5.1
The diagnostics advisory committee (section 10) considered evidence from a number of sources.
Guidance
5 Outcomes
How outcomes were assessed
5.2
A systematic review of the effectiveness of contrast-enhanced ultrasound using SonoVue compared with contrast-enhanced CT and contrast-enhanced MRI was undertaken by the external assessment group. The outcome measures included:
-
the effect of testing on the treatment plan (for example, surgical or medical management, or palliative care)
-
the effect of pre-treatment testing on clinical outcome (for example, overall survival, progression-free survival)
-
prognosis – the ability of the test result to predict clinical outcome (for example, overall survival, progression-free survival, response to treatment)
-
test accuracy and number of people or lesions for which no conclusive diagnostic information could be obtained with contrast-enhanced ultrasound using SonoVue.
5.3
Radiation exposure was not considered a relevant outcome because the population is mostly older adults in whom additional incident cancers as a result of imaging are likely to be minimal.
5.4
A systematic review of the evidence on cost effectiveness for SonoVue was undertaken by the external assessment group. The external assessment group constructed multiple de novo models. The outcomes of interest for the modelling were costs and the morbidity and mortality associated with the investigation and characterisation of focal liver lesions and their treatment. These included survival and health-related quality of life, including the impact of adverse events associated with treatment (such as chemotherapy).
5.5
Diagnostic technologies themselves do not usually have direct evidence for health-related quality of life, and the de novo models therefore followed a linked evidence approach in which intermediate outcomes (results of the test/s) were linked to the care pathway to estimate clinical outcomes and hence quality-adjusted life year (QALY) gains. Costs and QALYs were assigned to SonoVue and the comparators.
Clinical effectiveness
5.6
A total of 17 studies in 18 publications were included in the assessment. All of the included studies were test accuracy studies:
-
7 concerned the use of contrast-enhanced ultrasound with SonoVue for characterising focal liver lesions identified during routine monitoring of people with cirrhosis
-
4 assessed the performance of contrast-enhanced ultrasound with SonoVue for investigating potential liver metastases in people with known primary cancers (mostly colorectal cancer)
-
6 concerned the use of contrast-enhanced ultrasound with SonoVue for characterising incidentally detected focal liver lesions.
5.7
Only 1 of the studies of test accuracy included in this assessment reported information on adverse events related to testing. In this study there were no adverse events associated with contrast-enhanced ultrasound with SonoVue. There was no information about the comparator (contrast-enhanced MRI with gadolinium). A large, retrospective safety study of contrast-enhanced ultrasound with SonoVue in abdominal imaging did not meet the inclusion criteria for this assessment but reported data from 23,188 investigations in 29 centres in Italy. This study found 29 incidents of adverse events, of which 2 were graded as serious, 1 as severe, 3 as moderate and 23 as mild. There were no fatal adverse events. Most non-serious adverse events resolved without intervention.
5.8
All included studies were published in 2006 or later. Most were conducted in Europe (most in Italy or Spain). Two studies reported funding from the manufacturer of SonoVue.
Results by clinical indication
5.9
Test accuracy data in relation to each clinical indication assessed are summarised below.
Characterising focal liver lesions identified during monitoring of people with cirrhosis
5.10
Studies conducted in people with cirrhosis during routine monitoring all concerned the differentiation of hepatocellular carcinoma from other lesion types in small to medium (less than 30 mm) focal liver lesions. The definition of a positive test for hepatocellular carcinoma varied across studies. Studies assessing contrast-enhanced MRI used 3 contrast agents: gadolinium (a vascular contrast agent), SPIO (a hepatocyte-specific contrast agent) and Gd-EOB-DTPA (a 'combined' vascular and hepatocyte-specific contrast agent). There was no consistent evidence for any significant difference in test performance between the 3 imaging modalities (contrast-enhanced ultrasound, contrast-enhanced CT and contrast-enhanced MRI) and the 3 MRI contrast media assessed. When the definition of hepatocellular carcinoma given in the EFSUMB guidelines (arterial phase enhancement followed by portal-venous washout) was used, estimates of the sensitivity and specificity of each of the imaging modalities varied across studies. There was some evidence, from 1 study comparing contrast-enhanced ultrasound and contrast-enhanced MRI using gadolinium, that these imaging techniques may be better at ruling out hepatocellular carcinoma in focal liver lesions between 11 and 30 mm (sensitivities were 92% and 95% respectively) than in small focal liver lesions 10 mm or less (sensitivities 27% and 73% respectively). However, this study did not use the EFSUMB definition of hepatocellular carcinoma. It is therefore possible that some of the variation in sensitivity estimates in studies of focal liver lesions smaller than 30 mm may be a result of differences in the size distribution of focal liver lesions included. The evidence suggested that contrast-enhanced ultrasound alone may be adequate to rule out hepatocellular carcinoma for focal liver lesions between 11 mm and 30 mm.
