Enhancements to angioplasty for peripheral arterial occlusive disease (PAD): systematic review, cost-effectiveness assessment and expected value of information analysis

Population

The population was participants with symptomatic PAD undergoing endovascular treatment for disease distal to the inguinal ligament. Patients with either IC or CLI were included.

Interventions

Interventions were techniques used as an adjunct to, or as a replacement for, balloon angioplasty in the peripheral circulation.

These were as follows: absorbable stents, self-expanding stents (SESs), balloon-expandable stents (BESs), DESs, stent-graft, atherectomy, CB, cryoplasty, radiation by EVBT or EVRT, DCB and laser angioplasty.

Comparator

The comparator was conventional PTA. Bail-out stenting was included as a possible comparator for any of the interventions, BMSs were considered as a comparator for DESs, and sham radiation was included as a possible comparator for radiation interventions.

Outcomes

Reported outcomes included patency or restenosis measures, need for reintervention, disease-specific and generic measures of QoL, clinical status, exercise tolerance or walking distance, pain (patient-reported pain scores and analgesic use), limb salvage, complications and adverse events. Cost outcomes are discussed in Chapter 4.

Study design

Initially, RCTs were searched. As data were available from these, other study types from further down the accepted hierarchy of evidence were not sought. Meta-analyses and systematic reviews of RCTs were sought to identify RCTs that met the inclusion criteria of this review.

Interventions

Pharmacological interventions, combined surgical procedures and devices that have been withdrawn, such as older laser angioplasty devices, were not considered, as well as interventions above the inguinal ligament (aortoiliac segment).

Publication types

Studies that were published only in languages other than English, studies based on animal models, and preclinical and biological studies were excluded, as were narrative reviews, editorials and opinion pieces. Reports published as meeting abstracts were excluded only when insufficient details were reported to allow inclusion.

Study selection was made by one reviewer and checked by another, based on the above inclusion and exclusion criteria. Citations were sifted by title and abstract, and those remaining after abstract sift were sifted by full papers. Studies excluded at full-paper screening were placed in Appendix 2.

Data extraction, critical appraisal and synthesis

Data were extracted by one reviewer and checked by another, using a standardised form. The forms are shown in Appendix 3. Data were extracted with no blinding to authors or journal. Quality was assessed according to criteria based on NHS Centre for Reviews and Dissemination (CRD) Report No. 4. 9 Quality assessment forms are shown in Appendix 4. Prespecified outcomes were tabulated and discussed within a descriptive synthesis. For some interventions, meta-analyses were precluded as a result of differences in outcomes. For example, definitions of patency varied across trials and there were also differences in populations, interventions and length of follow-up. When appropriate, meta-analyses were undertaken using fixed- and random-effects methods. Meta-analyses were carried out using Review Manager 5.1 (The Nordic Cochrane Centre, Copenhagen, Denmark). The Mantel–Haenszel methods have been shown to be more reliable than other methods when there are relatively few studies with small sample sizes, so these were employed, with both fixed and random effects, as recommended by the Cochrane Collaboration. 10

Study selection

The search of electronic databases yielded 9501 article citations with duplicates removed. Additional searching yielded one reference and two conference presentations. The sifting process is shown in Figure 1, a flow diagram adapted from PRISMA recommendations. 8 Title sifting excluded 8175 citations. There were 1329 abstracts sifted. In total, 95 references were full-text screened. Appendix 2 shows 34 studies that were excluded at the full-paper sifting stage with reasons for exclusions.

FIGURE 1.

Flow diagram of study selection (based on a revised version of the PRISMA diagram). 8

There were 40 RCTs accepted into the review, published in 61 references, comprising 53 articles from peer-reviewed journals with additional data in eight conference presentations (Table 1). Following literature searches, the Zilver PTX25–27 trial published an additional paper,71 which confirmed the results included from abstracts.

There was one RCT of absorbable metal stents (AMSs), five RCTs of SESs and six RCTs of BESs. There were three trials of DESs, of which one concerned paclitaxel and two sirolimus. There was one trial of stent-graft, two of atherectomy, two of CB and two of cryoplasty. Of the 10 RCTs of radiation, eight employed EVBT and two employed EBRT. Three RCTs of DCB were included and five RCTs of laser angioplasty. Trials of stents, stent-graft, CB, cryoplasty and DCB versus PTA allowed bail-out stenting in the PTA group, when deemed medically necessary. Bail-out atherectomy was permitted in one atherectomy trial (Vroegindeweij et al. 36), and, of the radiation trials, the comparator PTA group had oral placebo in two RCTs (Gallino et al. , 42 Zehnder et al. 45) and sham radiation in four RCTs (Krueger et al. ,47,48 Vienna-3,51–53 Fritz et al. ,57 Therasse et al. 58).

Further study details are shown in Appendix 3.

Critical appraisal

Appendix 4 shows the quality assessment for the included studies. Method of allocation concealment was considered adequate in 11 of the trials [AMS INSIGHT (bio-absorbable metal stent investigation in chronic limb ischaemia treatment),11 Becquemin et al. ,19 Grimm et al. ,21 Rand et al. ,22 Vroegindeweij et al. ,23,35,36 Rastan et al. ,31 Tielbeck et al. ,36 Amighi et al. ,38 Dick et al. ,39 VARA (VAscular RAdiotherapy trial),54 FemPac (Femoral Paclitaxel trial)64]. Both the method used to generate allocation sequences and the method of allocation concealment were considered adequate in seven of these trials (AMS INSIGHT,18 Becquemin et al. ,19 Grimm et al. ,21 Rastan et al. ,31 Amighi et al. ,38 Dick et al. ,39 VARA,54 FemPac64). For other trials, reporting of randomisation methods was unclear.

There was blinding for assessors in at least one of the study outcomes in 20 trials [Dick et al. ,12 FAST (Femoral Artery Stenting Trial),14 ABSOLUTE (randomized balloon angioplasty vs. stenting with nitinol stents in the superficial femoral artery),16–18 Becquemin et al. ,19 Rand et al. ,22 SIROCCO (SIROlimus-Coated COrdis self-expandable stent trial),28–30 Rastan et al. ,31 Amighi et al. ,38 Spiliopoulous et al. ,41 Diehm et al. 44 analysis of Gallino et al. 42 and Zehnder et al. 45 trials, Krueger et al. ,47,48 Vienna-3,51–53 Wyttenbach et al. ,55,56 Fritz et al. ,57 Therasse et al. ,58 THUNDER (local taxane with short exposure for reduction of restenosis in distal arteries),61–63 FemPac,64 Lammer et al. 68]. Blinding of clinicians to the endovascular techniques used in these studies would have been difficult or impossible. One trial (FemPac64) mentioned that the blinding of clinicians was attempted, but the difference in appearance of DCB and uncoated balloons meant that clinicians were likely to know which intervention was being used. There was explicit blinding of patients in eight trials (SIROCCO,28–30 Rastan et al. ,31 Krueger et al. ,47,48 Vienna-3,51–53 Fritz et al. ,57 Therasse et al. ,58 THUNDER,61–63 FemPac64).

Intervention and control groups were largely comparable at baseline in all trials. Some trials reported one variable that was not equal across treatment groups at baseline [AMS INSIGHT,11 Dick et al. ,12 RESILIENT (randomised study comparing the Edwards self-expanding LifeStent with angioplasty alone in lesions involving the superficial femoral artery and/or proximal popliteal artery),15 Zilver PTX,25–27 SIROCCO,28–30 Rastan et al. ,31 Nakamura et al. ,34 Vroegindeweij’s group,35–37 THUNDER,60–62 Fisher et al. 67]. When studies measured more outcomes than they reported, this was because of future expected reports [LEVANT I (the Lutonix paclitaxel-coated balloon for the prevention of femoropopliteal restenosis trial),59,60 Spies et al. 69].

Only one trial had an imbalance in dropouts between treatment groups (Hagenaars et al. 46). An analysis of patients in their allocated groups according to the intention-to-treat (ITT) principle was available for all trials, although for two trials (Gallino et al. , 42 Zehnder et al. 45) this was only available for the combined analysis of these two trials (Diehm et al. 44).

Absorbable metal stent

One RCT identified compared AMS with PTA (AMS INSIGHT11) in CLI patients. The AMS INSIGHT11 trial provided patency data on 94 lesions at 6-month follow-up (Tables 2 and 3). AMS fared significantly worse than PTA (p = 0.013) in terms of restenosis measured by core-lab quantitative vessel analysis (QVA). A patency measure including major amputation or target lesion revascularisation (TLR) as failure showed no significant difference between treatment groups. For adverse events, a measure including major amputation or death did not find any significant difference between groups at 1-month follow-up (Table 4).

Self-expanding stent

Five RCTs compared SESs with PTA. The populations comprised mostly IC patients, but also some CLI patients.

Three RCTs (Dick et al. ,12 RESILIENT,15 ABSOLUTE16–18) showed an advantage for SES over PTA in terms of restenosis (Table 5). Of these, one study (ABSOLUTE16–18) had only a trend favouring SES at 6 months but significant results at 1 and 2 years, whereas the other studies reached and maintained significance at 3–6 months (Dick et al. 12) and 6–12 months (RESILIENT15). One RCT found no significant difference between groups at 1-year follow-up (FAST14). Meta-analysis (Figures 2–7) for restenosis at 6 months using the studies ABSOLUTE16–18 and Dick et al. 12 produced a relative risk (RR) for SES with reference to PTA of 0.49 with a 95% CI of 0.32 to 0.76 by fixed-effect analysis. By random-effect analysis, the RR was 0.50 (95% CI 0.32 to 0.77). Both analyses significantly favoured SES over PTA (p = 0.002). Restenosis at 12 months, using the studies ABSOLUTE,16–18 Dick et al. 12 and FAST14, produced a RR of 0.68 (95% CI 0.53 to 0.87) by fixed-effect analysis (p = 0.003). By random-effect analysis, the RR was 0.67 (95% CI 0.52 to 0.87), significantly favouring SES over PTA (p = 0.002).

Of the four RCTs that reported a need for reintervention, three showed no significant difference between groups [VascuCoil (intracoil femoropopliteal stent trial),13 FAST,14 ABSOLUTE16–18] (Table 6). One study found an advantage for SES over PTA, with fewer SES participants needing TLR/target vessel revascularisation (TVR) at 6–12 months following the procedure (RESILIENT15). Rutherford category was studied by two RCTs, neither of which found a significant difference between SES and PTA treatment groups (FAST,14 ABSOLUTE16–18) (Table 7).

Treadmill protocols were used by two studies (FAST,14 ABSOLUTE16–18) to assess walking capacity (Table 8) and both found a significant advantage for SES over PTA at 6–12 months. ABSOLUTE16–18 found that by 24 months the difference between treatment groups was no longer significant. Maximum walking capacity, as reported by the patients, was reported as significantly better with SES than PTA in one study (Dick et al. 12). One study (RESILIENT15) found no significant difference between groups, as measured by the walking impairment questionnaire, as both groups improved significantly from baseline. RESILIENT15 reported that the PTA group reported more claudication pain at 12 months (p = 0.009).

The two RCTs investigating QoL (RESILIENT,15 ABSOLUTE16–18) found no significant differences between treatment groups SES and PTA on measures of Short Form questionnaire-8 items (SF-8) or Short Form questionnaire-36 items (SF-36) by ITT analysis (Table 9). There were no significant differences between treatment groups SES and PTA in terms of complications, in any of the five included RCTs (Dick et al. ,12 VascuCoil,13 FAST,14 RESILIENT,15 ABSOLUTE16–18) (Table 10).

