4 Evidence and interpretation
The Appraisal Committee (appendix A) considered evidence from a number of sources (appendix B). No submissions were received from the manufacturers of the technologies considered. The evidence base comprised the evidence presented by the Assessment Group and the personal perspectives of the nominated experts. The objective of the appraisal was to determine whether it is clinically and cost effective to scan routinely all those with first-episode psychosis by either structural MRI or CT techniques compared with the standard practice of carrying out selective radiological examinations contingent on clinical findings suggestive of an underlying structural cause.
4.1.1 The Assessment Group identified 25 studies that had been conducted for different purposes and had a wide range of study populations. The relevant studies identified had varying objectives and only two of the studies were conducted in the UK. Nine studies were considered to relate to first-episode psychosis. Two of the studies in first-episode psychosis involved MRI scanning, six involved CT scanning and one study involved both techniques. All the studies included for the clinical effectiveness review by the Assessment Group had varying patient populations, and a high level of methodological heterogeneity. There was incomplete reporting of results and sampling bias, which the Assessment Group thought was likely to affect the results. Consequently, a quantitative meta-analysis of the study results was not possible.
4.1.2 Studies that included people with first-episode psychosis did not generally explain how this term was defined, and this could be important given the lack of precision in defining what is meant by the term first-episode psychosis (see section 2.1). Based on a review of the 25 studies identified, the Assessment Group estimated that MRI scanning resulted in findings that would influence clinical management in approximately 5% of people with psychosis (range of 0–10%). The corresponding figure for CT scanning was approximately 0.5% (range of 0–5%). However, these estimates are subject to considerable uncertainty, given the nature of the studies and also the possibility that studies that do not demonstrate the usefulness of the technology remain unpublished.
4.2.1 A systematic review of studies on the cost effectiveness of structural neuroimaging in first-episode psychosis found no relevant economic evaluations. Nor was any evidence found on differential treatment responses to antipsychotic drugs in organic and functional psychoses or on quality-of-life benefits following early diagnosis (from routine screening). Because of the lack of data to populate a comprehensive decision-analytical model, the Assessment Group used a threshold analysis to estimate the cost effectiveness of routine scanning as compared with the standard diagnostic strategy of selective scanning contingent on clinical findings suggestive of an underlying structural cause of first-episode psychosis. A threshold analysis predicts the quality-adjusted life year (QALY) gain required for a technology to be regarded as cost effective. By combining the incremental cost of routine scanning with cost-effectiveness thresholds of £20,000 and £30,000 per QALY, the QALY gains needed to make routine scanning cost effective (or the QALY losses that could be tolerated if the strategy is cost saving) are estimated. A 12-month time horizon was assumed in the Assessment Group's threshold analysis. It was assumed that people considered to have functional psychoses will receive a predefined sequence of atypical antipsychotic medications.
4.2.2 The Assessment Group noted that some organic causes of psychosis cannot be diagnosed using MRI or CT scans. The Assessment Group's threshold analysis therefore considered the case of an organic psychosis caused by a brain tumour or cyst diagnosed after routine or selective scanning. The threshold analysis assumed that treatment of a brain tumour was not altered as a result of earlier detection with an MRI or CT scan. The analysis also assumed no deterioration in disease state when detected at a later stage with selective scanning compared with early-stage detection with routine scanning.
4.2.3 The cost of treatment for a brain tumour or cyst is common to both the routine and selective scanning strategies (using MRI or CT), because it was assumed that, even with selective scanning, diagnosis (and subsequent treatment) of a brain tumour or cyst would be achieved within the 12-month time horizon of the threshold analysis. It was assumed that patients' response to antipsychotic medications is monitored over an 8-week period. The costs associated with this monitoring phase were determined by a proportional split of people receiving either hospital care or home care. The Assessment Group estimated test accuracy rates for detecting brain tumours or cysts to be 100% for MRI and above 90% for CT scans. It was assumed that the prevalence of brain tumours or cysts in a population of people with psychosis was 5%. This was based on MRI scanning having a sensitivity rate at or close to 100%. Also, the probability of detecting a brain tumour or cyst after an MRI scan was estimated to be 5% based on the Assessment Group's review of the evidence from studies that reported scans affecting clinical management.
