Data collection overview

To conduct the economic evaluation, it was necessary to capture information on costs and consequences. Intervention costs were captured using a combination of methods including case report forms (CRFs), an adapted Client Service Receipt Inventory (CSRI) at 6 months’ and 12 months’ follow-up, and intervention delivery data collected by physiotherapists and the trial team. The EQ-5D-5L¹²⁷ was completed at baseline and the 6-month and 12-month follow-up. Utility values derived from the EQ-5D-5L were used to calculate QALYs for the primary analysis. For use in the sensitivity analysis, we also collected secondary care use data from NHS Digital. Data on inpatient hospital spells and outpatient attendances over the duration of the trial were sourced from HES data for financial years 2015–16, 2016–17 and 2017–18.

Costs: identifying resource use and costs

The costs within the analysis were divided into four components:

1. direct intervention costs (e.g. physiotherapy time and patient materials)
2. broader health-care/PSS costs (e.g. attendance at pain clinic)
3. wider costs (e.g. informal care)
4. set-up costs (e.g. intervention training costs).

The primary analysis adopted an NHS and PSS perspective and was concerned with the costs of delivering the intervention within an NHS setting. Thus, the primary analysis was concerned only with the direct intervention costs and the broader health-care and PSS costs. Set-up costs and wider costs were considered within the secondary analysis. This section first outlines the delivery costs and then set-up and training costs.

Direct intervention costs

Direct intervention costs were the costs associated with the introduction of the intervention compared with the usual-care group. All participants received usual care, which involved a 5-minute contact with a specialist BCN who provided usual-care leaflets (BCC6³³ and BCC151³⁴). In addition to leaflets, the intervention group then received a physiotherapist-led exercise programme. Resource use was captured prospectively alongside the trial and we summarise the collection of resource use components in Table 28.

TABLE 28

Resource use: direct intervention costs for usual care and exercise intervention

Broader health-care costs

Health-care resource use was captured primarily through section D of the CRF at 6 and 12 months (Table 29). Data on health-care use were collected for inpatient care, outpatient care, community health care, medication and equipment provided. HES data were obtained for 242 patients who had reached 12 months from randomisation by the end of the 2017–18 financial year, for use in secondary analysis. The resource use data collected within the CRFs were the primary source of cost data within the trial. Other wider costs considered within secondary analyses included out-of-pocket costs, privately purchased equipment and private health-care costs. A further analysis included set-up costs, which included resource use associated with training physiotherapists.

TABLE 29

Resource use: broader health-care, wider and intervention set-up resource use

Estimating quality-adjusted life-years

We used QALYs within the health economic evaluation, as per national recommendations.¹²⁶ QALYs combine quantity and QoL into a single metric. To calculate QALYs, it was necessary to obtain health state values for trial participants over multiple time points. We used the EQ-5D-5L, a five-dimension measure of HRQoL recommended by NICE.^126,127 There are value sets, also referred to as tariff values, that allow the calculation of utility values associated with each and every state generated by the EQ-5D-5L measure.¹³¹ At the time of writing, NICE preferred the use of the van Hout et al.¹³² algorithm;¹³² hence this value set was used to calculate utility values.

Health states were measured prospectively using the EQ-5D-5L at three time points: baseline, 6 months and 12 months. Health state values as measured by the EQ-5D-5L were combined with time to calculate QALYs by calculating the area under the curve using the trapezium rule.¹³³ This method assumes that the health states reported at each time point were linearly interpolated. Participants who died during follow-up were given an EQ-5D-5L score of zero at subsequent follow-ups beyond the date of death.

Secondary health economic outcome

The primary clinical outcome was the DASH upper limb disability questionnaire.³⁶ A secondary analysis used this clinical outcome measure to consider the cost per DASH point associated with the exercise intervention.

Estimating cost-effectiveness

To examine cost-effectiveness, it was necessary to jointly assess the incremental costs and incremental effects. In its most simple form, an incremental cost-effectiveness ratio (ICER) was presented. An ICER was calculated as follows:

Although ICERs provide a point estimate for cost-effectiveness, by themselves they do not characterise the uncertainty that may surround them. ICERs are estimated through the analysis of sample data, which may be subject to large variability, and there are inherent difficulties associated with characterising uncertainty around ratios, for example when the CI of the ICER overlaps zero and when the effect size approaches zero. Consequently, the net benefit approach is favoured for assessing the cost-effectiveness and characterising uncertainty within this analysis.