Meta-analyses

Self-expanding stent versus percutaneous transluminal balloon angioplasty

Restenosis at 6 months: using the studies ABSOLUTE16–18 and Dick et al. ,12 there was no substantial heterogeneity between studies. Fixed- and random-effect analyses gave similar results (see Figures 2 and 3).

FIGURE 2.

Forest plot of comparison: 1 SES vs. PTA, restenosis 6 months fixed two studies.

FIGURE 3.

Forest plot of comparison: 1 SES vs. PTA, restenosis 6 months random two studies.

Restenosis at 12 months: using the studies ABSOLUTE,16–18 Dick et al. 12 and FAST,14 there was no significant heterogeneity among studies. The overall effect was similar for fixed- and random-effect analyses (see Figures 4 and 5).

FIGURE 4.

Forest plot of comparison: 1 SES vs. PTA, restenosis 12 months fixed three studies.

FIGURE 5.

Forest plot of comparison: 1 SES vs. PTA, restenosis 12 months random three studies.

Restenosis at 12 months – using the studies ABSOLUTE16–18 and Dick 2009,12 which had been used for the 6-month restenosis analyses – gave non-significant heterogeneity. Overall effect was similar for fixed- and random-effect analyses (see Figures 6 and 7).

FIGURE 6.

Forest plot of comparison: 1 SES vs. PTA, restenosis 12 months fixed two studies.

FIGURE 7.

Forest plot of comparison: 1 SES vs. PTA, restenosis 12 months random two studies.

Balloon-expandable stent

Six RCTs compared BESs with PTA.

All six included RCTs reported restenosis, and four of these studies, of which two had only IC patients and two had approximately twice as many IC as CLI patients (Becquemin et al. ,19 Cejna et al. ,20 Grimm et al. ,21 Vroegindeweij et al. 23), found no significant difference between BES and PTA (Table 11). One study of CLI patients (Rand et al. 22) reported a significant advantage for BES over PTA, whereas one study of CLI patients (Zdanowski et al. 24) reported that PTA had an advantage over BES. Meta-analyses for restenosis at 6 months, using the studies of Cejna et al. 20 and Rand et al. ,22 gave a RR of 0.49 (95% CI 0.24 to 1.02) for both fixed- and random-effect analyses, with a non-significant trend favouring BES (p = 0.06) (Figures 8–11). Restenosis at 12 months, using the studies Becquemin et al. ,19 Cejna et al. ,20 Grimm et al. ,21 Vroegindeweij et al. 23 and Zdanowski et al. ,24 gave a non-significant treatment group difference by fixed-effect (RR 1.19; 95% CI 0.85 to 1.68; p = 0.31) and random-effect analyses (RR 1.18; 95% CI 0.85 to 1.66; p = 0.34). Restenosis at 24 months, using the studies of Cejna et al. 20 and Grimm et al. ,21 gave a non-significant treatment group difference by fixed-effect (RR 1.17; 95% CI 0.59 to 2.35; p = 0.65) and random-effect analyses (RR 1.16; 95% CI 0.59 to 2.29; p = 0.67).

Neither of the two studies (Grimm et al. ,21 Zdanowski et al. 24) that reported a need for reintervention found a significant difference between BES and PTA treatment groups (Table 12). One study (Grimm et al. 21) investigated walking distance, and found similar results between groups. Although the PTA group had a slightly larger increase in walking distance, no statistic for the difference between groups was reported (Table 13). All six included RCTs reported complications (Table 14), and none of the studies showed a significant difference between groups for BES and PTA.

Meta-analyses

Balloon-expandable stent versus percutaneous transluminal balloon angioplasty

Restenosis at 12 months: using the studies of Becquemin et al. ,19 Cejna et al. ,20 Grimm et al. ,21 Vroegindeweij et al. 23 and Zdanowski et al. ,24 there was no significant heterogeneity. The overall effect was similar for fixed- and random-effect analyses (see Figures 8 and 9).

FIGURE 8.

Forest plot of comparison: 2 BES vs. PTA, restenosis at 12 months fixed.

FIGURE 9.

Forest plot of comparison: 2 BES vs. PTA, restenosis at 12 months random.

Restenosis at 24 months: using the studies of Cejna et al. 20 and Grimm et al. ,21 there was no significant heterogeneity. The overall effect was similar for fixed- and random-effect analyses (see Figures 10 and 11).

FIGURE 10.

Forest plot of comparison: 2 BES vs. PTA, restenosis at 24 months fixed.

FIGURE 11.

Forest plot of comparison: 2 BES vs. PTA, restenosis at 24 months random.

Drug-eluting stent

Three RCTs of DESs were included. One RCT compared paclitaxel-eluting stents with PTA, with participants in the PTA arm having the potential to be further randomised to DES or BMS. 25 One RCT compared sirolimus-eluting stents with SESs. 30 One RCT compared sirolimus-eluting stents with stents coated with placebo. 31

The RCT of paclitaxel-eluting stents (Zilver^® PTX^®, Cook Medical, Bloomington, IN, USA) reported a significant advantage for DES over PTA for restenosis at 12 months (Table 15), and also for survival free of amputation, TLR or worsening of Rutherford category (Table 16). Of the two RCTs of sirolimus-eluting stents, one study found no treatment effect for DES and BMS for restenosis (SIROCCO28–30), and the other found a significant advantage of DES over BMS for luminal narrowing (Rastan et al. 31) (Table 17). Neither of these studies found significant differences between groups in terms of the need for reintervention (Table 18).

One study (Rastan et al. 31) found a significant advantage for SES over BMS for improving Rutherford category (Table 19), although this advantage appeared at 12 months and was not seen 6 months post intervention. The two RCTs of sirolimus-eluting stents found no significant differences between groups for adverse events (Table 20).

Stent-graft

One RCT was identified that compared stent-graft with PTA (Saxon et al. 32,33). IC and CLI patients were included, with most having IC. This RCT reported significantly superior results for stent-graft compared with PTA in terms of restenosis, after up to 24 months follow-up (Table 21). This RCT also reported significantly superior results for stent-graft compared with PTA in terms of clinical status (Tables 22 and 23). Complications were similar between treatment groups, although there was a borderline significant effect of increased rates of thigh pain for stent-graft compared with PTA (Table 24).

Atherectomy

Two RCTs comparing atherectomy with PTA in IC patients were included. One RCT (Nakamura et al. 34) found no significant difference in restenosis rates between atherectomy and PTA at 6 months (Table 25). One RCT (Vroegindeweij et al. ,35,36 Tielbeek et al. 37) found an advantage for PTA over atherectomy for restenosis at 1-year follow-up, although this no longer reached significance at 2-year follow-up.

One RCT (Vroegindeweij et al. ,35,36 Tielbeek et al. 37) found no significant difference in clinical status between atherectomy and PTA, with both groups showing improvement after 1 month, and some continuation of improvement after 12 months (Table 26). Between-group statistics were not reported for complications (Table 27), but neither study (Nakamura et al. ,34 Vroegindeweij et al. ,35,36 Tielbeek et al. 37) suggests significant differences between atherectomy and PTA.

Cutting balloon

Two RCTs were identified that compared CB with PTA, with mostly IC, but some CLI, patients. All patients in the trial of Dick et al. 39 had prior stents and the study investigated femoropopliteal in-stent restenosis, whereas the study of Amighi et al. 38 looked at short de novo superficial femoral artery lesions. One RCT (Amighi et al. 38) showed a borderline significant trend favouring PTA over CB for restenosis. The other RCT (Dick et al. 39) found no significant difference in restenosis between CB and PTA (Table 28).

One study (Dick et al. 39) showed similar rates of need for reintervention for CB and PTA groups (Table 29). One RCT (Amighi et al. 38) showed a trend favouring PTA over CB for rates of asymptomatic patients (Tables 30 and 31). Both studies (Amighi et al. 38 and Dick et al. 39) showed similar levels of complications between CB and PTA groups (Table 32).

Cryoplasty

Two RCTs were included that compared cryoplasty with PTA in IC and CLI patients. Neither RCT (Jahnke et al. ,40 Spiliopoulos et al. 41) found a significant treatment group effect between cryoplasty and PTA for restenosis (Table 33).

One study (Spiliopoulos et al. 41) found a significant advantage for PTA over cryoplasty, in terms of fewer patients needing reintervention (Table 34). One study (Jahnke et al. 40) showed similar levels of improvement in clinical status for cryoplasty and PTA (Table 35). Both studies (Jahnke et al. ,40 Spiliopoulos et al. 41) showed similar levels of complications between cryoplasty and PTA groups (Table 36).

Radiation

In this review, 10 RCTs were included that compared radiation with PTA in majority IC and CLI patients. Of these, eight employed EVBT, and two used EBRT.

Endovascular brachytherapy studies

For restenosis (Table 37), three studies (Zehnder et al. ,45 Krueger et al. ,47,48 Vienna-351–53) showed a significant advantage for EVBT over PTA, although, for one of these studies (Krueger et al. 47,48), the advantage at 6 months was not maintained at 2 years, and two studies (Gallino et al. ,42 Hagenaars et al. 46) showed a trend favouring EVBT (Gallino et al. 42 trial significance value not calculated between two arms presented here, as it was part of a four-arm trial). Two studies (Vienna-2,49–51 VARA54), and one combined analysis with long-term follow-up of two included studies (Diehm et al. 44 analysis of Gallino et al. 42 and Zehnder et al. 45 trials), found no significant difference between EVBT and PTA (Table 38). Meta-analyses of restenosis at 6 months using VARA54 and Vienna-249–51 trials (Figures 12–15) gave a RR of 0.93 (95% CI 0.62 to 1.39; p = 0.72) by fixed-effect analysis. By random-effect analysis, the RR was 1.00 (95% CI 0.70 to 1.44; p = 1.00). At 12-month follow-up, restenosis rates based on the meta-analyses of Diehm et al. ,44 VARA54 and Vienna-351–53 had a RR of 0.63 (95% CI 0.48 to 0.83) by both fixed-effect (p = 0.001) and random-effect (p = 0.0008) analyses, significantly favouring EVBT over PTA.

TABLE 37 - Endovascular brachytherapy and restenosis

a/a, as above; PSVR, peak systolic velocity ratio.

Restenosis rates calculated from reported patency.

Study	Follow-up	Definition of restenosis/patency	PTA analysed (n)	PTA patients with restenosis (%)	Radiation analysed (n)	Radiation patients with restenosis (%)	Comparative statistic
Gallino et al. 2004,42 Bonvini et al. 2003,43 Diehm et al. 200544	6 months	> 50% restenosis measured by duplex ultrasound	84	42a	81	17a
Zehnder et al. 2003,45 Diehm et al. 200546	12 months	> 50% recurrent obstruction defined by duplex ultrasound	56	42	44	23	p < 0.028
Gallino et al. 2004,42 Zehnder et al. 2003,45 Diehm et al. 200544	12 months	50% or more diameter reduction by digital subtraction angiography	75	29.3a	72	17.3a	p = 0.16
	24 months	a/a	75	36.9a	72	35.7a	p = 0.16
	36 months	a/a	75	52.9a	72	35.7a	p = 0.16
Hagenaars et al. 200246	6 months	> 50% diameter stenosis defined by angiography	16	31.3	8	0	p = 0.08
Krueger et al. 2002, 200447,48	6 months	> 50% diameter reduction within the former stenotic section defined by angiography	15	46.7	15	0	p = 0.006
	12 months	a/a	15	33.3	15	0	p = 0.042
	24 months	a/a	15	33.3	15	13.3	p = 0.39
Vienna-249–51	6 months	Angiographically verified stenosis of > 50% narrowing of the luminal diameter within the recanalised segment compared with the diameters of normal segments. In a patient who only underwent duplex ultrasound a PSVR ≥ 2.4 was used to indicate restenosis	29	69	15	73.4
	60 months	a/a	37	32.4	37	43.2
Vienna-351–53	12 months	> 50% reduction of arterial lumen determined angiographically or, when patients refused, with duplex ultrasound. PSVR > 2.4 indicated 50% restenosis	67	67.1	67	41.7	p < 0.05
VARA54	6 months	≥ 50% restenosis of the treated segment defined by duplex ultrasound	29	31	23	22	p = 0.45
	12 months	a/a	27	44	23	35	p = 0.49

TABLE 38 - Endovascular brachytherapy and late lumen loss

a/a, as above; MRI, magnetic resonance imaging; NS, non-significant; SD, standard deviation.