4.2.4 The base-case threshold analysis incorporating the above assumptions found that the strategy of routine scanning with MRI was cost saving. These cost savings were sensitive to the following assumptions: the time period during which a brain tumour or cyst is undetected and antipsychotic medications are provided under selective scanning; the dosage and costs of antipsychotic medications; and the proportional split of people receiving hospital and home care during the monitoring phase. The greatest cost saving was apparent when the largest proportion of people were hospitalised during the monitoring phase. A 50/50 split between hospital and home care had the largest impact on incremental costs. Under a conservative assumption that no people were hospitalised (0/100 split), routine structural neuroimaging using MRI was still cost saving.
4.2.5 At a threshold value for willingness to pay for an additional QALY of £20,000, and under the conservative scenario of a 0/100 split in hospital/home care, a QALY loss of 0.011 for the full cohort and 0.228 for people with brain tumours or cysts only is needed to offset cost savings. The Assessment Group stated that, under its base-case assumptions, QALY losses needed to render routine MRI scanning not cost effective would have to be large.
4.2.6 The base-case threshold analysis for CT scanning also showed that the scenario that achieved the greatest cost saving was that with the largest proportion of people receiving hospitalised care. However, even when this proportion was assumed to be zero, the antipsychotic drug dosage was assumed to be low and the duration of antipsychotic treatment was assumed to be only 6 months, a routine scanning strategy remained cost saving. Threshold analysis suggested that the QALY loss (needed to render routine CT scanning not cost effective) is greatest in the scenario where the proportion of hospitalised care is largest (50%), the dose of antipsychotics is highest, and the duration of antipsychotic treatment is 12 months under selective scanning and for people with false negative routine CT scans. Under a conservative assumption of no hospitalised care, the QALY loss needed to render routine CT scanning not cost effective would have to be large, if the base-case assumptions regarding the probability of detecting a brain tumour or cyst after a scan are correct.
4.2.7 The Assessment Group conducted a number of sensitivity analyses, one of which varied the prevalence rate of brain tumours or cysts to 0.5% and 1%, respectively. The results of this sensitivity analysis showed that for MRI routine scanning was no longer cost saving at these prevalence rates. Therefore, for MRI to be cost effective, a QALY gain would be needed. Under all scenarios (duration of untreated psychosis, hospital and home care split, dose of antipsychotic medications), the maximum QALY gain needed to make MRI cost effective at an incremental cost-effectiveness threshold of £30,000 per QALY was small: 0.007 and 0.005 for the full cohort at 0.5% and 1% prevalences of brain tumours or cysts, respectively. At an incremental cost-effectiveness threshold of £20,000 per QALY, the corresponding maximum QALY gains were 0.010 and 0.007 for the full cohort at 0.5% and 1% prevalences of brain tumours or cysts, respectively.
4.2.8 When the prevalence rate of brain tumours or cysts was set at 0.5% and hospital care was given in 20% of cases or fewer, routine scanning was no longer cost saving and a QALY gain was needed to make CT scanning cost effective at conventional thresholds. For all scenarios with a 50/50 split of hospital/home care, routine CT scanning was cost saving. When prevalence was set to 1%, routine CT scanning was cost saving under all scenarios.
4.2.9 The analyses carried out by the Assessment Group suggest that routine structural neuroimaging would be cost saving if the base‑case assumptions regarding the probability of detecting a brain tumour or cyst after a scan are plausible. The maximum acceptable QALY loss for MRI to be cost effective ranged from 0.011 to 0.039, and for CT the maximum acceptable QALY loss ranged from 0.017 to 0.043. These results appear robust to variations in the various parameters investigated, except for variations in the prevalence rates of brain tumours or cysts in people with psychosis.
4.2.10 In conclusion, the threshold analysis showed that, if the prevalence of organic psychosis due to a brain tumour or cyst lies in the region of 5%, then, under the Assessment Group's assumptions, routine structural neuroimaging is cost saving. If the prevalence of organic psychoses is close to 0.5%, then, under the Assessment Group's assumptions, MRI is no longer cost saving, and CT is only cost saving if 50% of people receive hospital care. However, evidence for determining the true prevalence of treatable lesions in the population under test is extremely limited.
