Health economics analysis: results

Overview of the health economics analysis

The protocol for the DIAMONDS trial22 envisaged that economic modelling might be required if visual outcomes differed between treatment groups. However, the protocol was designed to minimise any visual loss. As per standard clinical practice, laser treatment was repeated following the initial session at the discretion of the treating ophthalmologists if it was felt to be required. Similarly, as per standard clinical care, rescue treatment with anti-VEGF therapy was allowed if rescue criteria were met. Intravitreal steroid injections could also be used if needed, at the discretion of the treating ophthalmologist.

Subthreshold micropulse laser was found to be not only non-inferior but indeed equivalent to SL. Thus, modelling was not required. This chapter provides the results from the within-trial cost-effectiveness analysis comparing SML with SL. The chapter presents (1) missing data by treatment group, (2) resource use and economic costs for different health resource categories, (3) health outcomes (EQ-5D-5L utility, NEI-VFQ-25 and VisQoL scores) and (4) cost-effectiveness results of the base-case and sensitivity analyses.

Results of the health economics analysis

Missing data

Table 14 shows the degree of missing health economics data by treatment group and follow-up time point. The missing data pattern was non-monotonic as individuals with missing data at one follow-up time point could have subsequent data entries. For example, there are more missing EQ-5D-5L data at 12 months than at 24 months. Overall, the percentage of missing data was low, ranging between 2% and 21% across all variables.

Economic costs and health-care resource use

Table 15 summarises the NHS costs associated with resource use among complete cases (i.e. cases with complete cost data) by cost category and follow-up period. We present the costs of the first laser procedure separately from that of the repeated laser procedures. Nonetheless, the total costs of laser therapy for each patient includes costs of the first laser procedure plus any subsequent laser retreatments they had (as captured in the trial CRF). The mean direct intervention cost for the first laser procedure was £47.11 (SE £2.65) for the SML group compared with £41.16 (SE £2.30) for the SL group; the mean difference of £5.95 was not statistically significant at the 5% level (95% CI: –1.00 to 12.90; p = 0.09). Equipment costs constituted a small proportion of the total costs as we assumed that the machine will be used to perform laser procedures on 3000 patients per year (see Appendix 1, Table 23). Further details on variables used to calculate laser treatment costs, for example times to complete laser treatments, are provided in Appendix 1, Table 26 and in Appendix 2, Figure 9.

The mean total NHS and PSS costs were lower in the SML group than the SL group (£897.83 vs. £1125.66) between baseline and 24 months post randomisation; the mean difference of £227.83 was not statistically significant at the 5% level. Table 16 presents more granular economic cost data by time period. The table also captures information on the number of participants who accessed a service over specified time periods.

The total costs in Tables 15 and 16 differ because one uses only data from participants with compete cost data at all time points whereas the other refers to participants with complete cost data at each of the specified time points. Outpatient visit costs are higher in those with complete attendance, as would be expected. The main difference is in anti-VEGF therapy rescue costs, which are much higher in participants with missing data.

Economic costs and resource use summary

• The mean number of laser treatments was 2.4 in the SML group and 1.9 in the SL group, difference 0.48 (p = 0.002). Of these treatments, 80% and 86%, respectively, were given in the first 12 months (Table 9).
• The proportion of patients requiring rescue treatment (almost all with anti-VEGF therapy – only one patient had a steroid injection) in the study eye was 18% in the SML group and 21% in the SL group, and the difference was not statistically significant (Table 9). About half of the patients receiving anti-VEGF therapies did so in the first 12 months (9.8% in the SML group and 12.9% in the SL group) (see Appendix 1, Table 27).
• The proportion of participants that met the criteria for rescue treatment at least once was 33% for the SML group and 31% for the SL group, therefore of those who at any one time point met the criteria, only 54% and 68%, respectively, were treated (see Appendix 1, Table 27). Some of those not treated had only temporary increases of a few micrometres in CRT that resolved without treatment.
• The mean number of anti-VEGF treatments in the SL group was skewed by five patients who received more than 10 treatments. None of the participants in the SML group required 10 or more injections.
• Laser treatment costs were based on the time from the patient entering the laser room to the time of their leaving the room, as recorded in the CRF.

Health outcomes

The EQ-5D-5L utility scores showed no statistically significant differences between the two laser groups (Table 17). The EuroQol-5 Dimensions (EQ-5D) visual analogue scale (VAS) scores followed a similar pattern across time periods for the two treatment groups (Table 18).

National Eye Institute Visual Functioning Questionnaire – 25 and Vision and Quality of Life Index outcomes

The VisQoL analysis showed no statistically significant differences in utility scores between the two treatment groups for any of the VisQoL dimensions at any follow-up time point (see Appendix 1, Table 28). Similar results were observed for the NEI-VFQ-25 subscales (Figures 5 and 6, and Appendix 1, Table 29). Overall baseline (pre laser treatment) utility scores from analysis of all six VisQoL dimensions and NEI-VFQ-25 subscales are detailed in Appendix 1, Tables 28 and 29 and in Report Supplementary Materials 3 and 4. The tables show that there are no statistically significant differences in utility scores for the VisQoL or the NEI-VFQ-25 subscales between baseline and 24 months.

FIGURE 5 The baseline NEI-VFQ-25 subscale and composite scores in participants treated with SML vs. SL. (Figure not reproduced here; see the original article.)

FIGURE 6 The NEI-VFQ-25 subscale and composite scores in participants treated with SML vs. SL at 24 months. (Figure not reproduced here; see the original article.)

Cost-effectiveness analysis

The cost-effectiveness results are presented in Table 19 with the SL group as the referent and the SML group as the comparator (i.e. SML minus SL) for the estimation of ICER values. The time horizon is the 24-month follow-up period of the trial. The joint distributions of costs and outcomes for the base-case analysis are represented in Figures 7 and 8.

FIGURE 7 Cost-effectiveness scatterplot with 95% confidence ellipses at 24 months for the base case. Imputed, additionally controlled for baseline utilities and minimisation variables (previous use of anti-VEGF therapy and previous laser treatment). (Figure not reproduced here; see the original article.)

FIGURE 8 Cost-effectiveness acceptability curves at 24 months for base case. Imputed, additionally controlled for baseline utilities and baseline minimisation variables. (Figure not reproduced here; see the original article.)

Base-case analysis

Over the 24-month follow-up period, participants in the SML group, compared with the SL group, experienced a non-statistically significant increase of 0.008 QALYs (circa 3 days of good quality of life) (95% CI –0.059 to 0.075 QALYs). In addition, the mean NHS and PSS costs were lower in the SML group compared with the SL group (mean cost difference –£365, 95% CI –£822 to £93). However, it should be noted that the CI for the cost difference is wide and ranges from cost saving to cost increasing. The ICER for the base-case analysis indicates that SML dominates, as average costs for this intervention were lower and average benefits were marginally higher than those for SL.

Assuming cost-effectiveness thresholds of £15,000 per QALY, £20,000 per QALY and £30,000 per QALY, the probability that SML was cost-effective compared with SL ranged from 71% to 80%, and the NMB associated with SML was positive (Table 19). However, the QALY difference of 0.008 was not significant. The results of the sensitivity analyses follow a similar pattern to the base-case analysis.

Discussion

Total costs were higher in the SL group, but the difference seemed to be driven by a higher number of anti-VEGF rescue injections, largely owing to five patients in the SL group having 10 or more injections. The CI around total costs were wide and overlapped, so findings should be interpreted with caution. There was no statistically significant difference in EQ-5D-5L scores and no clinically relevant difference on visual outcomes; SML and SL were found to be clinically equivalent.