Study	Follow-up	Definition of late lumen loss	PTA analysed (n)	PTA lumen	Radiation analysed (n)	Radiation lumen	Comparative statistic
Hagenaars et al. 200246	6 months	Change in lumen area from immediately post procedure to 6-month follow-up (mm2)	16	mean –1.6 mm (SD 5.1)	8	mean 4.3 mm (SD 6.8)	p = 0.03
Wyttenbach et al. 2004, 200755,56	24 hours	Lumen area gain (%) from baseline detected via cross-sectional MRI	10	86%	10	67%	Reported as NS
	3 months	a/a	10	40%	10	106%	p = 0.026
	24 months	a/a	10	30%	10	82%	p = 0.047

FIGURE 12.

Forest plot of comparison: 4 EVBT vs. PTA, restenosis at 6 months fixed two studies.

FIGURE 13.

Forest plot of comparison: 4 EVBT vs. PTA, restenosis at 6 months random two studies.

FIGURE 14.

Forest plot of comparison: 4 EVBT vs. PTA, restenosis at 12 months fixed three studies.

FIGURE 15.

Forest plot of comparison: 4 EVBT vs. PTA, restenosis at 12 months random three studies.

Need for reintervention rates were not significantly different between EVBT and PTA (Gallino et al. ,42 Krueger et al. ,47,48 Vienna-2,49–51 Vienna-3,51–53 VARA,54 and the combined Gallino et al. 42/Zehnder et al. 45 analysis reported by Diehm et al. 44) (Table 39). One RCT (VARA54) and one combined analysis with long-term follow-up of two included studies (Diehm et al. 44 analysis of Gallino et al. 42 and Zehnder et al. 45 trials) found no significant difference between EVBT and PTA in terms of clinical improvement (Table 40).

TABLE 39 - Endovascular brachytherapy and need for reintervention

a/a, as above.

Study	Follow-up	Definition of reintervention	PTA analysed (n)	PTA patients undergoing reintervention (%)	Radiation analysed (n)	Radiation patients undergoing reintervention (%)
Gallino et al. 2004,42 Bonvini et al. 2003,43 Diehm et al. 200544	6 months	Revascularisation needed	75	11	69	6
Zehnder et al. 200345	12 months	Repeat dilatation or surgery	56	23.2	44	6.8
Krueger et al. 2002, 200447,48	6 months	TLR	15	0	15	0
	12 months	a/a	15	0	15	0
	24 months	a/a	15	13.3	15	6.7
	6 months	TVR	15	0	15	6.7
	12 months	a/a	15	0	15	13.3
	24 months	a/a	15	13.3	15	26.7
Vienna-249–51	60 months	TLR	51	64.7	51	62.7
	60 months	TVR	51	72.5	51	70.6
Vienna-351–53	12 months	TLR	46	30.4	50	10
	12 months	TVR	46	0	50	4
	12 months	Bypass surgery	46	0	50	2
VARA54	12 months	Mandatory TLR; PTA or bypass surgery	29	21	22	18

TABLE 40 - Endovascular brachytherapy and clinical improvement

a/a, as above.

Study	Follow-up	Definition of clinical status	PTA analysed (n)	PTA outcome (95% CI)	Radiation analysed (n)	Radiation outcome (95% CI)	Comparative statistic
Gallino et al. 2004,42 Zehnder et al. 2003,45 Diehm et al. 200544	12 months	Sustained clinical improvement was defined as survival without repeat revascularisation and with an ABPI > 0.1 and/or an upwards categorical shift in clinical symptoms according to the Rutherford classification	75	84.3 (72.7 to 91.3)	72	82.4 (71.1 to 89.6)	p = 0.26 by log-rank (cumulative rates)
	24 months	a/a	34	82.1 (69.8 to 89.8)	37	69.8 (56.5 to 79.7)	p = 0.26 by log-rank (cumulative rates)
	36 months	a/a	25	76.4 (62 to 86)	25	67.5 (53.9 to 77.9)	p = 0.26 by log-rank (cumulative rates)
VARA54	6 months	Change in Rutherford classification (median)	27	2	23	2	p = 0.75
	12 months	a/a	27	2	23	2	p = 0.39

The RCT (Krueger et al. 47,48) reporting walking capacity found no significant differences between groups for EVBT and PTA on measures of pain-free walking distance or total walking distance up to 12 months post intervention (Table 41), with similar results up to 24 months. The patient-reported leg pain scores (Krueger et al. 47,48) were also similar between EVBT and PTA following intervention, although there was a borderline significant trend (p = 0.05) at 12 months favouring EVBT over radiation. Reported complications (Table 42) were similar for EVBT and PTA (Gallino et al. ,42 Vienna-3,51–53 VARA54).

TABLE 41 - Endovascular brachytherapy and walking capacity

a/a, as above.

Study	Follow-up	Definition of walking capacity	PTA analysed (n)	PTA outcome	Radiation analysed (n)	Radiation outcome	Comparative statistic
Krueger et al. 2002, 200447,48	1 month	Pain-free walking distance (m) (mean) (treadmill 3 km/h, slope 12 degrees)	15	288.1 ± 193.9	15	308 ± 191.2	p = 0.68
	6 months	a/a	15	307.0 ± 170.2	15	339.4 ± 185.3
	12 months	a/a	15	297.9 ± 205.8	15	329.2 ± 185.5	p = 0.72
	1 month	Total walking distance (m) (mean) (treadmill 3 km/h, slope 12 degrees)	15	321.1 ± 176.0	15	344.5 ± 171.7	p = 0.72
	6 months	a/a	15	345.4 ± 174.8	15	395.9 ± 140.1
	12 months	a/a	15	357.8 ± 170.4	15	393.0 ± 143.0	p = 0.59
	1 month	Walking distance leg pain scores at interview (max. 35)	15	25.7 ± 5.9	15	28.4 ± 4.5	p = 0.18
	6 months	a/a	15	25.7 ± 6.2	15	27.3 ± 5.3	p = 0.28
	12 months	a/a	15	22.2 ± 8.1	15	26.1 ± 3.7	p = 0.05

TABLE 42 - Endovascular brachytherapy and complications

Study	Follow-up	Definition of complication	PTA analysed (n)	PTA patients with complications (%)	Radiation analysed (n)	Radiation patients with complications (%)
Gallino et al. 2004,42 Bonvini et al. 200343	6 months	Late acute thrombotic occlusion	75	0	69	4.3
Vienna-351–53	12 months	Amputation	46	2.2	50	0
VARA54	Perioperative	Vessel thrombosis/early occlusion	33	3	27	3.7

Drug-coated balloon

Three RCTs were identified that compared DCB angioplasty with conventional (uncoated balloon) PTA. For all studies, the type of drug utilised was paclitaxel. Most of the patients across the studies had IC, although some had CLI.

Two studies (THUNDER,61–63 FemPac64) reported a significant advantage for DCB over PTA for restenosis rates (Table 46). When meta-analysed for restenosis at 6-month follow-up, these studies gave an RR of 0.40 (95% CI 0.23 to 0.69; p = 0.001), by both fixed- and random-effect analyses. Late lumen loss (LEVANT I59,60) and postintervention lumen diameter difference (THUNDER61–63) showed a significant treatment effect favouring DCB over PTA (Table 47).

Need for reintervention rates were lower in DCB than in PTA treatment groups (Table 48), significantly favouring DCB over PTA in two RCTs (THUNDER,61–63 FemPac64); the significance level was not reported in the other study (LEVANT I59,60) for TLR. Rates of TVR were also lower in the DCB than in the PTA group (LEVANT I59,60). TLR at 6-month follow-up, by meta-analysis of FemPac,64 LEVANT I59,60 and THUNDER61–63 trials, which showed some heterogeneity (Figures 16–21), produced a RR of 0.26 (95% CI 0.10 to 0.68; p = 0.006) by random-effect analysis. This significantly favoured DCB over PTA, which was also the case at 24-month follow-up using the FemPac64 and THUNDER61–63 trials (RR 0.27; 95% CI 0.16 to 0.47; p < 0.00001).

FIGURE 16.

Forest plot of comparison: 5 DCB vs. PTA, restenosis at 6 months fixed.

FIGURE 17.

Forest plot of comparison: 5 DCB vs. PTA, restenosis at 6 months random.

FIGURE 18.

Forest plot of comparison: 5 DCB vs. PTA, TLR at 6 months fixed.

FIGURE 19.

Forest plot of comparison: 5 DCB vs. PTA, TLR at 6 months random.

FIGURE 20.

Forest plot of comparison: 5 DCB vs. PTA, TLR at 24 months fixed.

FIGURE 21.

Forest plot of comparison: 5 DCB vs. PTA, TLR at 24 months random.

Two studies reported Rutherford category (Table 49). One study found no treatment group effect (THUNDER61–63), and one study (FemPac64) reported a borderline significant group difference, with more patients improving in the DCB group than in the PTA group at 6 months. However, in the latter study, by 18–24 months post intervention there was no significant difference between the groups, with the PTA group remaining stable and the improvement lessening in the DCB group, although both groups still improved from pre intervention. Complications and adverse events (Table 50) showed no significant treatment effects between DCB and PTA groups (LEVANT I,59,60 THUNDER,61–63 FemPac64).

Laser angioplasty

Five RCTs were included that compared laser angioplasty with PTA. Restenosis at 12-month follow-up was reported by one trial (Lammer et al. 68), which found no significant treatment effect between laser and PTA (Table 51).

One study reported clinical success (Belli et al. 65,66) measured by symptoms and peripheral pulses, and found a borderline significant trend favouring PTA over laser angioplasty (Table 52). Procedural complications were similar in laser and PTA groups (Belli et al. ,65,66 Lammer et al. ,68 Spies et al. ,69 Tobis et al. 70), with the exception of dissection, which was significantly more frequent with laser angioplasty than with PTA (Table 53).

Hunink et al.77

• Indication: IC and CLI.

• Lesion type: stenosis and occlusion.

• Site: femoropopliteal.

• Comparator: PTA.

• Interventions: BS or no treatment.

• Costs: 1990 US dollars.

• Health utilities: Torrance multiattribute scale.

(1990 US dollars presented in Hunink et al. 77 These values were updated to 1999 US dollars in Muradin and Hunink. 80)

Together the comparator and interventions constitute three treatments. Six specific treatment strategies were compared. Initial PTA could be followed by any of the three treatments. Initial BS could be followed only by NT or graft revision. No treatment completed the strategies under consideration. A maximum of two treatments per patient were modelled.