Net benefit regression framework

The net benefit regression framework was chosen as it has several strengths: (1) it transforms the cost/QALY data from a ratio into a continuous variable, allowing for easier manipulation while often normalising the data; (2) by combining costs and outcomes, it can seamlessly account for correlation between the two end points; (3) it allows easy control for baseline and covariate imbalances;¹³⁵ (4) it can correct for clustering using a multilevel framework; (5) it effectively deals with uncertainty around the decision-makers’ willingness to pay (WTP) for the health outcome of interest; (6) it facilitates the generation of CEACs to present decision uncertainty; and (7) it is relatively straightforward to implement in Stata using multiple imputation data.

Characterising uncertainty

It is important to present uncertainty in a manner that decision-makers can easily interpret. CEACs are a graphical representation of the probability that an intervention is cost-effective at different levels of WTP. NICE recommends that WTP thresholds of £20,000 and £30,000 per QALY are included in the CEAC when assessing uncertainty.¹²⁶ For a range of WTP thresholds, including those specified by NICE, CEACs were created to characterise uncertainty within cost-effectiveness estimates.

Sensitivity analyses

A number of sensitivity analyses were conducted to examine the uncertainty surrounding trial results. These sensitivity analyses included:

• Complete case analysis. This analysis considered only complete cases.
• Cost per DASH point. Should the intervention group be associated with higher costs than the usual-care group, then the costs per DASH point were to be estimated.
• Costing from a societal perspective. In this sensitivity analysis, wider societal costs were included within the cost-effectiveness analysis. This included NHS health costs, private costs and over the counter (OTC) medication.
• Incorporating training within the evaluation. Site staff were trained both centrally and at hospital sites, this analysis used a conservative approach whereby it was assumed each site was trained separately, with up to two trial staff undertaking training for 4 hours at each hospital site.
• Excluding high-cost cancer health-care use. This analysis limited costs to intervention costs, community care costs, outpatient physiotherapy, outpatient pain clinics, outpatient complementary therapies/exercise facilities and analgesics.
• Using HES cost data instead of CSRI data for hospital costs. This sensitivity analysis re-ran the primary analysis for the 242 participants with 12 months of complete data post randomisation, prior to the HES cut-off date (31 March 2018) and used HES data for costing hospital costs instead of CSRI inpatient and outpatient data. As these hospital data are obtained centrally, we assumed that these data were complete. Inpatient spells during the study and other hospital-based care costs were estimated by linking hospital episode data with Health Resource Groups, using the Reference Cost Grouper software¹³⁶ and then costed using NHS reference costs.¹²⁸

Resource use: NHS and Personal Social Services resources

Health-care resource use by type and quantity of resource for both trial groups at 6 and 12 months, along with statistical tests for difference between the two groups, is reported in Tables 30 and 31.

TABLE 30

Inpatient and outpatient resource use

TABLE 31

Community, social, medication and other wider health-care resource use

Inpatient resource use

Inpatient resource use was separated into breast cancer-related and non-breast cancer-related stays. We observed a decrease in resources used between 6 months and 12 months, with the proportion of inpatient contacts falling in both groups, and mean inpatient days falling from 1.8 to 0.4 for breast cancer-related inpatient stays, and from 0.3 to 0.2 for non-breast cancer inpatient stays. At 6 months there was no difference in resource use between the two groups for either breast cancer stays or non-breast cancer stays. There was no difference in the number reporting breast cancer-related inpatient stays (p = 0.25) or non-breast cancer-related inpatient stays (p = 0.81) by trial arm, and this was reflected in costs, with very little difference between the two groups (p = 0.89). At 12 months, there was no difference in contacts with inpatient services between breast cancer stays (p = 0.15) and non-breast cancer stays (p = 0.60). When considering mean breast cancer inpatient days, the number of days per person was lower in the intervention group than in the usual-care group (0.07 vs. 0.72 days, respectively; p = 0.04). Considering multiple testing and the very few non-zero values, this finding should be treated with caution. This is reflected in the relatively small and non-significant cost difference between the two groups at this time point.

Outpatient resource use

Similar to inpatient use, we observed a fall in the number outpatient contacts from 6 to 12 months (Table 30). The proportion of participants reporting breast cancer outpatient contacts fell in both groups, with the mean number of breast cancer contacts falling from 18 to 8. However, non-breast cancer contacts remained relatively steady, with an average of one contact per participant at each time point. Regarding the proportion of participants having breast cancer contacts, there were no differences between the two groups at 6 months (p = 0.08) or 12 months (p = 0.89). In terms of the proportion having non-breast cancer contacts, there were no differences at 6 months (p = 0.37) or 12 months (p = 0.74). There was, however, a statistically significant difference (p = 0.02) in non-breast cancer contacts at 6 months, with the intervention group having, on average, 1.3 contacts, compared with 0.6 contacts in the usual-care group. Again, multiple testing should be considered when interpreting this result. There were no differences at 12 months. In terms of cost, there was no significant differences between the two groups at either of the two time points for breast cancer and non-breast cancer outpatient contacts.

