in a later search (see below).

TABLE 26 - Net benefits of treatment at a threshold of £20,000 (derived from Gohel et al.136)

Treatment	Net benefits at £20,000 (£)
Conservative treatment	70,440
UGFS (LA)	76,291
EVLA (LA)	77,769
RFA (LA)	77,770
Surgery (day case)	77,778
EVLA (GA)	77,165
RFA (GA)	77,196
Surgery (inpatient)	77,020

The studies are quite heterogeneous in terms of varicose vein inclusion criteria, and procedures may have varied, but all were unilateral treatments, with the possible exception of Subramonia and Lees,89 where it is not stated. In all studies stripping was done under GA as a day case, but some of the other procedures (FS, EVLA) were done either under LA or the type of anaesthesia was not reported.

Table 27 shows the costs of varicose vein treatments reported in RCTs that conducted primary collection of resource and cost data in the UK. All costs have been inflated to 2011/12 prices. 144 The Michaels et al. study2 had much longer follow-up (2 years) than the other two studies (< 3 months) and the total health-care costs include retreatment. The total initial hospital costs of treatment are more similar. Each of the three studies compares a different treatment (FS, RFA and LS) with stripping.

TABLE 27 - Costs (£) of varicose vein procedures, UK primary studies, £2011/12

GP, general practitioner.

Item		Bountouroglou 200690		Subramonia 2010134		Michaels 20062		Lattimer 2012153
		FS	Surgery	RFA	Surgery	LS	Surgery	FS	EVLA
Number of patients		30	30	47	41	41	36	54	56
Surgeon		52.06	118.99	102.32	68.16
Assistant		21.80	49.83	49.62	33.06
Anaesthetist		0.00	118.99	102.32	68.16
Anaesthetic assistant		0.00	25.42	36.86	24.56
Nursing		30.26	69.16	66.24	44.14
Portering				5.90	5.90
Subtotal staff time		104.11	382.39	363.25	243.98
Consumables	Stripper (surgery)/sodium tetradecyl sulphate (foam)/catheter (RFA)	5.96	10.87	652.45	0.00
	Antiembolism stockings	22.25	8.90
	Unspecified			59.31	59.31
	Sterile supplies	102.41	102.41
	Anaesthetic	3.61	10.63
Subtotal consumables		134.23	132.81	711.77	59.31
Theatre recovery		50.85	105.93
Ward time		328.40	625.03
Day-case ward				250.07	250.07
Subtotal facilities		379.25	730.96	250.07	250.07
Ultrasound		63.56	0.00	43.89	0.00
Medical attendance/non-protocol outpatient visits		10.11	0.00	3.89	8.91
Subtotal miscellaneous		73.67	0.00	47.78	8.91
Capital and overhead		91.15	178.44	140.70	101.00
Total initial treatment hospital costs		782.41	1424.60	1513.58	663.27	251.14	895.75
Retreatment						52.84	27.12
Hospital admissions/visits						0.00	12.56
GP/practice nurse/district nurse visits				11.31	23.87	2.09	2.38
Total health-care costs				1524.88	687.14	306.06	937.80	230.24	724.72

It can be seen that there is considerable variation between the three studies2,89,90 in the reported costs of stripping, varying from £663 to £1425. In fact, only that of Michaels et al. 2 falls within the IQR of the national reference costs (£765–1355). 52 The study characteristics shown in Table 25 do not explain these variations. All the studies compared different treatments with stripping, so their costs relative to stripping can be calculated. Table 28 shows the estimated costs of varicose vein procedures, assuming the cost of stripping is the national reference cost for primary unilateral day-case procedures,52 and the relative costs for other procedures based on the data from the RCTs shown in Table 27. The validity of the assumption that the national reference cost for all procedures was reasonably representative of the costs of stripping, although accounting for only 52% of activity, was explored using the limited HES activity data (see Table 1) and the cost ratios shown in Table 28. The limitation of the HES data is that EVLA and RFA activity are combined, as is that for FS and LS. Assuming an equal split in activity between treatments for both these groups gives an estimate for the cost of stripping of £1076. The effect on the differences in costs between the newer treatments and stripping (FS +£35, RFA –£100, EVLA –£80) is insignificant in the context of their magnitude and the lack of effect on the results (see Costs).