The authors used a patient-level ‘multistate transition model’ using a lifetime horizon programmed in Borland C (Borland, Scotts Valley, CA, USA). The perspective was that of the health-care system. Input and results were disaggregated by lesion type and graft material. CLI was subdivided into ‘rest pain’ and ‘necrosis’.

Costs for repeat procedures are assumed to be equal to the initial procedure cost. Annual follow-up costs are also provided, depending on whether or not the patient maintained patency, or if they had an amputation.

Quality of life was based on the Torrance multiattribute scale as valued by two vascular surgeons, two interventional radiologists and an internist. Utility values are based on the patient’s indication, and are also altered if the patient receives successful treatment or if the patient receives an amputation. These states are further divided depending on whether or not major morbidity (see below) is present. Procedure-specific decrements are also applied.

Initial success and patency rates are based on a previous systematic review. Disease progression was not modelled. Operative mortality rates were based on 26 studies, and depended on the type of operation and whether or not the patient was ‘high risk’ – defined as being aged 65 years and over with CLI and/or documented coronary artery disease. Procedure-related complications were modelled as the development of non-fatal systemic morbidity (which includes major cardiopulmonary, renal or cerebrovascular complications). Long-term mortality was modelled as an excess per cent, based on the ABPI (2% above the annual risk for individuals with ABPI > 0.3, 12% above the annual risk for individuals with ABPI ≤ 0.3). For a sensitivity analysis, a RR of 3.1 is used, regardless of indication.

Based on an incremental cost-effectiveness ratio (ICER) threshold of US$50,000 per quality-adjusted life-year (QALY), the authors come to the following conclusions:

• Initial PTA is recommended for all patients with stenoses, and claudicants with occlusive lesions.

• Initial BS is recommended for patients with both CLI and occlusions.

The authors only presented selected results. QALYs gained range from 2.7 to 7.4 for stenosis and 2.6 to 7.0 for occlusions. Costs range from US$15,000 to US$43,000 for stenosis and US$24,000 to US$51,000 for occlusions.

A variety of univariate sensitivity analyses were performed, with most of the parameters varied according to observed ranges within the literature. Multiway sensitivity analyses considered ‘optimistic’ and ‘pessimistic’ scenarios. Results were found to be most sensitive to procedural mortality and morbidity rates.

Sculpher et al.76

• Indication: IC and CLI.

• Lesion type: occlusions.

• Site: not stated.

• Comparator: PTA.

• Interventions: PTA with laser-assisted PTA on acute failure.

• Costs: 1993/94 UK pounds.

• Health utilities: European Quality of Life-5 Dimensions (EQ-5D) used; SF-36 values also available.

(Intervention uses data from a study on disease in the femoropopliteal arteries.)

For this cost–utility analysis, a two-part model was used, with a decision tree for initial revascularisation outcomes; these outcomes are then used as the starting health states in a Markov model that employed a lifetime horizon. Only the decision tree explicitly modelled the effects of the intervention. With the exception of death, all probabilities in the decision tree were taken from a single RCT (Lammer et al. 68). This RCT also showed that primary laser-assisted PTA was dominated by primary PTA, so this intervention was not considered.

If the initial operation (with or without laser assistance) failed, then patients may have BS and/or an amputation. There does not seem to be a limit on the number of BS operations that a patient may receive; in addition, patients with IC are able to receive repeat PTA (with or without laser assistance). Bilateral disease is not considered.

A crucial limitation concerning the cost-effectiveness data is that the cost-effectiveness of the laser when used as a secondary intervention (on immediate failure) is based on only seven patients. Long-term cost-effectiveness is based on a Markov model with a cycle length of 1 month, with a time horizon of 25 years. Disease progression was not modelled. General mortality is based on a Gompertz function, adjusted for an increased RR owing to having PAD (RR = 2 for IC and 3 for CLI). All other transition probabilities are independent of time and based on a mixture of published studies, an audit of patients’ notes at John Radcliffe Hospital in Oxford (where a co-author worked) or the clinical judgement of one of the co-authors. The paper does not state which probabilities came from which source. Procedure-related mortality is dependent on indication (IC or CLI) for BS but not for PTA. The secondary use of the laser is assumed not to result in any procedure-related deaths. Procedure-related complications are not modelled.

Utility values were elicited for four health states (IC, CLI and amputation above/below the knee) using both the time trade-off (TTO) method and the EuroQol visual analogue scale (EQ-VAS). Values for successfully treated patients were assumed to equal one. Two samples were used during elicitation: one of 36 health-care professionals (with a 100% response rate), and a random sample of the public (size not stated). As the values elicited were very similar for the two samples, only the results from the latter are used. In the base case, TTO values were used, with EQ-VAS values used in a sensitivity analysis; this did not have a noticeable impact on the results.

Costs are broken down into one-off costs based on procedure type (with an additional cost of angiography for any procedures during the Markov model) and monthly costs based on health state (cured, IC, CLI, amputee). For one-off costs, a breakdown of inpatient and outpatient costs is presented.

For each indication, the numbers in each health state and the numbers receiving a repeat operation (PTA or BS) are presented in 5-yearly increments.

For IC, the secondary use of a laser increases life-years from 6.78 to 6.79 and QALYs from 5.78 to 5.87, while increasing cost from £3669 to £3929. This gives an ICER of £3040 per QALY.

For CLI, the secondary use of a laser increases life-years from 5.44 to 5.46 and QALYs from 4.40 to 4.46, while increasing cost from £8716 to £8823. This gives an ICER of £1180 per QALY.

Sensitivity analyses showed that the following uncertainties had the greatest effect on results:

• The proportion of patients ‘cured’ following a successful operation (assumed = 100%).

• Annual utilisation of the laser (affecting its cost per operation).

• The proportion of patients in whom CLI recurs after reocclusion (assumed = 100%).

• Patency rates following PTA among CLI patients.

• The effectiveness of the laser.

de Vries et al.78

• Indication: IC.

• Lesion type: stenosis and occlusion.

• Site: above tibia.

• Comparator: PTA.

• Interventions: exercise and bypass surgery.

• Costs: 1995 US dollars.

• Health utilities: EQ-5D.

The authors compared five different treatment strategies involving sequences of exercise and PTA. BS was also included as an option in some of the strategies when PTA was deemed to be unsuitable. Exercise is an excluded intervention in our research, so results for this are not discussed here. The study presented an in-depth breakdown of outcomes for PTA and BS, broken down by site and lesion type, which are discussed here.

The authors presented results at both the aortoiliac level and the femoropopliteal level; the latter are of interest for this report. Rates of procedural mortality and systemic complications are taken from Hunink et al. ,77 as previously described. In addition, de Vries et al. 78 also include rates for angiographic investigations. For an amputation, mortality rates are presented separately for patients below and above the age of 75 years; rates for systemic complications were assumed not to vary with age.

As regards patency data, only 2-year results are provided. These are broken down by indication and intervention (PTA or BS). For BS, there is a further subdivision by graft type and, for PTA, there is a further subdivision by lesion type. The values used for the femoropopliteal level are taken from Hunink et al. 77

Health utility values for amputation and CLI are taken from Sculpher et al. 76 Values for IC and asymptomatic disease are taken from two other studies. Utility values associated with systemic complications are based on reported values for myocardial infarction survivors. Costs are taken from a mixture of published studies and the Medicare database. They are different from the costs used in any of the previous economic evaluations.

Holler et al.79

• Indication: CLI.

• Lesion type: occlusions. (Not entirely clear.)

• Site: not stated.

• Comparator: PTA.

• Interventions: BS, prostaglandin E1 (PGEI) or no treatment.

• Costs: euros, year not stated.

• Health utilities: EQ-5D.

This study looked at treatment strategies, with patients able to experience a maximum of two treatments. As PGE1 is an excluded intervention, only the information provided for PTA and BS are considered.

Cost-effectiveness data were based on a systematic review of German- and English-language literature. Values from studies were weighted by their sample size and the median value was taken. The probability of staying within the same health state (CLI or IC) is calculated based on the logical constraint that transition probabilities must sum to one. For patients with IC, the probability of dying was assumed to be the same as that for a 70-year-old German male (taken from life tables). Mortality rates vary depending on the initial treatment, unless a patient receives an amputation, in which case the probability of mortality is independent of initial treatment. It is assumed that patients with IC do not have an amputation.

Cost data are based on a survey of a patient sample (time period and setting not stated), which included 147 patients with IC, 92 with CLI and 40 who had had an amputation. Treatment costs are applied on a yearly basis, and are different for the two indications. The cost of an amputation is independent of the initial treatment.

Data on QoL are based on the EQ-5D questionnaire given to a sample of 280 patients with PAD. Separate values are given depending on initial treatment and indication. As with cost, the value for having had an amputation is independent of the initial treatment.

The BASIL trial (Forbes et al.82)

• Indication: Severe ischaemia. (CLI, but without the restriction that ABPI < 50 mmHg.)

• Lesion type: stenosis and occlusion.

• Site: infrainguinal.

• Comparator: PTA.

• Interventions: BS.

• Costs: 2006/07 US dollars.

• Health utilities: EQ-5D used; SF-36 values also available.

This is the only economic evaluation that was conducted alongside a clinical trial (ISRCTN 45398889). Detailed 12-month outcomes for the BASIL trial have been published (Bradbury et al. 84). Data from further follow-up have been presented in a number of publications (Forbes et al. ,82 Bradbury et al. 85–88).

The authors note that their inclusion criteria are different from the technical definition of CLI, but it was felt by our clinical expert (JAM) that they reflect CLI as defined in every-day practice, and thus the results of the BASIL trial are assumed for this evaluation to apply to CLI patients.

Between August 1999 and June 2004, the BASIL trial randomised 452 patients to a treatment strategy of either PTA first or BS first. There were a small number of crossovers; the economic evaluation uses an ITT analysis.

All of the data used in the model come from the BASIL trial. QoL was measured using the Vascular Quality of Life Questionnaire, the generic SF-36 health survey and EQ-5D. The EQ-5D is used within the economic evaluation. Cost data are based on hospital-related activity only. Both costs and utilities are discounted at 3.5% per annum.

Statistical regression methods were used to calculate incremental costs and incremental QALYs, with non-parametric bootstrapping used to assess uncertainty. As the costs data exhibited a heavy skew, the results from three different regression methods were reported. These are reproduced in Table 55, along with the corresponding ICERs. Although the results are not presented in UK pounds, it is clear that BS would not be considered cost-effective by decision-makers such as NICE using any of the three methods given current, and historic, exchange rates.

The National Institute for Health and Care Excellence cost-effectiveness analysis83

• Indication: IC.

• Lesion type: stenosis and occlusion.

• Site: iliac/femoropopliteal. (Analysed separately; only the latter is considered here.)

• Comparator: PTA (with selective stenting).

• Interventions: PTA (with primary stenting), unsupervised exercise, supervised exercise, BS.

• Costs: 2009/10 UK pounds.

• Health utilities: EQ-5D.

This economic evaluation considered two-stage treatment strategies. BS is considered only as a second-line treatment (giving four different first-line treatments). Neither PTA with primary stenting nor unsupervised exercise is considered as a second-line treatment (giving three different second-line treatments), resulting in (3 × 4) 12 different treatment strategies. A 13th strategy of PTA with selective stenting and supervised exercise (and no secondary treatment) is also evaluated.

A Markov model is employed using 3-monthly cycles. The analysis takes the perspectives of the NHS and personal social services. Both costs and QALYs are discounted at 3.5% per year.

