4 Evidence

The diagnostics advisory committee (section 8) considered evidence on the integrated multiplex polymerase chain reaction (PCR) tests for identifying gastrointestinal pathogens in people with suspected gastroenteritis from several sources (section 9). Full details of all the evidence are in the committee papers.

Clinical effectiveness

4.1 In total, 23 studies reported data for the xTAG Gastrointestinal Pathogen Panel and the FilmArray GI Panel, all of which were observational studies. One study was retrospective and used stored samples, and the remaining 22 were prospective. No data were available for the Faecal Pathogens B assay. All studies reported diagnostic accuracy and other intermediate outcome data; no studies reported clinical end outcomes.

4.2 Most studies did not differentiate between people in hospital and people in the community, and it was assumed that they were a mixed population. Eleven studies included people from European countries, with 3 of these studies done in the UK. The studies were often poorly reported and used an inadequate reference standard. In most studies, it was unclear how participants were selected. In 10 of the studies, there was concern about the applicability of the population included in the studies to the decision question, and all 15 studies that included a reference standard were rated as high risk for applicability concerns relating to the reference standard.

Diagnostic accuracy

4.3 Diagnostic accuracy data were presented as positive and negative agreement and were included in exploratory random-effects meta-analyses. Positive and negative agreement data were meta-analysed and reported for both the intervention compared with the comparator and the comparator compared with the intervention.

xTAG Gastrointestinal Pathogen Panel

4.4 The accuracy of the xTAG Gastrointestinal Pathogen Panel was reported in 19 studies, 2 of which also included the FilmArray GI Panel. The number of pathogens on the xTAG test panels varied: 15 pathogens in 16 studies; 11 pathogens in 2 studies; and 14 pathogens in 1 study. The test is currently marketed as containing 15 pathogens.

4.5 Only 8 studies included enough information for inclusion in the meta-analyses. Two of these studies, both of which used the 15‑pathogen version of the xTAG Gastrointestinal Pathogen Panel, were based in the UK; Halligan et al. (2014) was the largest study with 2,187 samples and Pankhurst et al. (2014) included 839 samples.

4.6 The results of the exploratory meta-analyses for the xTAG Gastrointestinal Pathogen Panel are summarised in table 4. There was substantial heterogeneity in the pooled summary estimates. The overall positive and negative agreement between xTAG compared with conventional testing is high but positive agreement falls when conventional testing is compared with the xTAG Gastrointestinal Pathogen Panel. This is because a large proportion of samples that are positive with the xTAG Gastrointestinal Pathogen Panel are confirmed by positive testing, but when the xTAG Gastrointestinal Pathogen Panel is used as the benchmark, conventional testing appears to report fewer positive results. This suggests that the xTAG Gastrointestinal Pathogen Panel detects more pathogens than conventional testing, but the clinical significance of the increased diagnostic yield is unknown.

Table 4 Accuracy estimates for xTAG Gastrointestinal Pathogen Panel

Comparison

Positive agreement

Negative agreement

I 2

Pooled result (95% CI)

p value

I 2

Pooled result (95% CI)

p value

xTAG GPP versus conventional testing

83%

0.929

(0.898 to 0.955)

<0.001

95%

0.982

(0.976 to 0.988)

<0.001

Conventional testing versus xTAG GPP

97%

0.678

(0.580 to 0.770)

<0.001

77%

0.998

(0.997 to 0.999)

<0.001

Abbreviations: CI, confidence interval; I2, Cochran Q statistic; xTAG GPP, xTAG Gastrointestinal Pathogen Panel.

FilmArray GI Panel

4.7 The accuracy of the FilmArray GI Panel was reported in 6 studies, 2 of which also included the xTAG Gastrointestinal Pathogen Panel. Five studies reported a version of the FilmArray GI Panel that contained 23 pathogens and 1 study reported a version with 22 pathogens. The test is currently marketed as containing 22 pathogens.