TABLE 28 - Initial procedure costs estimated relative to stripping costs

Estimated – see text immediately below.

Procedure	Cost relative to surgery	Initial procedure cost (£)	Source
Surgery	–	1155	National reference costs52
FS	0.55	634	Bountouroglou 200690
RFA	2.28	2635	Subramonia 2010134
EVLA	2.02a	2338	See text below

None of the UK RCTs, or any detailed costing studies from other countries, included EVLA. One study shows the procedure times for EVLA and RFA to be the same,88 and a survey of expert opinion also indicates that the duration of procedure and facilities required are very similar to that for RFA, the key difference being the specialist catheter kits and generators that are used for the two procedures. 154 The differential between the costs for these two items was used to estimate the cost difference between EVLA and RFA. Prices for catheter kits vary between manufacturers. Most are in the range £200–300 for optical fibre kits (including access kits – EVLA) and £460–578 for RFA catheter and access kits (2009). 154 In the clinical trials, Biolitec laser kits (Biolitec AG, CeramOptec GmbH, Bonn, Germany) were commonly used for EVLA (£255) and VNUS Closure catheters for RFA (£578), including Subramonia and Lees;134 these were the costs used in the model. Subramonia and Lees134 do not explicitly cost the RFA generator, but rather use a standard formula for calculation of overhead costs, including capital charges, of 30% direct theatre costs. 134 However, as a laser generator costs approximately twice as much as a RFA generator (£20,000 compared with £10,000),154 an allowance needs to be made for the additional capital cost when comparing the costs of the two procedures. Assuming capital costs are depreciated over 5 years, an interest rate of 3.5%, and 50 patients are treated per year, the additional capital cost for EVLA compared with RFA is £44. This cost is clearly sensitive to the assumed number of patients treated per year. The activity data in the national reference costs show that the average number of day-case unilateral procedures per unit is 151,52 so, if EVLA were to be widely adopted, 50 patients per year seems a reasonable assumption. This gives an overall cost difference between RFA and EVLA of £279 (2009/10 prices) or £297 at 2011/12 prices.

Given the limited information on the costs of EVLA in particular, the costs literature search was rerun in August 2012 to identify any additional literature. One further UK RCT was identified that collected and reported cost data for EVLA and FS. 153 These were £725 and £230 respectively (cost year unspecified, 2012?). These costs are considerably lower than those reported in the other studies (see Table 27), and also inconsistent with the mean national reference cost for varicose vein day procedures presented earlier. However, the ratio of costs of FS to EVLA at 0.31 is similar to that derived from Table 28 (0.55/2.02 = 0.27).

There are limited data on the uncertainty in procedure costs. Only one study (Michaels et al. 2) reports the SD with the mean costs. These data were used to calculate the SE of the mean of the costs for LS and stripping. Assuming the SE of the distributions of mean costs are described by gamma distributions, the mean cost distributions for LS and stripping were each sampled 1000 times to give a distribution of the cost ratio for LS compared with stripping. This had a SE of 11% of the mean ratio value, and was normally distributed. It was assumed that this distribution applied to the cost ratios of other treatments compared with stripping.

Additional treatment to achieve a successful outcome

It has been assumed that patients whose treatment is an initial failure (by 1 month – determined from the time to failure distribution) are retreated with FS at least once. Sclerotherapy (FS or LS) was the most commonly reported method of short-term retreatment for incomplete occlusion in the effectiveness studies (Table 29). Two studies also report additional treatment81,103 not for unsuccessful occlusion, but for ‘residual side branches and accessory saphenous veins’. 103 This form of additional treatment is referred to here as top-up treatments. Very few studies report on further treatment within approximately 1 month of the initial procedure. All studies reporting further treatment within this time scale, whether for incomplete occlusion (retreatment) or not (top-up treatment) are shown in Table 29. It has been assumed that all of these studies are informative for top-up treatments, with zero being assumed for those cohorts for whom all retreatment was for incomplete occlusion. Note that this table differs from Table 11 in that it only includes studies that report retreatment, and not just treatment failure. In two studies (Bounteroglou et al. 90 and ElKaffas et al. 81) all treatment failures were retreated, so the numbers match those shown in Table 11, whereas in Rasmussen et al. 95 not all failures were retreated, and so the numbers are not the same. Furthermore, Table 29 includes Disselhoff et al. ,126 although the follow-up period is 6 weeks rather than 1 month (the inclusion criteria for Table 11) because of the paucity of data on top-up treatment.