Community care resource use

Regarding use of community care resources, the proportion reporting contacts fell over the 6–12 months time period (Table 31). Likewise, the mean number of contacts fell from six to three over this time period. However, there were no between-group differences in the proportion of community care contacts at 6 months (p = 0.07) or 12 months (p = 0.27). Likewise, there were no between-group differences in number of community contacts at 6 (p = 0.08) or 12 months (p = 0.92). There were no differences in cost between the two groups at either time point (p = 0.16 and p = 0.94).

Special equipment resource use: NHS/Personal Social Services provided

There was no difference in the number of individuals receiving specialist equipment/accessories (Table 31). When focusing on equipment provided by the NHS and PSS, there were no difference in terms of cost of equipment between the two groups, with there being, on average, less than a £1 cost difference to the NHS between the two groups at both time points.

Medication resource use: NHS provided

Given the severity and burden of illness of trial participants, it is unsurprising to find high levels of cost associated with medication use per participant (Table 31). Focusing on NHS medication, as opposed to OTC, high mean costs were observed at both 6 months and 12 months. At 6 months, there were negligible cost differences between the intervention group and the usual-care group (p = 0.97). At 12 months, although there was a large cost difference (≈£900) between the two groups, this was not statistically significant (p = 0.15). This discrepancy appeared to be due to a small number of participants within the usual-care group receiving high-cost cancer drugs.

Direct intervention resource use

The direct costs related to the intervention were based on physiotherapy logs for participants who had attended physiotherapy. Participants had a mean of four contacts with physiotherapist and the mean length of appointment was approximately 40 minutes. The addition of physiotherapy administrative time added 8 minutes to this consultation (Table 32). There were a number of other small intervention-related costs, including an exercise manual and planner given to each intervention participant. Therabands were also given after 1 month postoperatively. All participants received cancer rehabilitation leaflets during a short preoperative session with a breast cancer nurse (Tables 32 and 33). Compared with other health-care costs, these costs were relatively minor, with total intervention costs coming to just over £100 per intervention participant.

TABLE 32

Intervention resource use: physiotherapy appointments

TABLE 33

Other intervention costs

Other resource use: non-NHS/non-Personal Social Services

Resource use: impact on employment

Reported impacts on employment are presented in Table 34. Across both groups, there was an increase in the number and proportion of participants able to work at 12 months compared with 6 months, although this was not a statistically significant difference. There were no differences between groups in terms of the numbers who reported taking time off work, the number of days of work missed or lost income as a result of missing work.

TABLE 34

Employment-related outcomes

Direct intervention costs

The mean cost of physiotherapy appointments for those in the intervention group was £103 (Table 32). Both trial groups received information leaflets; however, these contributed very little to cost. For the intervention group, there were other small costs, such as a personalised exercise planner and manual, manuals for the physiotherapist (Table 33) and Therabands; these costs were again relatively small (£26). The total direct incremental cost associated with the intervention compared with the usual-care group was £129.

Inpatient costs

In both trial groups, the costs associated with inpatient breast cancer care were significant (Table 35). Total breast cancer-related inpatient costs per person were slightly (£69) higher in the usual-care group than in the exercise group, but the difference was not statistically significant (p = 0.84). Non-breast cancer-related inpatient costs were slightly (£49) higher in the exercise group than in the usual-care group but, again, the difference was not statistically significant (p = 0.59).

TABLE 35

Total costs: inpatient, outpatient, community care, medication and other costs

Outpatient costs

The total costs associated with breast cancer-related outpatient care was over £3600 per person in both groups (Table 35). This dwarfed the costs associated with non-breast cancer-related outpatient visits. There was minimal difference between the two groups in terms of total cost (£20). By comparison, non-breast cancer outpatient contacts were relatively low in both the usual-care (£239) and exercise (£455) groups. Although the cost per person for non-breast cancer contacts was £217 higher in the exercise group than in the usual-care group, the difference was not statistically significant.

Community care costs

The total community care costs were relatively small compared with the outpatient care costs, with the average participant costing £531 in the usual-care group and £460 in the intervention group (Table 35). Again, there were no significant differences between the two groups in terms of the cost of community care accrued (p = 0.38).

