Testing strategies for Lynch syndrome in people with endometrial cancer

3.9 Data on the accuracy of IHC alone (that is, without MLH1 promoter hypermethylation testing) were available from 5 studies (Berends et al. 2003, Chao et al. 2019, Lu et al. 2007, Rubio et al. 2016, Tian et al. 2019). Sensitivity values ranged from 66.7% to 100%. Specificity values ranged from 6.5% to 83.3%.

3.10 Data on the accuracy of MSI testing alone (that is, without MLH1 promoter hypermethylation testing) were available from 4 studies (Berends et al. 2003, Chao et al. 2019, Lu et al. 2007, Rubio et al. 2016). Sensitivity values ranged from 41.7% to 100%. Specificity values ranged from 69.2% to 88.9%.

3.11 There were data from 4 studies on the accuracy of IHC- or MSI-based testing strategies when these tests were done before MLH1 promoter hypermethylation testing. The studies varied in when promoter hypermethylation testing was done:

3.12 Four of the complete test accuracy studies assessed both IHC and MSI testing on the same population (Lu et al. 2007, Berends et al. 2003, Chao et al. 2019, Rubio et al. 2016). Point estimates for sensitivity ranged from 66.7% to 100% for IHC and from 41.7% to 100% for MSI. For specificity, point estimates for IHC ranged from 60.9% to 83.3%. For MSI the range was 69.2% to 89.9%. The EAG commented that there was no statistically significant difference between the tests.

3.18 The EAG used data from 1 study (Lu et al. 2007) to inform estimates of sensitivity and specificity for the different test strategies for the model. One study was used for consistency (that is, accuracy estimates produced from the same population) and to avoid illogical results, which may have happened if different studies were used for different strategies. The EAG did not consider that pooling results across studies was appropriate because the few studies identified were heterogeneous. Data from a recent meta-analysis (Snowsill et al. 2019), Chao et al. (2019) and the PETALS study (Ryan et al. 2020) were used in scenario analyses.

3.19 Age-related incidence of colorectal cancer for people with Lynch syndrome was taken from Snowsill et al. (2019). This was assumed to differ by which MMR gene was mutated and was estimated using gene specific data from the Prospective Lynch Syndrome Database. A log-normal distribution was fitted to the data to estimate the incidence of colorectal cancer over time. A hazard ratio of 0.387 (Järvinen et al. 2000) was applied to estimate the effect of colonoscopic surveillance on reducing the incidence of colorectal cancer. If a person was having colonoscopic surveillance because Lynch syndrome had been diagnosed, this was assumed to identify colorectal cancer at an earlier stage (as well as reducing incidence).

3.20 Incidence data for endometrial cancer were taken from the Prospective Lynch Syndrome Database (Dominguez-Valentin et al. 2020). The incidence differed by which MMR gene was mutated. A fitted piecewise linear model was used to estimate annual incidence at different ages. Data from Cancer Research UK on uterine cancer survival statistics were used for the incidence of death from endometrial cancer, assuming no difference for people with and without Lynch syndrome.

3.21 Female relatives with Lynch syndrome could choose to have hysterectomy with removal of both ovaries and fallopian tubes (bilateral salpingo-oophorectomy), which eliminated all future risk of endometrial cancer. The uptake of this surgery increased with age, from 20% at 35 years old to 80% at 75 years old. The EAG highlighted considerable uncertainty about the benefit of gynaecological surveillance, and variation in practice across the UK. In its base case, the EAG assumed all female relatives with Lynch syndrome who were 25 years or older (who had not had a hysterectomy) would have annual non-invasive gynaecological surveillance done by a GP. Of these, 10% would be referred for invasive surveillance (gynaecological examination, pelvic ultrasound, cancer antigen‑125 analysis and aspiration biopsy). Gynaecological surveillance was assumed to reduce mortality by 10.2% (Snowsill et al. 2017).

3.22 Most costs were taken from work done for previous NICE guidance on testing for Lynch syndrome after colorectal cancer (Snowsill et al. 2017). Hospital-related costs were from the most current NHS reference tables. The EAG used test costs from the UK Genetic Testing Network (confirmed by clinical experts) in the base case.

3.23 Baseline health-related quality of life for people in the model was calculated based on age and sex. Testing, a diagnosis of Lynch syndrome, surveillance and risk-reducing interventions were assumed to have no effect on health-related quality of life.

3.24 In the base case, a decrease in health-related quality of life for people with colorectal cancer was only assumed to occur at stage 4 (a multiplier of 0.789; Snowsill et al. 2017). Because this may underestimate the effect of colorectal cancer on a person's quality of life, the EAG did a scenario analysis in which people with stage 3 colorectal cancer also had a decrease in health-related quality of life. The health-related quality of life of people with endometrial cancer decreased by 0.036 (Snowsill et al. 2017) for 1 year.

3.27 The EAG did several scenario analyses in its main report:

3.28 In the fully incremental analysis in the base case, IHC then MLH1 promoter hypermethylation testing (strategy number 4) was the most cost-effective strategy. In all scenarios except scenario 8, this strategy had an ICER of less than £12,000 per QALY gained in fully incremental analyses. In scenario 8, the ICER was £20,740 per QALY gained. In all scenarios except scenario 4, all other strategies were either extendedly dominated (the ICER was higher than that of the next more effective alternative), fully dominated (had higher costs and worse outcomes than an alternative strategy) or had ICERs of over £90,000 per QALY gained (fully incremental analysis). In scenario 4, the ICER for IHC testing alone was £41,180 per QALY gained in the fully incremental analysis.

3.29 The EAG also did more scenario analyses in an addendum to its main report. In additional scenario 1, diagnostic accuracy estimates from a meta-analysis done for recent modelling work (Snowsill et al. 2019) were used instead of estimates from Lu et al. (2007). Accuracy data were only available for strategies using MSI and IHC alone (with or without subsequent MLH1 promoter hypermethylation; strategies 1 to 4 in this assessment). In fully incremental analysis, IHC with MLH1 promoter hypermethylation testing had an ICER of £10,464 per QALY gained and IHC alone had an ICER of about £100,000 per QALY gained. MSI and MSI done before MLH1 promoter hypermethylation were either dominated or extendedly dominated.

3.30 In additional scenario 2, accuracy data from Chao et al. (2019) were used. Only accuracy estimates for IHC and MSI alone were available. Here, MSI and no testing extendedly dominated IHC testing and MSI testing had an ICER of £10,455 per QALY gained compared with no testing. In Chao et al. higher estimates of both sensitivity and specificity were seen for MSI testing than IHC testing.

3.31 In additional scenario analysis 3, people with variants of uncertain significance and people who were assumed to have Lynch syndrome did not gain any benefit from surveillance and risk-reducing interventions (in the base case, they were assumed to get the same benefit as people with Lynch syndrome). IHC with MLH1 promoter hypermethylation had an ICER of £9,514 per QALY gained and dominated or extendedly dominated all other strategies.