TABLE 29 - Study data on retreatment and top-up treatments

SF, saphenofemoral; SJL, saphenofemoral junction ligation.

Study	Initial procedure	Time success defined	Retreatment for incomplete occlusion	n	Retreated	Top-up treatment	Top-up treatment (%)
Bountouroglou 200690	FS (+ SJL)	3 weeks	Additional foam	30	4	0	0.0
	Stripping	3 weeks	Additional liquid	28	2	0	0.0
Disselhoff 2008103	EVLA	6 weeks	LS (n = 36) or phlebectomy (n = 1)	56		37	66.1
	Cryostripping	6 weeks	LS (n = 33) or phlebectomy (n = 2)	55		35	63.6
ElKaffas 201181	RFA	Immediate	GSV stripping with SF ligation	90	6
			Post-intervention sclerotherapy	90		24	26.7
	Stripping	Immediate		90	0	0	0.0
Rasmussen 201195	FS	1 month	Foam sessions	144	5	0	0.0
	EVLA	1 month	Foam sessions	144	0	0	0.0
	Stripping	1 month	Foam sessions	144	0	0	0.0
Combined results top-up treatment	Stripping			262		0	0.0
	FS			174		0	0.0
	EVLA			200		37	18.5
	RFA			90		24	26.7
	EVLA/RFA			290		61	21.0

The numbers for each treatment procedure were combined, but the variation between studies in top-up treatment rates suggests differences in populations/practice. However, the derived rates do, overall, reflect expected differences, although there is clearly uncertainty in the rates. Clinical experience suggests that it is very rare for any sort of top-up treatment to be required after stripping. This is very different from the case with EVLA and RFA. If these are carried out under LA, which is the usual situation, it is almost impossible to get full clearance of all the varicose veins at the time of the initial procedure. The low top-up treatment rates in some studies may reflect a policy decision not to retreat some residual varicosities. A combined rate for EVLA and RFA was calculated as in principle the top-up treatment rate might be expected to be similar.

From Table 29 it can be seen that sclerotherapy is the most common mode of immediate retreatment and top-up treatment. It is assumed in the model that all immediate retreatment and top-up treatments are FS. Ongoing time to failure following initial failure or top-up treatment is determined in the model by the initial treatment mode.

Taking in to account top-up procedures, the total initial procedure costs used in the model are shown in Table 30.

TABLE 30 - Total initial procedure costs (including top-up treatments)

Procedure	Total cost (£)
Stripping	1154.91
FS	634.29
EVLA	2471.54
RFA	2768.91

Costs of retreatment

It is assumed the costs of retreatment itself are the same as the cost of treatment, but clinical opinion suggests that patients will also see their general practitioner (GP) on average 2.5 times, and attend two outpatient visits, including a duplex scan. The costs are shown in Table 31.

TABLE 31 - Retreatment resource use and costs

PSSRU, Personal Social Services Research Unit.

Item	Cost/number	Source
GP visit cost (£)	32.10	PSSRU144
Outpatient first attendance vascular surgery cost (£)	172.34	National reference costs52
Outpatient second (or further) attendance vascular surgery cost (£)	118.13	National reference costs52
Duplex scan cost (£)	59.04	National reference costs52
GP visits, n	2.5	Author estimate
Outpatient visits, n	2	Author estimate
Duplex scans, n	1	Author estimate
Total additional retreatment cost (£)	429.76

Retreatment following failure

Assumptions need to be made in the model regarding the mode of retreatment following late treatment failure. Data from the trials included in the effectiveness review were extracted on retreatment mode for each initial treatment, but shows no consensus for retreatment modes, although FS was the most commonly used method (see Appendix 6). In many trials only one mode of retreatment was reported, suggesting that the choice was procedure driven rather than patient driven.

