Health economics results

Quality of life and quality-adjusted life-years

Table 21 presents EQ-5D utility values at baseline and at 6 and 16 weeks for the three treatment pathways. QALYs were similar for the three groups and equate to approximately 55 days or just under 2 months over a 16-week period.

Table 22 provides a breakdown of EQ-5D-5L responses for each of the five domains of the EQ-5D-5L for participants for whom QALYs were generated. At baseline, no one indicated that they had no pain or discomfort, but this improved at 6 and 16 weeks for all three pathways. Similarly, the proportion of patients with no problems with mobility, usual activities and anxiety or depression also increased over time.

Resource use

The most commonly accessed resources were prescription services, podiatry, GPs, practice nurses and outpatient appointments (Table 23). For those attending accident and emergency, patients could attend between one and four times per period. Hospitalisation duration ranged from one to eight nights, and those patients who attended outpatient appointments could attend up to 12 times per period.

Table 24 presents a summary of resource use costs for each pathway. Costs were similar across the three pathways, with treatment costs being cheaper for A-P (mean £1424) and slightly more expensive for P-A (mean £1448) and D-P (mean £1452). Similarities were also observed for overall costs, with P-A, on average, having the lowest cost (mean £1942), with the average costs for A-P (mean £2012) and D-P (mean £2032) being slightly higher, but similar.

Order of treatment effect

Regression models were fitted to both costs and QALYs to examine the effect of order of receiving treatment pathways A-P, D-P and P-A. Table 25 presents the results with 95% bootstrap CIs for the order effect compared with receiving treatments in the first time period. The overall effect of order on costs and QALYs was not significant for the three treatment pathways, although it did approach statistical significance for the A-P pathway, with patients receiving A-P in the third time period more likely to have lower costs than patients receiving A-P in the first time period.

Cost-effectiveness analysis

Table 26 presents a summary of the costs and QALYs over 16 weeks for each of the treatment pathways, as well as the pairwise incremental costs, QALYs and cost per QALY gained. Note that participants were eligible to receive all three treatment pathways. However, participants may drop out of the study before receiving all three pathways. Therefore, 67 participants received A-P and D-P, 67 participants received A-P and P-A and 73 participants received D-P and P-A. The incremental QALY gain is small and is uncertain when comparing A-P with D-P and A-P with P-A. However, there is a gain for those on the D-P pathway, equating to approximately 2.5 days when comparing D-P with P-A. The incremental cost gain was not significant for any comparison and ranged between a mean of –£113 for A-P compared with D-P and a mean of £154 for A-P compared with P-A. Figure 14 shows the cost-effectiveness planes for each pathway comparison. On average, D-P is more costly and less effective than A-P, with 57% of points being in the south-west quadrant of the cost-effectiveness plane (mean ICER £7026). A-P and D-P were, on average, more costly but more effective than P-A, with the majority of points being in the north-east quadrant of the cost-effectiveness plane (A-P vs. P-A mean ICER £7441, 79% in north-east quadrant; D-P vs. P-A mean ICER £94,157, 91% points in the north-east quadrant, respectively) (Figure 15). For all comparisons, it is important to note that they are based on a small QALY gain and a small difference in costs between the two treatment pathways.

FIGURE 14 Incremental cost-effectiveness plane. (a) A-P vs. D-P; (b) A-P vs. P-A; and (c) D-P vs. P-A. (Figure not reproduced here; see the original article.)

FIGURE 15 Cost-effectiveness acceptability curve. (a) A-P vs. D-P; (b) A-P vs. P-A; and (c) D-P vs. P-A. (Figure not reproduced here; see the original article.)

Sensitivity analysis

Non-pairwise comparisons of the three treatment pathways gave ICERs of £4000 for A-P compared with D-P (with D-P on average being more costly but more effective), –£700,000 for A-P compared with P-A (with P-A being less costly and more effective) and £18,000 for D-P compared with P-A (with P-A being less costly and less effective). The incremental QALY difference is very small and uncertain, which is reflected in the difference in the ICERs between pairwise and non-pairwise comparisons.

EuroQol-5 Dimensions, five-level version, Devlin algorithm

Table 27 presents EQ-5D utility values at baseline and at 6 and 16 weeks for the three treatment pathways for the crosswalk and Devlin algorithm. Utilities and QALYs were higher for the Devlin algorithm than the crosswalk, as observed elsewhere,78 but were similar across the three groups for both methods.

The D-P and P-A pathways become less cost-effective, compared with A-P, when the Devlin algorithm is used to estimate the EQ-5D-5L (A-P vs. D-P mean ICER –£37,784 and A-P vs. P-A mean ICER £31,909), with the ICER for D-P compared with P-A being lower (mean ICER £13,097). However, there remains uncertainty in the cost-effectiveness of all pathways.

Including costs borne by participants and their carers

Participants were asked about additional costs, including costs relating to help, transport, changes to the home, specialist equipment, time off work, loss of pay and help from friends and relatives over the past 8 weeks (Table 28). The largest costs were those where adaptions to the home were needed, although this affected only two participants, and costs incurred from caring support from friends and relatives, which were needed by 40–50% of participants.

Over the three treatment pathways, the mean costs borne by participants are more than double the other costs reported for the CSRI (see Table 28) and these were £1207 (95% bootstrap CI £758 to £2376) for A-P, £1740 (95% bootstrap CI £1077 to £3309) for D-P and £2138 (95% bootstrap CI £881 to £6836) for P-A. Overall costs per pathway also increased and these were £2670 (95% bootstrap CI £2220 to £3840) for A-P, £3216 (95% bootstrap CI £2561 to £4807) for D-P and £3619 (95% bootstrap CI £2358 to £8322) for P-A (Table 29). The inclusion of costs to participants and their carers did change the mean ICERs, although the results remained uncertain. D-P was more expensive and less effective than A-P (mean ICER –£19,268). P-A was more costly and more effective than both A-P (mean ICER £21,405) and D-P (mean ICER £229,855), respectively.

Multiple imputation of missing EuroQol-5 Dimensions responses

Imputing the EQ-5D responses increased the sample size to 130 (i.e. the full sample who participated in the study). On average, this increased the QALYs for A-P (mean QALY 0.157, 95% bootstrap CI 0.145 to 0.168) and D-P (mean QALY 0.160, 95% bootstrap CI 0.149 to 0.171) compared with the observed cases (see Table 26). However, the mean QALY for P-A remained similar to that prior to imputing the data (mean QALY 0.152, 95% bootstrap CI 0.141 to 0.164). As with other sensitivity analysis, imputing missing values did change the mean ICERs, with results remaining uncertain.