4.8 Two US‑based studies were included in the meta-analyses. The results of the exploratory meta-analyses for the FilmArray GI Panel are summarised in table 5. There was substantial heterogeneity in the pooled summary estimates. As with the xTAG Gastrointestinal Pathogen Panel meta-analyses, the results suggest that the FilmArray GI Panel detects more pathogens than conventional testing, but the clinical significance of the increased diagnostic yield is unknown. The increased diagnostic yield of the FilmArray GI Panel is slightly less than that for the xTAG Gastrointestinal Pathogen Panel (1.5 compared with 1.2 times more pathology); but only 2 studies contributed data to the meta-analyses for this test.

Table 5 Accuracy estimates for the FilmArray GI Panel

Comparison

Positive agreement

Negative agreement

I 2

Pooled result (95% CI)

p value

I 2

Pooled result (95% CI)

p value

FilmArray versus conventional testing

89%

0.954

(0.897 to 0.991)

<0.001

88%

0.996

(0.993 to 0.998)

<0.001

Conventional testing versus FilmArray

81%

0.820

(0.761 to 0.872)

<0.001

80%

1.000

(0.999 to 1.000)

<0.001

Abbreviations: CI, confidence interval; I2, Cochran Q statistic.

xTAG Gastrointestinal Pathogen Panel compared with the FilmArray GI Panel

4.9 Two studies compared the xTAG Gastrointestinal Pathogen Panel with the FilmArray GI Panel. Khare et al. (2014) reported that both tests detect more pathogens than conventional testing, whereas Gu et al. (2015) reported that both assays detected a similar number of pathogens.

Causes of additional positive results

4.10 Nine studies commented on the possible causes of additional positive results from either the new technologies or conventional testing; that is, they reported whether additional positives associated with the new technologies arose because of these technologies having a broader coverage of pathogens than the comparator, or because the new tests are more sensitive. All 9 studies reported data for the xTAG Gastrointestinal Pathogen Panel and the FilmArray GI Panel. In 8 of the 9 studies, additional positive results were reported to have arisen because of both improved sensitivity and greater coverage. A further 3 studies reported that additional positive results with conventional testing occurred because pathogens or serotypes of pathogens not included in the new technologies were detected. Although the new technologies seem to have an increased detection rate, the absence of an accurate reference standard results in uncertainty about the clinical significance of the additional positive results.

Analysis of discordant results

4.11 Five studies reported the verification of discordant results, 1 of which reported results for the FilmArray GI Panel and the remaining 4 for the xTAG Gastrointestinal Pathogen Panel. In general, the results of the discordant analyses reported in favour of the new tests. However, PCR methods were often used as the verification test, which are likely to verify in favour of the new tests because they use the same analytical approach, particularly for bacterial pathogens. An important area of uncertainty is the clinical significance of detecting bacterial DNA compared with detecting viable bacteria with culture.

Test-failure rates

4.12 Eight studies reported test-failure rates; 2 for the FilmArray GI Panel, 5 for the xTAG Gastrointestinal Pathogen Panel, and 1 for both tests. Gu et al. (2015) reported results for both the xTAG Gastrointestinal Pathogen Panel, which had a failure rate of 5%, and the FilmArray GI Panel, which had no failures. In the 2 other studies reporting results for the FilmArray GI Panel, the test-failure rate ranged from 0.8% to 5.1%. Of the 5 studies reporting results for the xTAG Gastrointestinal Pathogen Panel, 2 reported test-failure rates per pathogen, which ranged from 7.8% to 17.12%, and 3 reported aggregated test-failure rates, that is, for all targets in the panel, which ranged from 2.1% to 19%.

Test-turnaround times

4.13 Laboratory turnaround times for the integrated multiplex PCR tests were reported in 8 studies, 1 of which (Spina et al. 2015) reported the time between sampling and testing with the FilmArray GI Panel only. Two studies, Gu et al. (2015) and Khare et al. (2014), reported results for both the xTAG Gastrointestinal Pathogen Panel and the FilmArray GI Panel. Both studies reported the turnaround time for the FilmArray GI Panel as about 1 hour (1 sample per run) and about 6.5 hours per run (multiple samples) for the xTAG Gastrointestinal Panel. A 1‑hour turnaround time for the FilmArray GI panel was also reported by Buss et al. (2015).