4.3.1 The Appraisal Committee reviewed the data available on the clinical and cost effectiveness of structural neuroimaging (using MRI or CT scanning) in first-episode psychosis, having considered evidence on the nature of the condition and the value placed on the benefits of structural neuroimaging by clinical specialists. It was also mindful of the need to take account of the effective use of NHS resources.
4.3.2 The Committee agreed that, because structural abnormalities in the brain progress over time, for people with a first episode of psychosis without signs or symptoms of additional pathology, the early positive detection and management of structural lesions after routine scanning could have health benefits where a treatable cause is found. The Committee expressed concern about whether it would be feasible to scan people who were particularly disturbed when they presented with acute psychosis. The Committee was reassured by the clinical specialist that some people may be more willing to undergo a neuroimaging scan than to have thorough clinical examinations, viewing it as less intrusive.
4.3.3 The Committee concluded from the evidence presented that there was substantial uncertainty about the true prevalence of structural lesions in the population under test. The Committee heard from the clinical specialist that the assumption of a 5% prevalence of organic psychosis may be an underestimate and that the figure could be as high as 10% because the studies reported had excluded people with any clinical sign of neurological abnormalities, which would reduce the likelihood of including people with psychosis of an organic cause in the study population. However, the clinical specialist considered that the figure for organic psychosis due specifically to a brain tumour or cyst may be less than 5%. The Committee noted that the prevalence estimate of 5% for organic psychosis due to a brain tumour or cyst was based on the results from studies of varying methodological quality and internal validity, and agreed that this estimate could not be relied on. The Committee further considered that incidental findings and false positives associated with neuroimaging may increase the anxiety levels of people with psychosis, leading to additional investigations and treatments, with questionable returns in terms of improved health outcomes from clinical care.
4.3.4 The Committee considered the evidence presented on the cost effectiveness of routine structural neuroimaging in first-episode psychosis. It discussed the tentative results of the Assessment Group's threshold analysis, which suggested that neuroimaging may be cost saving in a number of scenarios. The Committee noted that one limitation of the Assessment Group's threshold analysis is the uncertainty surrounding estimates of the prevalence of brain tumours or cysts in people with first-episode psychosis. It also noted that the analysis did not capture potential costs associated with false positives, the need for repeat investigations and subsequent treatments, and potential health benefits and losses. The Committee considered that, although the Assessment Group's approach was appropriate given the lack of data, substantial uncertainties existed about key parameters in the threshold analysis – in particular, the estimates of the prevalence rates of brain tumours or cysts in the population of people with first-episode psychosis.
4.3.5 The Committee further discussed the assumption in the Assessment Group's approach that people in whom structural lesions were identified by neuroimaging could discontinue antipsychotic medication and thereby eliminate subsequent costs for these drugs. The Committee heard from the clinical specialist that this may not routinely be the case if the lesion is not treatable and the psychotic symptoms persist. The Committee was aware that effects on mortality had not been considered in the threshold analysis and that the analysis did not consider possible deterioration in the underlying organic conditions as a result of late detection and diagnosis under selective scanning. The Committee concluded that a reliable estimate of the cost effectiveness of routine structural neuroimaging could not be made given the limitations on the data available.
4.3.6 On balance, the Committee agreed that, although routine scanning could have potential benefits from early detection of structural causes of first-episode psychosis, the current evidence base, particularly in relation to the prevalence of treatable lesions in the population under examination, was too weak to support a decision to implement routine use of MRI or CT scanning in people with first‑episode psychosis. The Committee agreed that this decision should not affect the current practice of using structural neuroimaging techniques selectively to exclude organic causes of psychosis where people's symptoms, or other aspects of their presentation, suggest a higher likelihood of an underlying organic cause.
4.3.7 The Committee considered that the limited evidence base to support routine scanning using structural neuroimaging techniques made it difficult to determine the clinical and cost effectiveness of routine structural neuroimaging versus selective scanning. The Committee concluded therefore that the use of structural neuroimaging techniques (either MRI or CT scanning) should not be recommended as a routine part of the initial investigations for the management of first-episode psychosis.