In practice, patients have a scan, the results of which determine further treatments. In terms of conventional methods, the general principle is that if there is recurrent reflux at the saphenofemoral junction (SFJ) then stripping is required, whereas for those who have reflux in the long saphenous vein without saphenofemoral incompetence or have some recurrent varicosities it is more likely that sclerotherapy, either LS or FS, would be used. The newer modalities of EVLA and RFA are possible as repeat procedures only where there is an incompetent long saphenous vein that can be treated in this way. None of the effectiveness studies reports the use of RFA for secondary procedures, but two do report the use of EVLA following stripping. 87,96 It has been assumed that 60% of retreatments are surgical procedures and 40% are FS.

Utility values

The model requires (1) the utility associated with symptomatic varicose veins; and (2) the short-term post-operative loss of utility reflecting adverse effects of each procedure. To inform these parameters a search was undertaken to identify utility values for varicose veins in the literature (see Identification of studies). This identified 975 unique references. The literature identified as being relevant to these two sets of parameters is discussed in turn.

Results

The results of the PSA are presented first as they are considered the most reliable as they take into account the distribution of the uncertainty in the model parameters, which is important particularly for the survival distributions, which are skew. The deterministic results are presented later for the purpose of univariate sensitivity analysis only. The results of the PSA with costs and QALYs discounted at a rate of 3.5% are shown in Table 34.

Although there is an element of retreatment, the total costs of treatment primarily comprise the initial treatment cost, and are similarly ordered, with RFA the most expensive procedure and foam the least costly option. All the novel treatments result in more QALYs than stripping at 10 years, but the QALY differences between stripping, EVLA and RFA are negligible: equivalent to less than a day in full health for EVLA compared with stripping.

Foam sclerotherapy is less costly than stripping and marginally more effective, and can thus be said to dominate stripping. The ICERs for EVLA and RFA in comparison with stripping show they are not cost-effective at usually accepted levels. 159 EVLA dominates RFA, as can be seen from a plot of the mean costs and QALYs of each treatment (Figure 10).

FIGURE 10.

Mean costs and QALYs at 10 years.

Given the negligible differences in QALYs between the treatments, the incremental net benefits primarily reflect the total cost differences between them (Table 35). At thresholds between £20,000 and £50,000 foam is the most cost-effective treatment, with a small probability of error.

Discussion

The analysis shows that any differences in benefits (QALYs) between the different procedures are negligible, but marginally favour the novel treatments relative to stripping. The time to treatment failure curves sourced from the mixed-treatment comparison described in Chapter 3, Quantitative analysis are all very similar. Disutility associated with post-operative pain, although not severe and limited to a few days' duration, affects the results in the short term (2 years), demonstrating the limited effects of time to failure on differential QALYs. There are differences in treatment costs and, although these are somewhat uncertain, with little differences in QALYs the incremental net benefits are primarily driven by costs. Our central estimate is that FS costs £530 less than stripping, and is marginally more effective, with a probability of being the most cost-effective treatment above 90% for willingness-to-pay thresholds in the range £20,000–50,000. This result is sensitive to the model time horizon. With FS having a higher failure rate (initially) than stripping, a shorter time horizon may result in fewer QALYs than stripping. With a short model time horizon (2 years) the result is also sensitive to the disutility associated with treatment. This parameter was derived from the mixed-treatment comparison of reported pain at approximately 10 days (see Table 22). By 10 days post-operative pain has already subsided, and therefore the analysis may not fully reflect differences between the treatments. Also the relationship between post-operative pain and utility was based on limited data. 86 However, the best possible estimate of these parameters has been used given the available data.

Endovenous laser ablation and RFA both cost more than surgery, and with very little difference in QALYs they cannot be considered cost-effective at the usual threshold of £20,000–30,000,159 a result that is robust to parameter variation and model time horizon. If their costs approach that for surgery they would be considered cost-effective.

There is uncertainty in the cost differentials between treatments, and in fact these are likely to vary with setting, and may vary over time. However, the differences in clinical effectiveness (time to recurrence, post-operative pain) are small. Threshold analysis shows that the additional costs of EVLA and RFA would have to be no more than £50 and £24, respectively, to be considered cost-effective at a threshold of £20,000.