4.14 Two further studies compared the turnaround time of the xTAG Gastrointestinal Pathogen Panel with conventional methods. A UK‑based study, Halligan et al. (2014), reported a laboratory turnaround time with the xTAG Gastrointestinal Pathogen Panel of 26.6 hours for an afternoon run compared with 10.4 hours for a morning run. When sample collection and transport time was included, the median turnaround time was 41.8 hours for an afternoon run with the xTAG Gastrointestinal Pathogen Panel compared with conventional methods, which ranged from 17.3 hours (Clostridium difficile testing) to 66.5 hours (bacterial culture). A study based in Germany, Kahlau et al. (2013), reported a median turnaround time of 1 day for the xTAG Gastrointestinal Pathogen Panel compared with 3 days for conventional methods (p=0.0000021). A study based in China, Deng et al. (2015), reported a turnaround time of 5 hours for the xTAG Gastrointestinal Pathogen Panel. A further study, based in the US, by Patel et al. (2014) reported the duration of laboratory technician hands‑on time to be 2.5 hours for the xTAG Gastrointestinal Pathogen Panel compared with 10 hours for conventional testing and a time to detection of 5 hours for the xTAG Gastrointestinal Pathogen Panel compared with 72 hours for conventional testing.

Frequency of detecting multiple pathogens in a sample

4.15 All 23 studies reported data on the frequency of the tests detecting multiple pathogens in a sample, which ranged from 4% to 58%. The proportion of samples with multiple pathogens detected in the 3 UK‑based studies ranged from 4% to 8%, and most samples contained 2 pathogens; the most common pathogens detected in combination were norovirus, Clostridium difficile, rotavirus, Campylobacter, Salmonella and Shigella.

Use of isolation facilities

4.16 Use of isolation facilities was reported in 2 studies. The UK‑based Halligan et al. (2014) study on using the xTAG Gastrointestinal Pathogen Panel, reported that in a hospital inpatient population a greater proportion of people with community-acquired infections were isolated compared with people with hospital-acquired infections (69% compared with 52.1%), but more people with community-acquired infections were subsequently removed from isolation (60.1% compared with 41.6%; p<0.01). The reasons for this difference are not clear. The median time that people with a community-acquired infection spent in isolation ranged from 1 day with either adenovirus or Entamoeba histolytica to 4 days with Clostridium difficile or Escherichia coli O157. For people with hospital-acquired infection, the median time spent in isolation ranged from 0 days with Salmonella to 13.5 days with Cryptosporidium. The most common reasons for not removing people from isolation included the person being colonised or infected with multi-drug-resistant organisms, being immunocompromised, or having a respiratory viral infection.

4.17 A US‑based study, Rand et al. (2015), concluded that 24.6% (25/102) of people who had negative test results with the xTAG Gastrointestinal Pathogen Panel could have been removed from isolation.

Change in clinical management

4.18 One study, Coste et al. (2013), which was done in France and included retrospective testing with the xTAG gastrointestinal Pathogen Panel and other molecular panel tests, reported that 18% (9/49) of patients had an intestinal endoscopy, and that changes to immunosuppressive therapy (including dose reduction) were seen in 24% (13/54) of episodes of diarrhoea. The authors did not change management recommendations on the basis of molecular panel test results because the study was retrospective, but reported that if the study had been prospective, they could potentially have been changed, with 5 of 9 colonoscopies and 6 of 13 therapy changes in immunosuppressed patients possibly being avoided.

Cost effectiveness

Systematic review of cost effectiveness

4.19 One study met the inclusion criteria (Goldenberg et al. 2015) for the systematic review of existing economic evaluations. The study had limited generalisability because the outcomes reported for the xTAG Gastrointestinal Pathogen Panel were simulated and were not reported as quality-adjusted life years (QALYs).