Investigating potential liver metastases in people with known primary malignancy
5.11
Studies of the diagnosis of liver metastases using imaging with vascular contrast media (contrast-enhanced ultrasound, contrast-enhanced CT and contrast-enhanced MRI with gadolinium), in which definitions of a positive imaging test were reported, gave various descriptions of peripheral rim enhancement as the criteria for liver metastases. Two studies also reported data for contrast-enhanced MRI with SPIO. There was no evidence of any consistent difference in test performance between the 3 imaging modalities and the different contrast media assessed. Per patient sensitivity estimates, from 2 studies, were generally high: 83% for all imaging modalities and both MRI contrast agents in 1 study of people with colorectal cancer and more than 95% for both contrast-enhanced ultrasound and contrast-enhanced CT in a second study of people with various primary cancers (mostly colorectal cancer). The only previous systematic review of contrast-enhanced ultrasound with SonoVue for the diagnosis of liver metastases did not include any comparator tests and reported sensitivities ranging from 79% to 100%. The limited data available indicate that contrast-enhanced ultrasound alone may be adequate to rule out liver metastases in people with known primary malignancies.
Characterising incidentally detected focal liver lesions
5.12
The primary outcome measure reported by studies conducted in people with incidentally detected focal liver lesions was test accuracy for the differentiation of malignant from benign liver lesions. Studies consistently used definitions of the imaging criteria for hepatocellular carcinoma and liver metastases which were similar to those reported in the EFSUMB guidelines on the use of contrast-enhanced ultrasound. All studies reported no significant difference in the accuracy of contrast-enhanced ultrasound, contrast-enhanced CT and contrast-enhanced MRI for characterising focal liver lesions. The pooled estimates of sensitivity for the detection of 'any liver malignancy' were approximately 95% for both contrast-enhanced ultrasound and contrast-enhanced CT. The pooled estimates of specificity were 94% and 93% respectively, based on data from 4 studies. The single study comparing contrast-enhanced ultrasound with contrast-enhanced MRI used gadolinium for MRI in all people, with the addition of SPIO in an unspecified number. This study reported sensitivity estimates of 91% and 82% respectively, and corresponding specificity estimates of 67% and 63%. Data from 1 study indicated that combined imaging using both contrast-enhanced ultrasound and contrast-enhanced CT did not increase sensitivity when a positive result on either modality was treated as 'test positive'. This, combined with the high estimates of sensitivity, indicates that contrast-enhanced ultrasound alone may be adequate to rule out liver malignancy in people with incidentally detected focal liver lesions.
Economic analysis
5.13
Four studies were identified that met the inclusion criteria for an economic analysis of the use of SonoVue in contrast-enhanced ultrasound.
5.14
Although all the studies were of reasonably good quality, they did not fully address the cost effectiveness of SonoVue as defined in this assessment. Limitations included restricted information about disease management and progression, choice of equipment and administrative procedures in different settings, inclusion of costing elements in the calculation, and health outcomes. Zaim et al. (2011) was the only study that modelled disease management and reported health outcomes relevant to this assessment, but the follow-up was only 24 months.
5.15
The external assessment group performed a de novo analysis to address specifically the decision problem for this evaluation and to estimate the cost effectiveness of SonoVue in England.