Procedural costs (for PTA, BS and amputation) were taken from 2009/10 NHS Reference Costs. 89 For PTA and BS, the proportion of procedures that were elective or non-elective was based on expert opinion, with slight differences between the initial and repeat procedures. Ongoing costs were modelled only for patients who had undergone an amputation. Costs incurred in the first year were different from those incurred in follow-up years; both were based on a mixture of expert opinion and the 2010 Personal Social Services Research Unit. 90

Quality of life data for patients with IC were based on the values reported by the studies included in the evaluation. Only reports of EQ-5D or SF-36 (when sufficient data were available for them to be mapped to EQ-5D) were included, the final values used being the average of the included values. Data for patients with CLI or an amputation were taken from Sculpher et al. 76

It is assumed that PTA does not affect subsequent rates of mortality or morbidity and that repeat procedures have the same effectiveness as the initial procedure. Failure was taken to include both a loss of patency and symptom deterioration requiring reintervention. Perioperative complications, amputations and deaths were taken from an audit reported by the Royal College of Surgeon’s of England. 91 For patients with IC, there were no amputations or deaths. Based on expert opinion, these probabilities were felt to be non-zero, and therefore values of 0.5 amputations and 0.5 deaths were added to the numerator (and subtracted from the denominator) of the audit. Rates of failure and the amount of patients needing a reintervention are based on expert opinion and are modelled as fixed (time-invariant) amounts. The requirement for reintervention is assumed to vary depending on lesion type (stenosis or occlusion); the prevalence of lesions among patients with IC is based on expert opinion.

Progression to CLI was assumed to be independent of treatment strategy, with a 3-month probability of 0.1% (based on a value of 2% over 5 years). It was assumed that 25% of patients with CLI will receive an amputation as a primary intervention and that 25% will die each year (modelled by 3-month probabilities of 6.9% and 3.9%, respectively).

For patients with IC and femoropopliteal disease, the NICE CEA concluded that there were only four treatment strategies that were neither dominated nor extendedly dominated. These are detailed in Table 56.

Model description

A discrete-event simulation model (DESM) was developed in Simul8© 17.0 (Simul8 Corporation, Boston, MA, USA) to determine the cost-effectiveness of each enhancement compared with conventional angioplasty alone. A DESM was used in preference to a state-transition model primarily because of the large number of patient characteristics that required tracking over time. A DESM also more appropriately models time to event based on stochastic distributions.

Patient population

The population considered was patients with symptomatic PAD suitable for endovascular treatment for disease distal to the inguinal ligament. A lifetime horizon was used.

The patient population was subdivided into those with IC and those with CLI. The clinical classifications of these subgroups are presented in Table 57.

Differences in anatomical features were not explicitly modelled. These include features such as proximity to bifurcations, stenosis versus complete occlusions and length of occlusion. These differences were not considered because of a lack of available evidence for the comparator and interventions.

With two exceptions, the effectiveness of all interventions was evaluated in the femoropopliteal arteries. The exceptions were BMSs, which were evaluated in both the femoropopliteal and infrapopliteal arteries, and sirolimus-eluting stents, which were evaluated in the infrapopliteal arteries. As base-case data (for PTA) were available only for the femoropopliteal arteries, the results of evaluations considering the infrapopliteal arteries should be viewed as exploratory.

Interventions and comparators

The base-case analysis considers patients receiving conventional PTA with secondary bare-metal stenting if immediate (acute) failure occurs. Acute failure is defined as either failure of the operation or restenosis within 30 days of the operation. The interventions considered for this research are listed and described in Clinical effectiveness results in Chapter 3 on clinical effectiveness. Based on the results of the clinical effectiveness research, it was decided that there would be little value in including some of the interventions in the economic evaluation, as they were likely to be dominated by either the base case or a comparator (as they were less effective and likely to be more costly). Explicit costs for these comparators were not calculated; instead, it was noted that, because they are all enhancements to PTA, they will be more expensive than PTA. Hence, the following interventions were immediately excluded in the assessment of cost-effectiveness (the sections describing their clinical effectiveness can be found in Clinical effectiveness results in Chapter 3):

• AMSs

• atherectomy

• EBRT

• laser angioplasty.

No distinction was made between SESs and BESs, as (in general) use of the former has replaced use of the latter. As with the NICE CEA, the use of either of these stents is referred to as use of BMSs. CBs were not included in the economic evaluation, as they were recalled by their manufacturer because of a potential shaft separation of the catheter during operation (www.fda.gov/MedicalDevices/Safety/RecallsCorrectionsRemovals/ListofRecalls/ucm062951.htm referenced in White and Grey92).

The two patient populations (IC and CLI) are analysed separately. Owing to a lack of evidence, the treatment effect of each intervention is assumed to be the same for the two patient populations. It should be noted that in most trials the majority of participants have IC. Natural history data for the two patient populations (for example, patency rates for the comparator and time to amputation) vary. The effectiveness of BS (modelled as a second-line treatment) also varies by patient population.

Each intervention may be used as the initial treatment instead of (or with) PTA, with secondary stenting if required. In addition, the use of conventional PTA with secondary DESs (paclitaxel) was also reported in one study (Dake et al. 71). Because of this, paclitaxel-eluting stents were included as two interventions: one for their use as the initial treatment (no secondary stenting was required, so no distinction is made for this intervention) and one for their use only on acute failure.

To summarise, the base-case comparator and included interventions in the femoropopliteal arteries are:

• PTA with secondary BMSs (base case)

• primary BMSs

• PTA using a DCB

• primary DESs (paclitaxel)

• PTA with secondary DESs (paclitaxel)

• stent-graft

• cryoplasty

• EVBT.

In the infrapopliteal arteries they are:

• PTA with secondary BMSs (base case)

• primary BMSs

• primary DESs (sirolimus).

Outcomes

The main model outcome is the incremental cost per QALY gained. A secondary outcome of incremental cost per life-year gained is also presented.

Model structure

The structure of the decision model is presented in Figure 23. Events are modelled such that each event triggers changes in a patient’s health state. Patients can enter the model with either one leg or two; if the patient has two legs, the status of both legs is modelled. For simplicity, on receiving an amputation (to either leg), the only possible events for a patient are procedure-related death or general mortality. Patients can enter the model with either IC or CLI; these two groups are modelled and analysed separately.

FIGURE 23.

Diagram of the structure of the decision model.

For illustrative purposes, Figure 23 has been depicted with three groups of events, entitled ‘30-day outcomes’, ‘While patent’ and ‘On loss of patency’. On entering a group, the time to each event (or the probability that it occurs) is calculated, with the event occurring first being the next simulated event.

For example, upon entering the ‘On loss of patency’ group, the time to develop contralateral symptoms and experience disease progression and time to general mortality are all calculated. The probabilities of requiring and receiving a reintervention are also calculated and compared with random numbers (drawn from the uniform distribution on [0,1]) to see if they occur. If a patient were modelled as receiving a reoperation, this is given a time to event of 1 week. The event with the shortest time to occurrence then becomes the next simulated event.

In the following discussion, a reoperation refers to receiving PTA or BS only; it does not include receiving an amputation. For the purposes of brevity, the term ‘reoperation’ is also used to include the situation in which an individual receives an operation on a contra-lateral limb while the first limb remains asymptomatic.

The structure of the model from the perspective of a patient is presented in Figure 24, which shows the health states modelled. Patients enter the model with either IC or CLI. It is assumed that after a successful operation patients move into the asymptomatic health state, where they remain until they either die or suffer a loss of patency (failure). If a failure occurs, then, as with Sculpher et al. 76 and Hunink et al. ,77 it is assumed that the patient returns to their health status prior to the operation.

FIGURE 24.

Diagram of the health states modelled.

As with Sculpher et al. 76 and Hunink et al. 77 it is assumed that spontaneous improvement from CLI to IC (in the absence of an operation) does not occur. It should be noted that, if a patient’s operation fails (at any time), then they return to their health state prior to the operation, not their health state when entering the model. This affects IC patients; if they progress to CLI, then it is not possible for them to enter the IC health state again.

Patients may also develop contralateral symptoms (PAD in their other leg), so, for example, a patient with CLI in one leg may also develop IC in their other leg. As the status of each leg is tracked separately, Figure 24 actually represents the health states (and permissible transitions) for each leg.

Patients enter the model when undergoing their initial endovascular operation (which varies depending on the intervention or comparator considered). During the 30 days following an operation (the perioperative period), the following events may occur:

1. mortality attributable to the intervention or comparator

2. mortality not attributable to the intervention or comparator (general mortality)

3. acute failure (loss of patency)

4. success; defined as none of the above.

In addition, there is a probability of the patient developing a complication during the perioperative period. This is assumed to result in a (ongoing) utility decrement and cost, but it does not affect subsequent transitions, so it is not modelled as a separate event. Once a patient develops a complication, it is assumed that they remain with this complication for life.

General mortality is taken from life tables. 93 Mortality attributable to the intervention or comparator is not removed from the life tables, as the numbers are small. The life tables are adjusted to reflect an increased RR of dying due to having PAD. Separate RRs are modelled for IC and CLI. It is assumed that this excess risk remains even if a patient experiences a successful operation, as it is based on the patient’s disease prior to the operation.

If the initial operation is a success, then the patient moves into the asymptomatic PAD health state, and postprocedural events-while-patent are modelled:

1. Late failure (loss of patency).

2. Develop contralateral symptoms: these may be either IC or CLI and are influenced by the patient’s disease prior to their last operation.

3. Amputation: this is the risk of amputation owing to progression of disease in the limb (not owing to the result of loss of patency at the treated site).

4. General mortality.

If a patient suffers a failure (loss of patency) at any time point, then the following events are possible:

1. A reintervention (PTA, BS or amputation) is required and received.

2. The patient’s disease progresses to CLI. (This is applicable only for patients with IC.)

3. The patient develops contralateral symptoms (as previously described).

4. General mortality.

There are three situations for which a patient may require a reoperation:

1. After loss of patency, a proportion of patients are modelled as experiencing the (immediate) return of symptoms. If symptoms do return, then the patient requires a reintervention.

2. The patient develops contralateral symptoms.

3. The patient experiences disease progression (to CLI).

It should be noted that loss of patency on its own is not sufficient to require a reoperation; the patient must also experience a return of symptoms. In contrast, the data used to inform the transition probabilities for developing contralateral disease or disease progression both imply that a reoperation will be required as a result of the event.

If patency is lost, the probability of experiencing a return of symptoms is modelled as an immediate event and is independent of the type of operation received. It does, however, depend on the patient’s health state prior to the operation. If symptoms do return, then the patient moves into their health state prior to the operation. If symptoms do not return (but patency is lost), then patients with prior IC remain in the asymptomatic health state, but patients with prior CLI return to the CLI health state. This is because there is evidence to suggest that alternative forms of therapy (exercise and pharmacotherapy) are effective in improving the QoL of patients with IC but not patients with CLI. 83,94

There are three situations for which a patient may require a reoperation but not receive it (for further details, see the section ‘Probability of reintervention following failure’):

1. The patient dies before the operation is received.

2. The assumed maximum number of permissible reoperations (two) has already been reached.

3. The patient’s lesion is not suitable for reoperation.

Patients may receive an amputation at any time, with higher rates of amputation observed for patients with CLI. It is assumed that the low risk of amputation for those with IC is a result of the progression of disease, rather than being directly attributable to restenosis at the site of the original lesion causing IC. The model does not distinguish between below-knee and above-knee amputations but uses average costs and utilities for amputation based upon the proportions in the BASIL Trial. 84

Time horizon, perspective and discounting

The time horizon of the model was 100 years to ensure that all differences in costs and benefits are captured within the model. The analysis takes the perspectives of the NHS and personal social services. Both costs and QALYs were discounted at a rate of 3.5% per year.