4.20 The study reported a cost–benefit analysis of the xTAG Gastrointestinal Pathogen Panel compared with conventional laboratory testing and took the perspective of the NHS. The time horizon for the model was the duration of the index episode of suspected infectious gastroenteritis. The economic model was based on a non-randomised, parallel testing study of samples from 800 patients, and assumed that the xTAG Gastrointestinal Pathogen Panel was 100% accurate. The clinical outcomes, which included treatment for the infection, admission to an isolation room, and discharge from hospital, were simulated for the xTAG Gastrointestinal Pathogen Panel arm of the model. This resulted in a higher pathogen-detection rate for patients in isolation using the xTAG Gastrointestinal Pathogen Panel (37.2%) compared with current practice (19.8%) and a reduced time in isolation from 2,202 to 1,447 days. The test costs associated with the xTAG Gastrointestinal Pathogen Panel were £22,283 greater than with conventional testing, but these were offset by savings associated with reduced use of isolation facilities, which led to cost savings of £44,482 compared with current practice. Sensitivity analyses showed that an overall reduction in isolation time of 252 days (based on a simulated population of 800 people) would be needed to make implementing the xTAG Gastrointestinal Pathogen Panel cost neutral.

Modelling approach

4.21 Five de novo economic models designed to explore the cost effectiveness of the xTAG Gastrointestinal Pathogen Panel and the FilmArray GI Panel were developed. Each model had a similar structure but the core inputs were varied to take account of the population included in the model. The analysis took the perspective of the NHS and personal social services. Clinical effectiveness was modelled from diagnostic accuracy data using a linked-evidence approach because no data on clinical end outcomes were identified. A model was developed for each of the following populations:

  • adults in hospital (base-case model)

  • younger children in hospital (Model 2)

  • people in the community (Model 3)

  • people who are immunocompromised and in hospital (Model 4)

  • people with a recent history of foreign travel (Model 5).

Base-case model structure

4.22 A decision tree model was developed, which included testing, isolation and treatment. A decision tree with a time horizon of 2 weeks was chosen because gastroenteritis usually resolves within this time, and so the model did not take into account any adverse events from treatment, persistent complications, readmissions, or mortality.

4.23 The model starts with patients having testing with either conventional methods or 1 of the new technologies (the xTAG Gastrointestinal Pathogen Panel or the FilmArray GI Panel). They are then either isolated or not isolated based on clinical judgement, and their subsequent care is determined by whether or not a pathogen is detected. The model structure was the same for patients having testing with conventional methods or the new technologies, but there were differences in the time to get test results and then whether patients were moved from isolation, treated or discharged earlier.

Model inputs

4.24 For all models, data on the prevalence of pathogens and agreement between the new technologies and conventional testing were taken from the clinical-effectiveness review. The proportions of people who were treated, isolated or discharged were estimated using expert opinion or were taken from Goldenberg et al. (2015). No discounting was applied to costs and effects because of the short time horizon of the model.

Costs

4.25 Resource use and costs included in the hospital-based models were:

  • cost of testing

  • bed days

  • cleaning

  • blood tests and other investigations, including flexible sigmoidoscopy and abdominal X‑ray.

4.26 Medicine and rehydration costs were included in both hospital and community-based models.

4.27 The test costs used in all models, which included consumables, staff costs and overheads, were:

  • conventional test: £66.18 per sample

  • xTAG Gastrointestinal Pathogen Panel: £37.10 per sample

  • Film Array GI Panel: £93.53 per sample.

4.28 Additional costs for confirmatory testing were applied to the integrated multiplex PCR tests when Clostridium difficile, Escherichia coli O157, Salmonella or Shigella were detected. Resource use and medicine costs were taken from published NHS reference costs and the British national formulary, although antimicrobial therapy is not recommended for most gastrointestinal infections. The length of hospital stay was determined by whether a pathogen was identified, whether symptoms persisted, and the identity of the detected pathogen. Length of stay ranged from 2 days for viruses to 19 days for Clostridium difficile. If a second test was needed, discharge was delayed.