5.16
The external assessment group conducted an analysis of contrast-enhanced ultrasound using SonoVue for assessing focal liver lesions in adults, in whom unenhanced ultrasound or other liver imaging has been inconclusive. Three separate models were used for 3 clinical applications for which the most data on test performance were available and experts suggested there was most likely to be clinical benefit:
-
cirrhosis surveillance (for characterising focal liver lesions identified through monitoring of people with cirrhosis)
-
investigating potential liver metastases in colorectal cancer
-
characterising incidentally detected focal liver lesions.
5.17
In each model, contrast-enhanced ultrasound with SonoVue was compared with contrast-enhanced CT, contrast-enhanced MRI using gadolinium and/or contrast-enhanced MRI using SPIO. Average costs, expected life years and expected QALYs per person were calculated for the above technologies.
Costs
5.18
Costs of contrast-enhanced and unenhanced ultrasound were informed by expert clinical opinion and cost information provided the manufacturer of SonoVue. The costs of using the contrast agent, including cannulation, were assumed to be £48.70 (estimate supplied by the manufacturer and agreed by clinicians). In addition, contrast-enhanced ultrasound was expected to take longer than unenhanced ultrasound. Therefore, the external assessment group used the difference between the reference costs of an ultrasound taking less than 20 minutes (£55) and an ultrasound taking more than 20 minutes (£71) as the additional time costs of contrast-enhanced ultrasound. The total additional cost was therefore estimated to be £65. This assumed that contrast-enhanced ultrasound is performed in the same appointment as the unenhanced scan. Costs of the other diagnostic tests were based on 2011 NHS reference costs.
Cost effectiveness
5.19
A model description, test accuracy data and results of the base-case and additional analyses are provided below for each of the 3 models.
Cirrhosis surveillance model
Model description
5.20
The model was a modified version of a model produced by the Peninsula Technology Assessment Group (the PenTAG cirrhosis surveillance model). The population consisted of people with a diagnosis of compensated cirrhosis entering a surveillance programme (aged 70 years or younger with no pre-existing medical conditions that would preclude treatment with liver transplant or hepatic resection [including current alcohol or intravenous drug misuse]). The time horizon was a lifetime and the cycle duration was 1 month. Patients in the model can develop hepatocellular carcinoma. In the base-case analysis monitoring takes place every 6 months, and stops when people reach 70 years.
Test accuracy data used in the model
5.21
It was assumed that the first test used for monitoring was unenhanced ultrasound. The test performance of unenhanced ultrasound used in the model is shown in table 1 and was based on the study by Bennett et al. (2002) as used in the health technology assessment report by Thompson Coon et al. (2007). This study was preferred over other studies because it distinguished between small, medium and large tumours, and had a reasonable sample size (n=200).
|
Tumour size |
Sensitivity |
|---|---|
|
Small |
0.11 |
|
Medium |
0.29 |
|
Large |
0.75 |
5.22
Additional imaging takes place when unenhanced ultrasound is inconclusive. About 43% of unenhanced ultrasounds were estimated to be inconclusive, based on information provided by the manufacturer of SonoVue. In the base-case analysis, the probability of identifying hepatocellular carcinoma and the proportion of people with a false-positive test result were taken from Leoni et al. (2010). Data from this study were used because diagnostic criteria matched the EFSUMB guidance on the use of contrast-enhanced ultrasound and the performance of contrast-enhanced ultrasound, contrast-enhanced CT and contrast-enhanced MRI with gadolinium was reported in the same population. The study included people with liver lesions between 1 cm and 3 cm. In the base-case analysis the external assessment group used these results to model the diagnostic accuracy for both small (less than 2 cm) and medium (2 cm to 5 cm) tumours (table 2). The sensitivity for identifying large hepatocellular carcinomas was assumed to be 100% for all confirmatory imaging tests, and this assumption was agreed by the clinical experts.