Medication

Medication was the second biggest driver of cost (Table 35). The costs associated with medication were high in both groups (≈£3000 per person). Although the cost per person of medication was £335 lower in the exercise group than in the usual-care group, the difference was not significant because in both groups costs varied considerably between individuals (p = 0.79). The costs of OTC medication were minimal compared with those of prescribed medication and were almost identical, at only £27 per person in both groups (p = 0.99).

Equipment

There were few reported costs related to specialist equipment provided by the NHS (Table 35). There was very little difference between the two groups in terms of NHS equipment costs. The intervention group reported slightly lower (–£29) privately purchased equipment costs than the usual-care group (p = 0.7).

Other non-NHS/non-Personal Social Services costs

Private health care costs are shown in (Table 35). The cost of private health care was slightly lower (–£13) in the exercise group than in the usual-care group but the difference was not statistically significant. Wider costs were higher in the exercise group than in the usual-care group (+£114), but once more the difference was not statistically significant (p = 0.23). Both groups reported significant loss of income as a result of time off work, although lost income per person was over £1000 lower in the exercise group than in the usual-care group. However, there was high missingness for this self-report variable and, given the large variability in this variable, the difference was not statistically significant (p = 0.26).

Secondary analysis 1: complete-case analysis

Secondary analysis 1 considered the cost-effectiveness results using the complete-case data. The results are shown in Appendix 3 in the form of a net benefit chart (Figure 11) and a CEAC (Figure 12). Again, the complete-case analysis provided supporting evidence for cost-effectiveness, with there being a 65% chance exercise is the more cost-effective option at a WTP of £20,000 per QALY, rising to 68% at a WTP of £30,000 per QALY.

Secondary analysis 2: cost per DASH point

As reported in Chapter 4 and in Table 12, the exercise intervention was associated with improved DASH scores and lower costs. Given this, exercise dominated usual care and so calculating a cost per DASH point was deemed unnecessary because of the problems associated with interpreting a negative ICER.

Secondary analysis 3: including societal costs

Secondary analysis 3 considered the impact of broadening the costing perspective from NHS and PSS to a societal perspective. This included other private health-care costs, private equipment purchases, OTC medication and other costs. Income losses were omitted owing to the lack of data for this variable. In terms of cost-effectiveness, this further strengthens the case for cost-effectiveness, with the intervention continuing to dominate usual care. The results of the cost-effectiveness analysis are presented in Appendix 3, Figures 13 and 14. As can be seen from the CEAC, the intervention at a threshold of £20,000 per QALY has an 83% chance of being more cost-effective than usual care when costed from a societal perspective.

Secondary analysis 4: including training cost

Across the 17 sites, a total of 312 hours of training time were accounted for, including the time of the trainers. The inclusion of these costs led to an increase in costs per participant in the exercise group of £55.54. As shown in Appendix 3, Figures 15 and 16, the inclusion of training costs had very little impact on the results of the cost-effectiveness analysis. In this analysis the probability of the intervention being cost-effective at a cost-per-QALY threshold of £20,000 falls marginally, to 76.8%.

Secondary analysis 5: excluding high-cost cancer treatment

This analysis considered a narrower costing perspective limited to those costs that are most likely to be affected by shoulder problems, such as upper limb stiffness and pain, rather than cancer more generally. This led to much lower cost estimates, with the mean costs falling to £732 (95% CI £649 to £815) per person. In this analysis, the intervention was associated with an increased cost per person of £106 (95% CI –£49 to £262). As shown in Appendix 3, Figures 17 and 18, at this limited cost perspective the probability of the intervention being cost-effective at a cost-per-QALY threshold of £20,000 per QALY increases to 97%. This reflects the low costs and reduced uncertainty around cost-estimates within this analysis while QALYs remain the same.

Secondary analysis 6: using Hospital Episode Statistics costs for hospital care

This analysis used HES data to calculate hospital costs instead of CSRI data. Given the timescales involved in obtaining HES data within the trial timeline, it was possible to obtain full 12-month data for only 242 (62%) of the recruited participants. According to this analysis, costs were slightly higher in the exercise group (£166), with a great deal of uncertainty surrounding the estimate (95% CI –£3849 to £4181). This is reflected in the CEAC shown in Appendix 3, Figure 19, in which the CEAC, while maintaining similar overall shape to the primary analysis, has shifted downwards slightly. This reflects the increased costs and smaller sample size, which manifest in increased levels of uncertainty. However, there remains a 62% chance that the intervention is more cost-effective than usual care at a threshold of £30,000 per QALY.