Health-related quality of life and quality-adjusted life year decrements

4.29 For all models, utility values were taken from the literature. One study, Minor et al. 2015, estimated healthcare costs associated with gastrointestinal infections in the US and estimated utilities for different pathogens as quality-adjusted life days lost. These utility values were converted to QALYs so they could be used in the model. The mean QALY losses from Minor et al. (2015) ranged from 0.0007 for adenovirus to 0.0126 for Escherichia coli O157. Expert opinion was used to estimate a QALY loss of 0.0137 for Clostridium difficile. There was no disutility included in the model for being hospitalised.

Main assumptions

4.30 The assumptions applied in the base-case analysis were:

  • Conventional testing was the comparator, and was 100% accurate.

  • The pathogen detection rate for conventional testing was 24.1%, 22.2% for the xTAG Gastrointestinal Pathogen Panel, and 22.9% for FilmArray GI Panel.

  • False negatives were identified using a second confirmatory test, which was taken because of persisting symptoms.

  • The daily throughput of the new technologies was 24 samples.

  • Test results were returned in 3 days for conventional testing, 1 day for the xTAG Gastrointestinal Pathogen Panel, and half a day for the FilmArray GI Panel, but this did not affect the length of stay.

  • The minimum total number of bed days for both conventional testing and integrated multiplex PCR tests was 3 days.

  • On admission, every patient had a full blood count and 30% of patients had further biochemistry tests.

  • A flexible sigmoidoscopy was done for 1% of patients if a pathogen was detected, and 10% of patients if a pathogen was not detected.

  • An abdominal X‑ray was done for 10% of patients if Clostridium difficile was detected, and for 10% of patients if no pathogen was detected. For all other pathogens detected, 5% of patients in the conventional testing arm and 3% of patients in the integrated multiplex PCR testing arm had abdominal X‑ray.

  • For each day in hospital, 60% of patients had oral rehydration (200 ml of dioralyte every 4 hours) and 30% had intravenous (IV) fluids (2 litres of sodium chloride 0.9%).

Base-case model results – adults in hospital

4.31 In the base-case model, no difference in length of stay between the interventions and the comparator was assumed. Both deterministic and probabilistic results were presented for the base-case model. In both the deterministic and probabilistic base case, the xTAG Gastrointestinal Pathogen Panel and the FilmArray GI Panel dominated current practice (that is, they were more effective and cost less).

4.32 The cost-effectiveness planes showed that the bootstrap estimates were spread across all 4 quadrants of the plane. The cost-effectiveness acceptability curves show that at a maximum acceptable incremental cost-effectiveness ratio (ICER) of £20,000 per QALY gained, the xTAG Gastrointestinal Pathogen Panel has a 57% probability of being cost effective and the FilmArray GI Panel has a 54% probability of being cost effective.

4.33 A scenario analysis was also done, in which the length of stay for the integrated multiplex PCR tests was decreased because of their shorter test turnaround time. If people are discharged a day earlier, the incremental cost saving increases to £679 per person for the xTAG Gastrointestinal Pathogen Panel (from £63 per person in the base case) and £677 for the FilmArray GI Panel (from £52 per person in the base case). The net monetary benefit also increases, but the 95% credible intervals for this calculation cross zero at a maximum acceptable ICER of £20,000 per QALY gained, showing that there is substantial uncertainty about the cost savings.

4.34 One-way deterministic sensitivity analyses showed that the base-case model results were most sensitive to changes in the number of bed days. The proportion of false-positive results also had an effect because this parameter drives additional testing and care, but the effect was relatively small compared with bed days.

4.35 The effect of changing the assumptions on daily throughput of tests was also explored in a sensitivity analysis. The daily throughput was reduced to 12 samples and increased to 48 samples; under both assumptions, the base-case conclusions were unchanged and the new technologies still dominated conventional testing (that is, they were more effective and cost less).

4.36 The cost of the comparator was changed from £66.18 to £20 in an additional analysis. Under this assumption, the new technologies remained dominant but the cost savings reduced to £26 for the xTAG Gastrointestinal Pathogen Panel and £24 for the FilmArray GI Panel.