In the model QALYs are determined by initial disutility associated with treatment and treatment failure. Treatment failure is not assumed to be necessarily symptomatic, but for those who are they are assumed to have a lower utility for a period of 6 months prior to retreatment. There is also a further disutility associated with the retreatment procedure, as for the initial procedure. Asymptomatic and symptomatic utility is assumed to be the same following all treatments. There is no direct evidence of relevant differences in utility following treatment. Nordon et al. 88 report the same increase in utility from baseline for EVLA and RFA at 3 months, and Carradice et al. 86 report the same utility for EVLA and surgery patients preoperatively, at 1 week and 1 year following treatment.

However, the mixed-treatment comparison of VCSS scores at approximately 1 year indicates slightly lower VCSS scores (i.e. less severe symptoms) in FS patients than in patients treated with stripping, despite a higher failure rate at this time (see Table 21). Studies reporting analyses of the relationship between disease-specific quality of life in varicose veins and generic measures have mixed results. Shepherd et al. 139 and Kahn162 found poor correlation between measures [AVVQ, Specific Quality-of-life and Outcome Response–Venous (SQOR-V) and SF-12 for the former, Venous Insufficiency Epidemiological and Economic Study (VEINES)-QOL (quality of life), VEINES-sym (symptoms) and SF-12 for the latter]. One study does report a significant relationship between VCSS and EQ-5D in a multivariate model which also included CEAP score, AVVQ and population characteristics. For a unit decrement in VCSS there is an increase in EQ-5D of 0.02. 163 If this relationship holds, the QALY difference between FS and stripping may be greater than shown by the model results, but does not change the conclusion from the baseline results as FS dominates stripping. It potentially does affect the QALY difference with surgery with shorter model time scales, and would mean that FS results in more QALYs than surgery irrespective of the time scale. The differences in the VCSS scores between EVLA and RFA with stripping were much smaller than for FS, and their inclusion does not change the cost-effectiveness of these treatments.

The model results are consistent with other studies in finding that QALY differences between treatments are very small. Only one relevant study was identified that reported incremental QALYs from trial data. 133 It found a difference of 0.01 QALYs between EVLA and cryostripping 2 years after treatment. In the modelling analysis by Gohel et al. 136 the maximum relevant difference in QALYs at 5 years was 0.115 between day-case surgery and FS. 136 However, in other respects the results of this model are different from those of Gohel et al. 136 Gohel et al. 136 estimated the costs of treatments from basic units of resource (day case, outpatient, equipment costs) and reports day-case surgery to be more costly than any of the novel treatments, contrary to more recently published cost studies showing the costs of EVLA and RFA to be greater than those for surgery. 90,134 Gohel et al. 136 also find surgery to be more effective than the novel treatments, on the basis of much more limited effectiveness data than used in the current analysis.

The strength of the model described in this report is that the treatment failure rates are derived from a comprehensive systematic review and meta-analysis using mixed-treatment comparison. Furthermore, the analysis of effectiveness allowed the shapes of the time to failure curves to vary, thereby avoiding any assumption about indefinite treatment effects. Nevertheless, the limitations of the failure data previously described are recognised, as well as the limited evidence of differences between treatments in post-procedure utility.

The economic analysis was undertaken from a UK NHS perspective. The effectiveness review included time to return to work or usual activities, which showed this to be quicker for FS and RFA than for stripping, and possibly also quicker for EVLA. This means that from both an NHS and a societal perspective FS is the most cost-effective option.

Search strategy

1. Varicose Veins/

2. varicose vein.tw.

3. varicose veins.tw.

4. vein, varicose.tw.

5. veins, varicose.tw.

6. varices.tw.

7. varix.tw.

8. varicosis.tw.

9. varicosit$.tw.

10. Saphenous Vein/

11. (saphenous adj2 vein$).tw.

12. (saphena adj2 vein$).tw.

13. or/1-12

14. laser ablation.tw.

15. evla.tw.

16. radiofrequency ablation.tw.

17. radio frequency ablation.tw.

18. radio-frequency ablation.tw.

19. rfa.tw.

20. vnus.tw.