Starting age

The starting age was based on data reported from the Swedish Vascular Registry,96 which was the only identified source that gave stratified estimates by indication (IC or CLI). The average (mean) age of a patient with IC receiving PTA was 66 years; for patients with CLI, the average age was 74 years. For comparison, where economic evaluations use (or state) their starting ages, for IC they range from 60 (de Vries et al. 78) to 67 (NICE CEA83). Neither Sculpher et al. 76 nor Hunink et al. 77 stratify their starting age by indication, both use a value of 65 years. Starting/average ages for CLI patients are not stated in the Holler et al. 79 and BASIL trials. 82

Starting age was not varied in probabilistic sensitivity analysis; the sensitivity of the base-case results to starting age was explored in a sensitivity analysis. For this, the variation reported in the Swedish Vascular Registry data96 was used to estimate a plausible range of ages. For patients with IC, a standard deviation of 10.4 about the mean of 66 was reported. For patients with CLI, a standard deviation of 9.3 about the mean of 74 was reported.

General mortality and excess risk

Patients with PAD were assumed to be at greater risk of general mortality than patients without PAD. In the majority of economic evaluations, this increased mortality was modelled as a RR, although the values employed, or suggested by the literature, can vary substantially. Further details are provided in Appendix 7.

Relative risks for IC (compared with the general population) vary from 1.6 to 4. The value of 3.1 quoted by Criqui et al. 97 is used for the base case; this source was also used by Hunink et al. 77 and in the NICE CEA. 83 It is also very similar to the value of 3.14 used by de Vries et al. 78

To ensure that patients with CLI do not have a lower probability of death than patients with IC, the RR of mortality for CLI is compared to that for IC. Two economic evaluations (de Vries et al. 78 and Hunink et al. 77) assume that this RR is 0; other values reported imply that the RR may be as high as 2.8. For the base case, a RR of 2 is used [based on data presented in the TASC-II (Trans-Atlantic Inter-Society Consensus II) guidelines (Norgren et al. 99)].

The RRs were included in the model by modifying 2009/10 UK life tables. 93

For the majority of the economic evaluations, it is not clear whether a successful operation reduces or removes any of this excess risk. The NICE CEA83 assumes that it does not reduce the risk at all, a view shared by our clinical expert (JAM). Hence, RRs after an operation (including amputation) remain the same as before the operation. It should be noted that this is a potential limitation; for example, for patients with CLI, successful treatment should remove some of the effects related to ischaemia of the limb.

Percutaneous transluminal balloon angioplasty failure

The meta-analysis of Hunink et al. 98 is employed in this evaluation; it was also used in four of the six economic evaluations. Of the alternatives, the NICE CEA83 bases its value on expert opinion and uses a fixed annual rate, whereas the BASIL trial82 reports failures only at 1 and 3 years. Further details of the Hunink et al. 98 meta-analysis are presented in Appendix 7.

Failure is defined as loss of patency; estimated failure rates over time are reproduced in Figure 25. To improve fit, failure during the perioperative period is modelled as a probability, as is failure during the first year (conditional on not failing during the perioperative period). Failure after the first year is modelled using a Weibull distribution (conditional on not failing during the first year).

FIGURE 25.

Percutaneous transluminal balloon angioplasty; cumulative failure rates over time for the two patient populations. Based on the meta-analysis of Hunink et al. 98

It should be noted that, if PTA fails during the perioperative period, then it is assumed that BS is always required and received. This assumption was employed in the economic evaluation of Hunink et al. ,77 and is used to reflect the fact that failures during the perioperative period usually indicate that repeat PTA would not be feasible (but it may be for longer-term failure).

30-day mortality following percutaneous transluminal balloon angioplasty or bypass surgery

Hunink et al. 77 provide the only economic evaluation to stratify mortality rates by patient status; this stratification is used within the model. Patients are deemed to be at ‘high risk’ of mortality if they are aged over 65, if they have a complication (as defined below) or if they have CLI; otherwise, they are at ‘low risk’ of mortality. High-risk patients have a probability of 30-day mortality following PTA of 3.2% and following BS of 4.7%. For low-risk patients, these values are 0.2% and 0.8%, respectively. It should be noted that, as the starting age of patients with IC is 65 years, all patients in the base case start with a high risk of mortality. The range of alternatives reported by Hunink et al. 77 is used to model uncertainty in these probabilities.

Complications during percutaneous transluminal balloon angioplasty or bypass surgery

A complication is defined as a non-fatal systemic complication (such as stroke, myocardial infarction and renal failure). For PTA, the BASIL trial (Bradbury et al. 84) is used to estimate the probability of a complication for the CLI population (6.75%), whereas the Royal College of Surgeon’s audit (Axisa et al. 91) is used for the IC population. These two sources are used because they reflect observed rates of complications. Axisa et al. 91 do not break down the complications by IC or CLI status, but the number of operations is broken down. This information is used with data from the BASIL trial84 to estimate a rate for IC (0.51%). For more details, see Appendix 7.

The number of complications reported by these two studies is reproduced in Table 62.

Estimates of complications during BS were modelled using a RR of 1.80, taken from the BASIL trial. 84 Although these data only relate to patients with CLI, they were used as it was felt that they provided the most plausible estimates. See Appendix 7 for more details.

Bypass surgery failure

This is modelled using the same meta-analysis as was used to model PTA failure. 98 This source was also used in four of the six economic evaluations. Of the alternatives, the NICE CEA83 did not model BS failure, and the BASIL trial84 reports failures only at 1 and 3 years.

As with PTA, life table estimates of patency for patients with IC and stenosis were presented. For BS, there was no statistically significant difference in patency by lesion type (hazard ratio not reported), so only the RR associated with having CLI was employed. The results are presented in Figure 26.

FIGURE 26.

Bypass surgery; cumulative failure rates over time for the two patient populations. Based on the meta-analysis of Hunink et al. 98

The meta-analysis98 also found that there was no difference between above-knee and below-knee operations, but that the type of graft material used affected the operation (with separate life tables presented for different graft types). Values for saphenous vein bypass are used in this analysis for two reasons:

• This type of operation was the most frequent in the BASIL trial (76%; 136/179). 84

• The NICE guidelines for PAD recommend using this type of graft when possible. 83

It should be noted that BS has a modelled perioperative failure rate of 0%.

Probability of reintervention following failure

It is assumed that, following failure during the perioperative period, a reintervention always occurs and it is always BS. If a patient experiences late failure, then three criteria must be met for a reintervention to take place:

1. Symptoms must return. For patients with IC, the values from the NICE CEA83 are used. It is assumed that 17.3% of patients with IC have an occlusion, as opposed to the 20% assumed by the NICE CEA83 (for more details, see Appendix 7: Percutaneous transluminal balloon angioplasty failure). This gives a probability of 28.3% that symptoms will return.

   For patients with CLI there are no direct data on the probability of symptoms returning following failure. The BASIL trial84 details the number of patients whose symptoms return, but not the number who lost patency. Instead, the failure rates used in this model are applied to the BASIL data, giving a probability of 72.7% that symptoms will return.

2. The individual must be eligible for a reintervention. Patients may be ineligible either because they have already received the maximum allowable number of interventions or because they are deemed to be physically ineligible. Based on discussions with our clinical expert (JAM), the maximum number of interventions (including the initial operation) was set at three; this is also the same number as was used by de Vries et al. 78

   Probabilities for being physically ineligible were taken from the only available evidence. de Vries et al. 78 assume that 5% of individuals with IC will be unsuitable for a reintervention. For patients with CLI, the most relevant data come from the BASIL trial,84 which states that, of 452 patients randomised to receive either PTA or BS, 14 did not receive any form of treatment. Removing the seven patients who did not receive a treatment because they died, the proportion ineligible is 1.55%.

3. The individual must not die before the planned reintervention occurs. Based on discussions with our clinical expert (JAM), it was assumed that the average time of reintervention following the return of symptoms was 1 week for patients with CLI and 1 month for patients with IC.

Type of reintervention and effectiveness

It is assumed that reinterventions are either PTA or BS. As previously mentioned, acute failure is always followed by a reintervention of BS. For the base-case analysis, reinterventions following late failure are always PTA; in a scenario analysis, this is changed to be always BS.

The BASIL trial86 is the only economic evaluation to compare the effectiveness of interventions when used as either the initial (first-line) or a follow-up (post-first-line) intervention. There was no evidence to suggest that there is any difference in patency rates between first-line and post-first-line interventions. This applied to both PTA and BS; for further details, see Appendix 7.

Disease progression

This is assumed to occur only following failure. Data from the analysis of Sculpher et al. 76 are used, as, of the evaluations that model disease progression, their assumptions regarding failure and disease progression are the closest to those used in this model. For further details, see Appendix 7.

Developing contralateral symptoms

Data were taken from an article by de Vries et al. ,100 as used in the economic evaluation of de Vries et al. 78 This is the only evaluation to consider contralateral symptoms. Values for contralateral symptoms requiring a reintervention are used, and adjusted to account for the fact that only 87% of contralateral symptoms (following infrainguinal disease) will also be in the infrainguinal arteries. de Vries et al. 78 state that, for patients with previous CLI, 67% of the contralateral symptoms are CLI (the rest being IC), whereas, for patients with previous IC, 10% of the symptoms are CLI. These values are used in this analysis. The development of contralateral symptoms is independent of the patient’s patency status.

Transition probabilities for a patient with CLI or IC are independent of how the disease was developed.

Amputations

The handling of amputation varies markedly across the economic evaluations. For this analysis, there are two key questions regarding the modelling of amputations.

Amputation-related mortality

Only three economic evaluations provide data on procedural mortality (Hunink et al. ,77 de Vries et al. ,78 NICE CEA83). There are no reported differences in rates between patients with IC and patients with CLI. Hunink et al. 77 use a value of 11.5%, and the NICE CEA83 uses a value of 12.9%. de Vries et al. 78 stratify their rates by age, with a rate of 9.8% among patients below the age of 75 years, and 14.7% above; these values are used in the model.

The NICE CEA83 is the only evaluation to model a change in the rates of general mortality following an amputation. Whereas the annual mortality rate of patients with CLI is 25%, patients experience a mortality rate of 35% in the first year following an amputation, followed by an annual rate of 19%. For this model, it is assumed that there is no difference in general mortality rates following an amputation.

Quality of life

There was wide variation in the QoL values employed in the existing economic evaluations. A detailed discussion of these is presented in Appendix 7. The values of Hunink et al. 77 are not used, as they were based on the abbreviated form of the Torrance multiattribute scale. All other evaluations used the EQ-5D.

Baseline (pre-treatment) values are taken from the analysis of Sculpher et al. 76 It is assumed that following patency failure an individual’s QoL returns to his or her pre-treatment value.

For IC, the value elicited by Sculpher et al. 76 (0.70) is nearly identical to those elicited by de Vries et al. 78 (0.71) and Holler et al. 79 (0.70). The value used by the NICE CEA83 is much lower (0.57 – this is the average of two studies); Spronk et al. 103 provide a value similar to that of Sculpher et al. 76 (12-month value: 0.77), whereas the value provided by Greenhalgh et al. 104 is much lower (12-month value: 0.48) – this is the only study to not directly measure EQ-5D (values are mapped from SF-36).