Model 2 results – young children in hospital

4.37 The structure used for Model 2 was the same as the base-case model, but some parameters were changed:

  • The pathogen detection rate was assumed to be 24.2% for conventional testing, 21.4% for the xTAG Gastrointestinal Pathogen Panel, and 22.3% for the FilmArray GI Panel. The pathogen prevalence data were also updated using expert opinion.

  • The probability of being isolated and treated after a positive result was lower in Model 2 and there was a greater probability of having no treatment and being discharged.

  • Fewer patients had a flexible sigmoidoscopy and an abdominal X‑ray.

  • QALY loss for rotavirus was changed to 0.0022 to reflect the values applied in UK studies on rotavirus vaccination in children.

  • The condition was treated with oral rehydration in 75% of patients and IV fluids in 20% of patients.

4.38 Both deterministic and probabilistic results were presented for Model 2. In both the deterministic and probabilistic analyses, the xTAG Gastrointestinal Pathogen Panel and the FilmArray GI Panel dominated current practice.

4.39 The cost-effectiveness planes showed that the bootstrap estimates are located in all 4 quadrants. The cost-effectiveness acceptability curves show that at a maximum acceptable ICER of £20,000 per QALY gained, the probability of being cost effective is 58% for the xTAG Gastrointestinal Pathogen Panel and 57% for the FilmArray GI Panel.

4.40 A scenario analysis was done, in which the length of stay for the integrated multiplex PCR tests was decreased because of their shorter test turnaround time. If people are discharged a day earlier, the incremental cost saving increases to £959 per person for the xTAG Gastrointestinal Pathogen Panel (from £73 per person when there is no difference in length of stay) and £969 per person for the FilmArray GI Panel (from £83 per person when there is no difference in length of stay). The net monetary benefit of the interventions also increases compared with current practice as bed days drop, and the 95% credible interval does not cross zero when the length of stay reduces by 1 day, suggesting that the net monetary benefit of the tests is likely to be positive under this assumption.

4.41 The cost of the comparator was changed from £66.18 to £20 in an additional analysis. As with the base-case model, under this assumption the new technologies stayed dominant but the cost savings reduced to £16 for the xTAG Gastrointestinal Pathogen Panel and £43 for the FilmArray GI Panel.

Model 3 results – people in the community

4.42 In this model, the structure of the base-case model was changed to remove the hospital-based decision nodes for admission, isolation and discharge. Changes were also made to some of the parameters:

  • Pathogen prevalence was taken from the Food Standard Agency's second study of infectious intestinal disease in the community.

  • The pathogen detection rate was 34.9% for conventional testing, 31.6% for the xTAG Gastrointestinal Pathogen Panel, and 32.5% for the FilmArray GI Panel.

  • The probability that a patient's symptoms resolved naturally was 0.75 and that the symptoms persisted was 0.25.

  • All patients were assumed to visit their GP before a test was taken.

  • In 10% of patients, oral rehydration was used to treat the condition.

4.43 The deterministic and probabilistic results were presented for Model 3. In both these analyses, the xTAG Gastrointestinal Pathogen Panel dominated current practice. In the deterministic analysis, the FilmArray GI Panel had an ICER of £1,653,939 per QALY gained and in the probabilistic analysis an ICER of £1,309,346 per QALY gained.

4.44 The cost-effectiveness planes showed that the xTAG Gastrointestinal Pathogen Panel is cost saving, whereas the FilmArray GI Panel is cost incurring. For both interventions, it is uncertain whether they are more or less effective than the comparator. The cost-effectiveness acceptability curves show that at a maximum acceptable ICER of £20,000 per QALY gained, the xTAG Gastrointestinal Pathogen Panel has almost a 100% probability of being cost effective, whereas the FilmArray GI Panel has a 6% probability of being cost effective.

4.45 One-way deterministic sensitivity analyses were also done, which showed that the model results were most sensitive to changes in the costs of both the interventions and the comparator.

4.46 The cost of the comparator was changed from £66.18 to £20 in an additional analysis. Under this assumption, the ICER for the xTAG Gastrointestinal Pathogen Panel is £518,112 per QALY gained and the FilmArray GI Panel's ICER increases to £3,264,373 per QALY gained. A threshold analysis showed that when the cost of the comparator was reduced to £36.60, the xTAG Gastrointestinal Pathogen Panel was no longer dominant.

