5 Outcomes

The Diagnostics Advisory Committee (appendix B) considered a review of the evidence by the External Assessment Group (EAG, appendix C).

How outcomes were assessed

5.1 The assessment consisted of a systematic review of clinical effectiveness data for the EOS system for the conditions described in the scope, followed by modelling to assess final patient outcomes and cost effectiveness when evidence was found. No studies followed patients to final outcomes, and therefore modelling was necessary to assess clinical effectiveness as well as cost effectiveness. Descriptions of the assessment process are contained in chapter 4 of the diagnostics assessment report. The only relevant data uncovered dealt with image quality and radiation dose reduction.

5.2 Three studies of image quality were found – two comparing the EOS system with X-ray film (Kalifa et al. 1998; Le Bras et al. unpublished data) and one comparing it with computed radiography (Deschênes et al. 2010). All found images from the EOS system to be comparable with or better than the comparator in most cases.

5.3 These three studies also reported radiation dose reductions with ratios of means ranging from 2.9 to 18.8, depending on the study and body part imaged. No direct comparisons with digital radiography were found, but similar dose reductions were assumed in the base-case models. A separate scenario assuming that digital radiography used two-thirds the radiation dose of computed radiography was also modelled.

5.4 Additional reviews were conducted to establish the impact of radiation dose reduction. Two different approaches to modelling cancer risk, identified in these reviews, were included in the assessment. These included data from a personal communication from the Health Protection Agency and data from the BIER VII phase 2 report.

Test accuracy: intermediate outcomes

5.5 No data meeting the inclusion criteria for the review were found to specifically compare the diagnostic accuracy of the EOS system with that of conventional radiological examinations beyond the three studies (described above), which showed comparable or better images. No evidence for sensitivity and specificity of the EOS system for any specific condition was uncovered.

Clinical outcomes

5.6 The only clinical outcomes assessed came from modelling, and were focused on the impact of radiation dose reduction in people with spinal deformities. Although direct evidence was available showing significant radiation dose reductions with the EOS system, modelling was needed to link dose reduction to reduced cancer occurrence. The base-case analysis used computed radiography as the comparator. Modelling of digital radiography was also performed as part of sensitivity analyses. Extensive modelling of the impact of radiation dose on future cancer was performed. The basic structure of the model is shown in appendix A. The modelling explored only the most prevalent forms of cancer, namely breast, lung, colorectal and prostate. In the base case, incremental quality-adjusted life years (QALYs) from cancer reduction as a result of radiation dose reduction varied by indication from about 0.0001 to 0.0009.

5.7 The original scope included spinal deformities (most of which were included in the model) and lower limb problems (which were excluded from the model because of a lack of evidence meeting inclusion criteria). In addition, the EOS system could be used for other conditions in which conventional radiography would usually be used. People with these other conditions might benefit from a reduced radiation dose as well. The EOS system also offers imaging enhancement for spinal conditions and lower limb problems, but the benefit associated with this enhancement could not be estimated from the existing evidence.

Cost and cost effectiveness

5.8 The EOS system costs 3–4 times as much as computed radiography machines and 2–3 times as much as digital radiography machines.

5.9 The cost-effectiveness analysis was performed using cancer reduction as the primary measure of benefit. The impact of throughput on cost effectiveness was modelled using three different assumptions about the throughput of the EOS system. The base-case throughput assumption (TA1) was based on using a single machine for the entire country and limiting use to only the number of cases of the studied conditions that actually exist in the country, with no other use of the machine. Additional throughput assumptions were based on full use of the machine for the indicated uses at the same throughput as computed radiography, namely 30 cases per day (TA2), or at a higher throughput, specifically, 48 cases per day (TA3). Because there are not enough cases of the indicated conditions to make full use of the machine, these last two assumptions were used to explore the impact on cost effectiveness of full use of the machine. If a machine that was fully used imaging the indicated conditions was found to be cost effective, then further analysis would have been needed to determine whether a machine partially used for the indicated conditions and also used for other conditions could still be cost effective. Thirty cases per day was the assumed rate of utilisation of the comparator. One reason the higher throughput of 48 cases per day may be justified is that the EOS system can take simultaneous PA and lateral images.

5.10 The base-case analysis showed the incremental cost-effectiveness ratio (ICER) to range from approximately £148,000 to over £15,000,000 per QALY gained, depending on the indicated use. The width of this range is primarily because the base case limits the use of the machine to the estimated number of cases of the studied conditions. For the throughput assumptions that are not limited by number of cases, the ICERs range from about £97,000 to £700,000 per QALY gained (TA2) and from £47,000 to £351,000 per QALY gained (TA3).

5.11 Additional scenarios modelled included:

  • earlier age for cancer diagnosis (55 years versus the average age of diagnosis in the population for the cancers modelled)

  • reduced discount rate for both costs and benefits (0% versus 3.5%)

  • further reductions in radiation dose (3 times the reduction of base case)

  • probabilistic modelling of QALYs gained from cancer reduction

  • increased cancer risk from radiation (using 1999 US data [BIER VII phase 2] versus newer models from the Health Protection Agency)

  • comparison with digital radiography (with digital radiography assumed to have a dose rate of two-thirds that of computed radiography).

5.12 None of these scenarios reduced the ICER to less than £30,000 using the throughput assumptions TA1 and TA2. The earlier age for cancer diagnosis or the alternative risk data reduced the ICER to less than £30,000 for scoliosis and Scheuermann's disease in adolescents for throughput assumption TA3.

5.13 Threshold analysis was performed to determine what level of additional benefits from imaging improvements would be required to reach levels that might normally be considered cost effective for each of the three throughput assumptions. This showed that additional QALYs required for cost effectiveness ranged from 0.0002 to 0.435, depending on the throughput assumptions and the condition being imaged. Threshold analyses of QALY gains required to reach an ICER of £20,000 under the six additional scenarios listed above varied from less than 0.001 to over 700 depending on the scenario, the condition, and the throughput assumptions.

  • National Institute for Health and Care Excellence (NICE)