There is much variation in the reported values for CLI. The BASIL trial82 (0.26) reports the most recent data, and directly elicits its values from CLI patients. However, there were high levels of comorbid cardiovascular disease in these patients. As the effect of cardiovascular disease is separately modelled, use of this data may not be appropriate. Sculpher et al. 76 elicited their value (0.35) from the general public, and thus the effect of comorbid disease should be less pronounced. This value was also used by de Vries et al. 78 and in the NICE CEA. 83 Holler et al. 79 elicited their value (0.60) from CLI patients; it is noted that this value is over twice that reported by the BASIL trial. 82

Quality of life following an amputation is taken from the analysis of Sculpher et al.,76 who separate their values by above-knee (0.20) and below-knee amputations (0.61). The proportions of these are taken from the BASIL trial (out of 41 amputations, 13 were above the knee and 28 were below the knee84). Both de Vries et al. 78 and the NICE CEA83 use the values of Sculpher et al. 76 The only other evaluation to report EQ-5D values following an amputation is that of Holler et al. 79 (0.52), although it does not state the proportion of above- and below-knee amputations.

The effect of systemic complications is assumed to have a multiplicative decrement on QoL. Only the NICE CEA83 and de Vries et al. 78 report the effects of these. The NICE CEA83 reports the effect of both myocardial infarction and stroke, with separate values for the first and subsequent years. Values following the first year are based on the arbitrary assumption that they are half that of the first year. de Vries et al. 78 only report the effect of myocardial infarction, assuming a constant effect.

The utility decrement reported by de Vries et al. 78 is used. This is primarily to keep the model simple, as there are no data to suggest that any of the interventions alters the probability of experiencing a systemic complication. In addition:

• Both Axisa et al. 91 and the BASIL trial84 indicate that an myocardial infarction is much more likely to occur than a stroke.

• The NICE CEA83 states that the derived decrement following the first year is based on the arbitrary assumption that it is half that of the first year.

Costs

The NICE CEA83 values costs using the same perspective and time frame (2009/10 NHS reference costs89) as this economic evaluation, so costs are taken from this with the following exceptions:

• The NICE CEA83 assumes no long-term costs for patients with IC or CLI. In contrast, Hunink et al. 77, Sculpher et al. 76 and Holler et al. 79 all assume that there are costs. Of these, Sculpher et al. 76 is the only evaluation to base their costs on assumed resource use. The costs of these are updated using 2009/10 costs89 and used in the model. For further details see Appendix 7.

• The long-term costs for patients following an amputation are 20.3% higher in the first year than in subsequent years in the NICE CEA. 83 To keep this model simple only the long-term costs are employed (these are applied at all years).

• The NICE CEA83 uses a slight increase in the cost of repeat PTA procedures (less than 1%) due to an (assumed) increased number of non-elective admissions. In this evaluation all repeat PTAs cost the same as the initial PTA, with the weight given to non-elective admissions based on their observed frequency of occurrence in the NHS reference costs data.

• The cost of systemic complications is divided into myocardial infarction and stroke, with an increased cost in the first 3 months in the NICE CEA. 83 Only the costs for myocardial infarction are used in this evaluation for the reason described in the QoL section. As the presence of complications results in an increased procedural cost, the increased cost in the first 3 months is not included, as this may lead to double counting.

The effects of complications

There were no data available that showed how clinical effectiveness, QoL or costs were affected by the presence of a complication. Hence, these are assumed to have the same effect on interventions as they have on PTA, namely they:

• do not alter subsequent transition probabilities

• have a multiplicative decrement on QoL

• increase procedural costs by 91.7%. (In comparison the costs for BS increase by 44.2%.)

Data on costs

Costs data were derived from two main sources. The cost of interventions involving stents (A, B, D, E, α and β) is based on the base-case cost of PTA with bail-out stents (£3348), adjusted for the cost of a stent (£900 for DESs and £500 for BMSs) and their frequency of use (two per patient when used as the primary procedure, 0.324 per patient when used as a bail-out procedure). These data are taken from the NICE CEA. 83 It should be noted that both types of DES are assumed to have the same procedural cost; however, as the two are applied to different sites, it is not possible to compare the two. It is assumed that the cost of C (paclitaxel-coated balloon) is equal to the base case plus the incremental cost of drug coating (taken to be the difference in costs between a DES and a BMS).

Data for the remaining interventions (EVBT, stent-grafts and cryoplasty) were taken from the literature. 105–107 Instead of adjusting quoted costs to 2009/10 UK pounds, the costs were compared to the quoted costs for the base case and the excess cost applied as a ratio to the base-case cost employed in this evaluation.

Data on clinical effectiveness

With two exceptions, the data on clinical effectiveness come from studies previously discussed in the assessment of clinical effectiveness (see Chapter 3, Results) and, therefore, will not be discussed further here. For each intervention, if multiple studies were available, these were meta-analysed, with the results presented in Results in Chapter 3. Sometimes there were multiple time points with data that could be used to inform clinical effectiveness data. In all instances, the data were judged to be consistent over time, and thus only one result was used, typically the 12-month values. For example, results of the meta-analysis of TLR for DCBs were available for 6 months and 12 months, with RRs of 0.24 and 0.27, respectively – the latter value is used in this evaluation.

Differences in late failure rates are conditional on any differences in acute failure and any differences in the return of symptoms are conditional on any differences in failure. For example, for DCBs a TLR RR of 0.27 is used as a proxy value for the RR of the return of symptoms. As DCBs have a RR for late failure of 0.40, the RR for return of symptoms is calculated as 0.27/0.40 = 0.68.

Data on the clinical effectiveness (late failure and return of symptoms) of cryoplasty were taken from the trial of Spiliopoulos et al. 45 Although this study is included in the assessment of clinical effectiveness, the results used here are adjusted for other variables (as reported in table 5 of the paper of Spiliopoulos et al. 45), and therefore differ from the unadjusted results reported in Results in Chapter 3 of this evaluation.

Data on paclitaxel-eluting stents are taken from a 2012 publication by Dake et al.,71 published after a systematic review of clinical effectiveness was undertaken.

Sources of evidence on each interventions used in the modelling are summarised in Table 65.

Cost–utility analysis: base case

Intermittent claudication: femoropopliteal arteries

Total costs and total QALYs for each intervention are displayed in Table 66, which is sorted by ascending price. Because the options are mutually exclusive, ICERs are presented based on a fully incremental analysis.

TABLE 66 - Full incremental analysis of PTA and all the potential interventions

Intervention		Costs (£)	QALYs	Incremental analysis
C	Paclitaxel-coated balloon	12,668	6.120	–
B	PTA with bail-out paclitaxel-eluting stents	13,032	6.081	Dominated by C
X(f)	PTA with bail-out BMSs	14,637	5.956	Dominated by C
A	PTA, no bail-out stenting	14,787	5.931	Dominated by C
D	BMSs	15,030	5.989	Dominated by C
E	Paclitaxel-eluting stent	15,692	5.993	Dominated by C
F	EVBT	15,891	5.984	Dominated by C
G	Stent-graft	16,171	5.989	Dominated by C
H	Cryoplasty	17,578	5.934	Dominated by C

Intervention C, paclitaxel (drug)-coated balloons, is both less expensive and more clinically effective than all of the other options and, therefore, it dominates them. Interventions C and B both dominate the comparator, whereas interventions A and H are dominated by it. The ICERs for the remaining interventions (vs. assumed standard care) are: D (£11,979), E (£28,701), F (£4150) and G (£46,318).

Any decisions regarding which interventions to adopt or fund would be based on the point estimates presented in Table 66. It is also important to look at uncertainty in the decision to adopt. These uncertainties are explored in the following sections. As part of the estimate of uncertainty, probabilistic sensitivity analysis was performed, with 1000 runs.

Figure 27 presents the incremental CEAC for the interventions and assumed standard care. This shows the probability of each procedure being cost-effective at various levels of willingness to pay (maximum acceptable ICER). Thresholds from £0 to £100,000 were tested. Of all the procedures, use of a DCB has the highest probability of being most cost-effective, and use of bailout DESs has the second highest probability of being most cost-effective for all willingness-to-pay thresholds. The probability of any of the other interventions being cost-effective is never greater than 1%. The actual probabilities for each procedure are presented in Table 67 for selected willingness-to-pay thresholds of £20,000, £30,000 and £50,000.

FIGURE 27.

Incremental cost-effectiveness acceptability curve for the base-case model results (all but two of the interventions have probabilities ≈ 0 for all willingness-to-pay thresholds).

TABLE 67 - Incremental probability (%) of being cost-effective for specified levels of willingness to pay

Threshold (£)	Intervention
	C	B	X(f)	E	D	A	G	F	H
20,000	61.8	37.0	0.6	0.3	0.2	0.1	0.0	0.0	0.0
30,000	61.9	36.8	0.4	0.4	0.3	0.1	0.1	0.0	0.0
50,000	62.6	35.5	0.2	0.8	0.7	0.0	0.1	0.1	0.0

Based on the values presented in Table 67, interventions C and B have the highest probability of being the most cost-effective. The cost-effectiveness plane for these interventions is presented in Figure 28, which shows the incremental clinical effectiveness and incremental costs of these interventions versus the comparator.

The cost-effectiveness plane shows that both of the presented interventions fall in all four quadrants, suggesting that there is a non-zero probability that each intervention could be dominated by the comparator (represented by points falling in the top-left quadrant).

FIGURE 28.

Cost-effectiveness plane showing incremental clinical effectiveness and costs of selected interventions vs. the comparator (base case).

Additional details of the two interventions B and C, along with the comparator, are presented in Tables 68 and 69. These show the main drivers for the observed differences in clinical and cost-effectiveness outcomes.

TABLE 68 - Breakdown of costs

Procedures exclude amputations.

Type of procedure	Average costs per patient (£)
	Comparator	Bail-out DESs	DCB
All proceduresa	8361	7851	7656
First procedure	3348	3461	3580
Follow-up procedures	4816	4198	3885
Amputations	198	192	191
Amputees	2401	2403	2295
IC	801	502	411
CLI	124	109	84

TABLE 69 - Breakdown of utilities and life-years

Health state	Average values per patient
	Comparator	Bail-out DESs	DCB
Life-years gained	7.80	7.78	7.78
QALYs	5.96	6.07	6.10
Asymptomatic	5.15	5.52	5.59
IC	0.73	0.47	0.44
Amputees	0.04	0.04	0.04
CLI	0.03	0.03	0.02

Table 68 shows that, although interventions B and C are both more expensive than the comparator, a large component of their cost saving comes from avoiding repeat procedures (by prolonging patency). These interventions also save costs by keeping patients out of the IC health state for longer.

Table 69 shows that, although there is no difference in extension to life offered by the interventions, they keep patients out of the IC health state for longer, resulting in greater QoL.

For both tables, differences in amputation outcomes are minimal, as expected given the assumption that all of the interventions are assumed to have no impact on time to amputation. As amputations are associated with large costs and decrements to QoL, if there is an effect of interventions on reducing these, the cost savings and increases in QoL shown here are likely to be even greater.

An analysis of the expected value of perfect information (EVPI) based on the method described in Claxton and Posnett109 was undertaken and the results are shown in Figure 29.

FIGURE 29.

Results of EVPI.