Model 4 results – people who are immunocompromised and in hospital

4.47 The structure used for Model 4 was the same as the base-case model, but changes were made to some of the parameters:

  • The pathogen detection rate, estimated using expert opinion, was 31.1% for conventional testing, 28.7% for the xTAG Gastrointestinal Pathogen Panel, and 29.4% for the FilmArray GI Panel. The pathogen prevalence data were also updated using expert opinion.

  • The proportion of patients having treatment for each pathogen was also updated using expert opinion.

  • The probability of being isolated after a pathogen was detected was greater than in the base-case model.

  • In patients with negative test results, 20% would have a flexible sigmoidoscopy.

  • An abdominal X‑ray would be done in 10% of patients with Clostridium difficile, 20% of people with no pathogen detected, 5% of people with other pathogens detected with conventional tests, and 2% of people with other pathogens detected with the new technologies.

  • Every patient had 2 full blood count tests.

4.48 Deterministic and probabilistic results were presented for Model 4. In both the deterministic and probabilistic analyses the xTAG Gastrointestinal Pathogen Panel and the FilmArray GI Panel dominated current practice.

4.49 The cost-effectiveness planes showed that the bootstrap estimates are located in all 4 quadrants, highlighting the uncertainty in the results. The cost-effectiveness acceptability curves show that at a maximum acceptable ICER of £20,000 per QALY gained, the probability of being cost effective was 55% for the xTAG Gastrointestinal Pathogen Panel and 57% for the FilmArray GI Panel.

4.50 The cost of the comparator was changed from £66.18 to £20 in an additional analysis. As with the base-case model, under this assumption the new technologies stayed dominant but the incremental cost savings reduced to £25 for the xTAG Gastrointestinal Pathogen Panel (from £81 when the comparator costs £66.18) and £30 for the FilmArray GI Panel (from £77 when the comparator costs £66.18).

Model 5 results – people with a recent history of foreign travel

4.51 The structure used for Model 5 was the same as Model 3 (people in the community with suspected gastroenteritis), but changes were made to some of the parameters:

  • Pathogen prevalence was estimated by clinical experts using Model 3 and data from the Food Standard Agency's second study of infectious intestinal disease in the community as a base case, and resulted in higher rates of bacteria and parasites in this model (Model 5).

  • The overall pathogen detection rate was 31.1% for conventional testing, 28.0% for the xTAG Gastrointestinal Pathogen Panel, and 29.2% for the FilmArray GI Panel.

  • There was a greater probability of people having treatment compared with Model 3.

  • In 10% of people, the condition would be treated with oral rehydration.

4.52 The deterministic and probabilistic results were presented for Model 5. In both analyses, the xTAG Gastrointestinal Pathogen Panel dominated current practice. In the deterministic analysis, the FilmArray GI Panel had an ICER of £1,020,674 per QALY gained and in the probabilistic analysis an ICER of £560,220 per QALY gained.

4.53 The cost-effectiveness planes show that the xTAG Gastrointestinal Pathogen Panel is cost saving, whereas the FilmArray GI Panel is cost incurring. For both new technologies, it is uncertain whether they are more or less effective than current practice. The cost-effectiveness acceptability curves show that at a maximum acceptable ICER of £20,000 per QALY gained, the probability of being cost effective is almost 100% for the xTAG Gastrointestinal Pathogen Panel whereas it is 6% for the FilmArray GI Panel.

4.54 The cost of the comparator was changed from £66.18 to £20 in an additional analysis. Under this assumption, the xTAG Gastrointestinal Pathogen Panel had an ICER of £356,931 per QALY gained and the FilmArray GI Panel's ICER increased to £4,203,556 per QALY gained. A threshold analysis showed that when the cost of the comparator was reduced to £36.80, the xTAG Gastrointestinal Pathogen Panel was no longer dominant.

  • National Institute for Health and Care Excellence (NICE)