|
Test |
Sensitivity for identifying small and medium tumours |
False-positive rates |
|---|---|---|
|
Contrast-enhanced ultrasound |
0.67 |
0.03 |
|
Contrast-enhanced CT |
0.67 |
0.03 |
|
Contrast-enhanced MRI with gadolinium |
0.82 |
0.01 |
Base-case cost effectiveness results
5.23
Contrast-enhanced ultrasound had the lowest discounted lifetime costs per person (£35,744), followed by contrast-enhanced CT (£36,124) and contrast-enhanced MRI with gadolinium (£36,807). Compared with contrast-enhanced ultrasound, contrast-enhanced CT was as effective but more costly, and was thus considered to be dominated by contrast-enhanced ultrasound. Contrast-enhanced MRI with gadolinium cost £1,063 more per person than contrast-enhanced ultrasound, but also yielded 0.022 more QALYs, giving an incremental cost-effectiveness ratio (ICER) of £48,454 per QALY gained. Contrast-enhanced ultrasound is more cost-effective than contrast-enhanced MRI at £20,000 per QALY gained because although less effective it costs less and the ICER for contrast-enhanced MRI compared with contrast-enhanced ultrasound is above this value.
Additional analyses
5.24
A range of additional analyses were performed by the external assessment group. Compared with contrast-enhanced MRI with gadolinium (and contrast-enhanced CT), contrast-enhanced ultrasound was the most cost effective option in many of the additional analyses, except when it was assumed that all positive (true and false) unenhanced ultrasound examinations were subject to confirmatory testing instead of only the inconclusive results, and when the proportion of people estimated to have an inconclusive unenhanced ultrasound was considerably lower (20% instead of 43%). These 2 analyses resulted in ICERs for contrast-enhanced MRI with gadolinium compared with contrast-enhanced ultrasound of £12,806 and £16,121 per QALY gained respectively (contrast-enhanced CT was dominated by contrast-enhanced ultrasound in both cases).
5.25
In probabilistic sensitivity analysis with over 5,000 replications, at £20,000 per QALY gained the probability that contrast-enhanced ultrasound, contrast-enhanced CT or contrast-enhanced MRI with gadolinium was most cost effective was 99%, 0% and 1% respectively.
Investigating potential liver metastases from colorectal cancer model
Model description
5.26
The model was a modified version of the model developed by Brush et al. (2011). The population consisted of people who had previously had surgery for primary colorectal cancer and who, during routine follow-up, were identified as potentially having a metastatic recurrence. The time horizon was a lifetime and the cycle duration was 1 year.
Test accuracy data used in the model
5.27
The test performance used in the base case was from Mainenti et al. (2010) because this study compared all 3 alternative tests (contrast-enhanced CT, contrast-enhanced MRI with gadolinium, contrast-enhanced MRI with SPIO) with contrast-enhanced ultrasound (table 3).
|
Test |
Sensitivity |
Specificity |
|---|---|---|
|
Contrast-enhanced ultrasound |
0.83 |
0.86 |
|
Contrast-enhanced CT |
0.83 |
0.96 |
|
Contrast-enhanced MRI with gadolinium |
0.83 |
0.96 |
|
Contrast-enhanced MRI with SPIO |
0.83 |
1.00 |
Base-case cost effectiveness results
5.28
In the base-case analysis, using the different imaging techniques to investigate potential liver metastases from colorectal cancer resulted in equal expected lifetime QALYs (8.364). Contrast-enhanced ultrasound and contrast-enhanced CT were the least costly tests, with expected lifetime costs of approximately £7,510 per person. Contrast-enhanced MRI with gadolinium (£7,688) and contrast-enhanced MRI with SPIO (£7,722) were both more costly than, and thus dominated by, contrast-enhanced CT and contrast-enhanced ultrasound. Contrast-enhanced ultrasound and contrast-enhanced CT were cost-effective technologies, with equal expected costs and effectiveness.