21. closurefast.tw.

22. rfitt.tw.

23. olympus.tw.

24. foam sclero$.tw.

25. Sclerotherapy/

26. sclerotherap$.tw.

27. 25 or 26

28. foam.tw.

29. 27 and 28

30. ugfs.tw.

31. transilluminated phlebectomy.tw.

32. tipp.tw.

33. radiofrequency obliteration.tw.

34. radio frequency obliteration.tw.

35. radio-frequency obliteration.tw.

36. rfo.tw.

37. or/14-24,29-36

38. 13 and 37

Search strategy

1. Varicose Veins/

2. varicose vein.tw.

3. varicose veins.tw.

4. vein, varicose.tw.

5. veins, varicose.tw.

6. varices.tw.

7. varix.tw.

8. varicosis.tw.

9. varicosit$.tw.

10. Saphenous Vein/

11. (saphenous adj2 vein$).tw.

12. (saphena adj2 vein$).tw.

13. or/1-12

14. laser ablation.tw.

15. evla.tw.

16. radiofrequency ablation.tw.

17. radio frequency ablation.tw.

18. radio-frequency ablation.tw.

19. rfa.tw.

20. vnus.tw.

21. closurefast.tw.

22. rfitt.tw.

23. olympus.tw.

24. foam sclero$.tw.

25. Sclerotherapy/

26. sclerotherap$.tw.

27. 25 or 26

28. foam.tw.

29. 27 and 28

30. ugfs.tw.

31. transilluminated phlebectomy.tw.

32. tipp.tw.

33. radiofrequency obliteration.tw.

34. radio frequency obliteration.tw.

35. radio-frequency obliteration.tw.

36. rfo.tw.

37. or/14-24,29-36

38. 13 and 37

39. Economics/

40. "costs and cost analysis"/

41. Cost allocation/

42. Cost-benefit analysis/

43. Cost control/

44. cost savings/

45. Cost of illness/

46. Cost sharing/

47. "deductibles and coinsurance"/

48. Health care costs/

49. Direct service costs/

50. Drug costs/

51. Employer health costs/

52. Hospital costs/

53. Health expenditures/

54. Capital expenditures/

55. Value of life/

56. exp economics, hospital/

57. exp economics, medical/

58. Economics, nursing/

59. Economics, pharmaceutical/

60. exp "fees and charges"/

61. exp budgets/

62. low adj cost).mp.

63. (high adj cost).mp.

64. (health?care adj cost$).mp.

65. (fiscal or funding or financial or finance).tw.

66. (cost adj estimate$).mp.

67. (cost adj variable).mp.

68. (unit adj cost$).mp.

69. (economic$ or pharmacoeconomic$ or price$ or pricing).tw.

70. or/39-69

71. 38 and 70

Search strategy

1. Varicose Veins/

2. varicose vein.tw.

3. varicose veins.tw.

4. vein, varicose.tw.

5. veins, varicose.tw.

6. varices.tw.

7. varix.tw.

8. varicosis.tw.

9. varicosit$.tw.

10. Saphenous Vein/

11. (saphenous adj2 vein$).tw.

12. (saphena adj2 vein$).tw.

13. or/1-12

14. laser ablation.tw.

15. evla.tw.

16. radiofrequency ablation.tw.

17. radio frequency ablation.tw.

18. radio-frequency ablation.tw.

19. rfa.tw.

20. vnus.tw.

21. closurefast.tw.

22. rfitt.tw.

23. olympus.tw.

24. foam sclero$.tw.

25. Sclerotherapy/

26. sclerotherap$.tw.

27. 25 or 26

28. foam.tw.

29. 27 and 28

30. ugfs.tw.

31. transilluminated phlebectomy.tw.

32. tipp.tw.

33. radiofrequency obliteration.tw.

34. radio frequency obliteration.tw.

35. radio-frequency obliteration.tw.

36. rfo.tw.