Figure 29 may be interpreted as showing that there is uncertainty in which treatment is more efficacious, with the result that EVPI increases as willingness to pay increases. Often one treatment is more efficacious and, thus, EVPI reaches a maximum; its value decreases as willingness to pay increases and the more efficacious treatment is adopted. In this situation, the decision of which treatment is the most cost-effective does not appear to be dependent upon willingness to pay (the maximum acceptable ICER to a decision-maker). Instead, the most cost-effective treatment is dependent on the clinical effectiveness of the various treatments, and the uncertainty about these treatment effects. This is shown by the CEAC of Figure 27, as the two treatments with non-negligible probabilities of cost-effectiveness have essentially flat curves. Because of this, increasing the maximum acceptable ICER will lead to an increase in the EVPI, as shown in Figure 29. This is in contrast to more commonly seen figures in which there is a trade-off between the cost of an intervention and its efficacy.

It is estimated that about 7% of persons aged ≥ 60 years have IC. 2 Applying this value to 2010 mid-year population estimates for the UK,1 and assuming that the information from this report will be of benefit for a 10-year horizon, gives a multiplier for the EVPI values of 9,847,740.

Critical limb ischaemia: femoropopliteal arteries

Total costs and total QALYs for each intervention are displayed in Table 70, which is sorted by ascending price. Because the options are mutually exclusive, ICERs are presented based on a fully incremental analysis.

TABLE 70 - Full incremental analysis of PTA and all the potential interventions

Intervention		Costs (£)	QALYs	Incremental analysis
C	Paclitaxel-coated balloons	49,890	3.402	–
B	PTA with bail-out paclitaxel-eluting stents	52,335	3.297	Dominated by C
D	BMSs	54,775	3.144	Dominated by C
E	Paclitaxel-eluting stent	55,012	3.157	Dominated by C
X(f)	PTA with bail-out BMSs	55,199	3.047	Dominated by C
G	Stent-graft	55,852	3.144	Dominated by C
F	EVBT	55,928	3.134	Dominated by C
A	PTA, no bail-out stenting	56,539	2.988	Dominated by C
H	Cryoplasty	58,097	3.003	Dominated by C

As with the results for patients with IC, intervention C, paclitaxel (drug)-coated balloons, is both less expensive and more clinically effective than all of the other options and, therefore, it dominates them. Interventions A and H are again dominated by the comparator, being both more expensive and less clinically effective.

Procedures which include some form of drug and/or the primary use of stents (interventions B to E) are both less expensive and more effective than the comparator of PTA with bailout stenting.

Endovascular procedures that have a different mechanism of action from the comparator (interventions F, G and H) are all more expensive. Only intervention H (cryoplasty) is also less effective. However, the majority of the excluded interventions (atherectomy, EBRT and laser) similarly have a different mechanism of action from the base case, but were excluded because they were known to be both more expensive and less effective. The ICERs for interventions G and F (vs. the comparator) are £6681 (G) and £8341 (F).

Any decisions regarding which interventions to adopt or fund would be based on the point estimates presented in Table 70. It is also important to look at uncertainty in the decision to adopt. These uncertainties are explored in the following sections. As part of the estimate of uncertainty, probabilistic sensitivity analysis was performed, with 1000 runs.

Figure 30 presents the incremental CEAC for the interventions and comparator. This shows the probability of each procedure being cost-effective at various levels of willingness to pay (maximum acceptable ICER). Thresholds from £0 to £100,000 were tested. Of all the procedures, use of a DCB has the highest probability of being most cost-effective and use of bailout DESs has the second highest probability of being most cost-effective for all willingness-to-pay thresholds. The probability of any of the other interventions being cost-effective is never greater than 0.5%. The actual probabilities for each procedure are presented in Table 71 for selected willingness-to-pay thresholds of £20,000, £30,000 and £50,000.

FIGURE 30.

Incremental cost-effectiveness acceptability curve for the base-case model results (all but two of the interventions have probabilities ≈ 0 for all willingness-to-pay thresholds).

TABLE 71 - Incremental probability (%) of being cost-effective for specified levels of willingness to pay

Threshold (£)	Intervention
	C	B	E	D	F	G	X(f)	A	H
20,000	76.9	22.4	0.2	0.3	0.1	0.1	0.0	0.0	0.0
30,000	75.1	24.2	0.3	0.2	0.1	0.1	0.0	0.0	0.0
50,000	73.6	25.4	0.4	0.3	0.1	0.2	0.0	0.0	0.0

Based on the values presented in Table 71, interventions C and B have the highest probability of being the most cost-effective. The cost-effectiveness plane for these interventions is presented in Figure 31, which shows the incremental clinical effectiveness and incremental costs of these interventions versus the comparator.

FIGURE 31.

Cost-effectiveness plane showing incremental clinical effectiveness and costs of selected interventions vs. the comparator (base case).

The cost-effectiveness plane shows that, for some realisations of intervention B (but not intervention C), results fall in the top-left quadrants, suggesting that there is a non-zero probability that it could be dominated by the comparator.

Additional details of the two interventions B and C, along with the comparator, are presented in Tables 72 and 73. These show the main drivers for the observed differences in cost-effectiveness outcomes.

TABLE 72 - Breakdown of costs

Procedures exclude amputations.

Type of procedure	Average costs per patient
	Comparator (£)	Bail-out DESs (£)	DCB (£)
All proceduresa	14,949	13,685	12,432
First procedure	3348	3461	3580
Follow-up procedures	8320	6877	5511
Amputations	3282	3347	3341
Amputees	32,478	33,731	32,600
IC	87	69	23
CLI	1262	808	668

TABLE 73 - Breakdown of utilities and life-years

Health state	Average values per patient
	Comparator	Bail-out DESs	DCB
Life-years gained	5.17	5.25	5.20
QALYs	2.99	3.24	3.32
Asymptomatic	2.20	2.50	2.65
IC	0.03	0.02	0.01
Amputees	0.50	0.53	0.53
CLI	0.27	0.18	0.13

Table 72 shows that, although interventions B and C are both more expensive than the comparator, a large component of their cost saving comes from avoiding repeat procedures (by prolonging patency). These interventions also save costs by keeping patients out of the CLI health state for longer.

Whereas the costs for amputees are of similar magnitude for the comparator and both interventions, the increased costs for intervention B (owing to natural variation) are almost the same as the decreased procedural costs (owing to intervention effect). Therefore, setting amputation costs to zero will result in even greater cost savings for intervention B relative to the comparator. Setting amputation costs to zero will not make intervention B cheaper than intervention C however.

Table 73 shows that the main driver for differences in QoL is keeping patients out of the CLI health state and in the asymptotic health state for longer.

An analysis of the EVPI based on the method described in Claxton and Posnett109 was undertaken and the results are shown in Figure 32.

FIGURE 32.

Results of EVPI analysis.

The results of the EVPI analysis for patients with CLI are very similar to the results of the analysis for patients with IC. Again, there is an indication of some uncertainty over the results, with EVPI increasing as willingness to pay increases. In this situation, the decision of which treatment is the most cost-effective does not appear to be dependent upon willingness to pay (the maximum acceptable ICER to a decision-maker). Instead, the most cost-effective treatment is dependent on the inherent effectiveness of the various treatments, and the uncertainty about these treatment effects. This is shown by the CEAC of Figure 30, as the two treatments with non-negligible probabilities of cost-effectiveness have essentially flat curves. Because of this, increasing the maximum acceptable ICER will lead to an increase in the EVPI, as shown in Figure 31. This is in contrast to more commonly seen figures in which there is a trade-off between the cost of an intervention and its efficacy.

It is estimated that about 0.4% of persons aged ≥ 60 years have CLI. 2 Applying this value to 2010 mid-year population estimates for the UK,1 and assuming that the information from this report will be of benefit for a 10-year horizon, gives a multiplier for the EVPI values of 562,728.

Scenario analysis 1: varying age

In the base-case analysis, the starting age for patients was 66 for those with IC and 74 for those with CLI. These values were not varied within probabilistic sensitivity analyses; instead, the sensitivity of the conclusions reached in the base case to starting age are explored here.

Incremental costs and incremental QALYs for each intervention relative to the comparator are shown in Tables 74–77. As many interventions either dominate or are dominated by the comparator, ICERs are not presented.

No intervention alters life-years gained; their effect on costs is to avoid repeat reinterventions and their effect on QALYs is to keep patients in the asymptomatic health state for longer. For QALYs, effectiveness (or lack of it) shows a mostly smooth relationship with age, with effects becoming more (less) pronounced at younger (older) ages. This pattern can be seen for both patient populations.

A similar pattern can be seen for costs, although it is less pronounced – particularly for patients with CLI. This is likely to be a result of the effect of amputations, which are both very costly and independent of the intervention used.

The tables show that interventions B and C dominate the comparator for all ages and for both patient populations. For patients with IC, this effect becomes very small after the age of (about) 80; for patients with CLI, the effect becomes very small after the age of (about) 85.

Scenario analysis 2: no amputation costs

The average costs of the interventions considered range from £3837 (A) to £7367 (H), whereas monthly costs for IC are £15 and for CLI are £52. In contrast, an amputation procedure costs an average of £9733, with follow-up monthly costs just under £2000.

The high costs associated with receiving an amputation mean that any variations in the rates of its occurrence are likely to overwhelm any intervention effects (with respect to costs). Convergence tests were performed to ensure that natural variation is unlikely to be an issue, and it is assumed that interventions do not affect time to amputation (as there is not enough evidence to suggest otherwise). Care was taken to minimise any indirect effects of interventions on amputation rates. For example, patent patients cannot experience ipsilateral disease progression and, hence, interventions prolonging patency also result in lower rates of progression to CLI. As patients with CLI have a shorter time to amputation, this could result in an indirect intervention effect on amputation rates. To avoid this, time to amputation is set based on a patient’s characteristics at model entry, and is not changed upon disease progression (the same applies to time to general mortality).

There remains in the model one indirect intervention effect on amputation rates. Patency avoids the need for reinterventions and thus avoids the slight mortality risk associated with a reintervention. Because of this, more effective interventions will (on average) keep patients alive for a slightly longer time, meaning that patients are slightly more likely to receive an amputation.

This scenario analysis looks at the impacts on the base-case results when amputation-related costs are removed. Costs and QALYs for the comparator and each intervention are shown for each patient population in Table 78. As with the base-case, results are ordered by ascending price.

There are some slight differences from the base-case results in the order of interventions ranked near the middle. For example, for patients with IC, in the base case intervention A was £150 more expensive than the comparator, but, in this analysis, it is now £67 cheaper.

The main conclusions of the base case – that intervention C dominates all others and intervention B also dominates the comparator – remain unchanged. This is to be expected, given that interventions are not assumed to effect amputation rates.

Scenario analysis 3: reduced clinical benefit owing to patency and cost-minimisation approach

The effects of interventions on prolonging patency are taken from the literature. However, there is some concern over the link between patency and clinical outcomes, such as the need for reinterventions and the effect on QoL. The available literature linking these two is very sparse, and the NICE CEA83 did not include differences in patency because of a lack of evidence.

In the base-case analysis, prolonging patency has the following effects:

1. It improves QOL by stopping the return of symptoms (either IC or CLI).

2. It saves future costs by preventing the need for a reintervention.

3. It stops patients with IC experiencing ipsilateral disease progression.

It is worth noting that each of these effects is diluted:

1. After losing patency, not all individuals will experience a return of symptoms.

2. After losing patency, not all individuals will require a reintervention.

3. Patients with IC can experience contralateral disease progression at any point.

This sensitivity analysis looks at the impact on the base-case results if assumptions C and A are removed. Assumption C was removed by setting time to ipsilateral progression at model entry, using an exponential distribution with a parameter of 249.5 (this is based on the rate of disease progression modelled in the NICE CEA). Results from this are shown in Table 79.