Additional analyses
5.29
A range of additional analyses were performed by the external assessment group. Analyses that had an impact on the results of the base-case analysis are summarised here. In the base-case analysis it was assumed that people who were incorrectly diagnosed with liver metastases would have a biopsy and the incorrect diagnosis would be discovered before treatment. If this is not assumed, and people could receive unnecessary treatment, the lower specificity of contrast-enhanced ultrasound had larger consequences. This led to contrast-enhanced ultrasound being both the most costly and the least effective option, and contrast-enhanced MRI with gadolinium dominating all other tests. When alternative sources of test performance were used, from Jones et al. (2011) and Clevert et al. (2009), contrast-enhanced ultrasound was the cost-effective option in both scenarios.
5.30
In probabilistic sensitivity analysis with 5,000 replications, at £20,000 per QALY gained contrast-enhanced CT had the highest probability of being cost effective (48%), followed by contrast-enhanced ultrasound (47%), contrast-enhanced MRI with gadolinium (3%) and contrast-enhanced MRI with SPIO (2%).
Characterising incidentally detected focal liver lesions
Model description
5.31
People with incidentally detected focal liver lesions can have a variety of conditions, ranging from malignant lesions such as hepatocellular carcinoma and metastases to different types of benign lesions. The prognosis and costs for people diagnosed with hepatocellular carcinoma were modelled using the cirrhosis surveillance model, whereas the prognosis and costs for people with liver metastases were modelled using the liver metastases model. The model took a lifetime time horizon. The costs, life years and QALYs for people with a malignancy other than hepatocellular carcinoma or metastases were assumed to be equal to those in people with hepatocellular carcinoma. However, it was known in advance of the modelling that the costs and QALYs for these people would have a limited effect on the cost effectiveness of contrast-enhanced ultrasound, because its sensitivity was very similar to that of the comparators and the prior probability of other malignancies was small.
Test accuracy data used in the model
5.32
The approach used in the base case was to take the results from the meta-analysis of 4 studies that compared contrast-enhanced ultrasound with contrast-enhanced CT for the differentiation of malignant and benign lesions. Table 4 illustrates the similar performance of the 2 tests.
|
- |
95% confidence interval (exact method) |
|
|---|---|---|
|
Sensitivity of contrast-enhanced ultrasound |
95.1% |
93.3% to 96.6% |
|
Sensitivity of contrast-enhanced CT |
94.6% |
92.7% to 96.1% |
|
Specificity of contrast-enhanced ultrasound |
90.4% to 96.3% |
|
|
Specificity of contrast-enhanced CT |
93.1% |
89.6% to 95.8% |
5.33
Only 1 study, Seitz (2010), compared the test accuracy of contrast-enhanced ultrasound with MRI (a sensitivity of 77.3% and 63.6% and a specificity of 75.0 and 76.7, respectively, were used in the base case). This study reported that all people in a subgroup had contrast-enhanced MRI with gadolinium, and that a subset of these people also had MRI with a SPIO contrast agent. It was difficult to determine the different accuracies of MRI with the 2 different contrast agents from the study, and therefore sections relating to the use of MRI in the characterisation of incidentally detected focal liver lesions refer to contrast-enhanced MRI overall.
Base-case cost-effectiveness results
5.34
The lower costs of contrast-enhanced ultrasound combined with slightly better test performance meant that contrast-enhanced ultrasound dominated both contrast-enhanced CT (contrast-enhanced ultrasound cost £52 less and yielded 0.0002 additional QALYs) and contrast-enhanced MRI (contrast-enhanced ultrasound cost £131 less and yielded 0.0026 additional QALYs).
Additional analyses
5.35
A range of additional analyses were performed by the external assessment group. Although these analyses changed the absolute costs and effectiveness of the different strategies, they did not lead to any significant changes in the incremental costs and effectiveness of contrast-enhanced ultrasound compared with contrast-enhanced CT or contrast-enhanced MRI. The cost of the tests was the most critical factor in the analyses. The impact of other factors (for example, prior probabilities of a particular diagnosis and costs of treatment) was minimal because the accuracy of the tests was so similar.
5.36
Probabilistic sensitivity analyses showed that the probability of contrast-enhanced ultrasound being cost effective compared with contrast-enhanced CT and contrast-enhanced MRI was greater than 95% at £20,000 per QALY gained.