37. or/14-24,29-36

38. 13 and 37

39. Economics/

40. "costs and cost analysis"/

41. Cost allocation/

42. Cost-benefit analysis/

43. Cost control/

44. cost savings/

45. Cost of illness/

46. Cost sharing/

47. "deductibles and coinsurance"/

48. Health care costs/

49. Direct service costs/

50. Drug costs/

51. Employer health costs/

52. Hospital costs/

53. Health expenditures/

54. Capital expenditures/

55. Value of life/

56. exp economics, hospital/

57. exp economics, medical/

58. Economics, nursing/

59. Economics, pharmaceutical/

60. exp "fees and charges"/

61. exp budgets/

62. (low adj cost).mp.

63. (high adj cost).mp.

64. (health?care adj cost$).mp.

65. (fiscal or funding or financial or finance).tw.

66. (cost adj estimate$).mp.

67. (cost adj variable).mp.

68. (unit adj cost$).mp.

69. (economic$ or pharmacoeconomic$ or price$ or pricing).tw.

70. or/39-69

71. 38 and 70

72. 13 and 70

73. 72 not 71

Search strategy

1. Varicose Veins/

2. varicose vein.tw.

3. varicose veins.tw.

4. vein, varicose.tw.

5. veins, varicose.tw.

6. varices.tw.

7. varix.tw.

8. varicosis.tw.

9. varicosit$.tw.

10. Saphenous Vein/

11. (saphenous adj2 vein$).tw.

12. (saphena adj2 vein$).tw.

13. or/1-12

14. laser ablation.tw.

15. evla.tw.

16. radiofrequency ablation.tw.

17. radio frequency ablation.tw.

18. radio-frequency ablation.tw.

19. rfa.tw.

20. vnus.tw.

21. closurefast.tw.

22. rfitt.tw.

23. olympus.tw.

24. foam sclero$.tw.

25. Sclerotherapy/

26. sclerotherap$.tw.

27. 25 or 26

28. foam.tw.

29. 27 and 28

30. ugfs.tw.

31. transilluminated phlebectomy.tw.

32. tipp.tw.

33. radiofrequency obliteration.tw.

34. radio frequency obliteration.tw.

35. radio-frequency obliteration.tw.

36. rfo.tw.

37. or/14-24,29-36

38. 13 and 37

39. "Quality of Life"/

40. "Value of Life"/

41. quality of life.tw.

42. Health Status Indicators/

43. health status indicator$.tw.

44. Health Status/

45. health status profile$.tw.

46. health related quality of life.tw.

47. Quality-Adjusted Life Years/

48. quality adjusted life.tw.

49. (qaly$ or qald$ or qale$ or qtime$).tw.

50. disability adjusted life.tw.

51. daly$.tw.

52. (sf36 or sf 36 or short form 36 or shortform 36 or sf thirtysix or sf thirty six or shortform thirtysix or shortform thirty six or short form thirtysix or short form thirty six).tw.

53. (sf6 or sf 6 or short form 6 or shortform 6 or sf six or sfsix or shortform six or short form six).tw.

54. (sf12 or sf 12 or short form 12 or shortform 12 or sf twelve or sftwelve or shortform twelve or short form twelve).tw.

55. (sf6D or sf 6D or short form 6D or shortform 6D or sf six D or sfsixD or shortform six D or short form six D).tw.

56. (sf20 or sf 20 or short form 20 or shortform 20 or sf twenty or sftwenty or shortform twenty or short form twenty).tw.

57. (sf8 or sf 8 or short form 8 or shortform 8 or sf eight or sfeight or shortform eight or short form eight).tw.

58. (euroqol or euro qol or eq5d or eq 5d).tw.

59. (hql or hqol or h qol or hrqol or hr qol).tw.

60. health related quality of life instrument.tw.

61. (aqol or a qol).tw.

62. assessment of quality of life instrument.tw.

63. (hye or hyes).tw.

64. health$ year$ equivalent$.tw.

65. health utilit$.tw.

66. utilit$.tw.

67. (hui or hui1 or hui2 or hui3).tw.

68. disutili$.tw.

69. rosser.tw.

70. quality of wellbeing.tw.

71. quality of well-being.tw.

72. qwb.tw.

73. willingness to pay.tw.

74. standard gamble$.tw.

75. time trade off.tw.

76. time tradeoff.tw.

77. tto.tw.

78. health impact survey$.tw.

79. or/39-78

80. 13 and 79