Clinical and technical evidence

A literature search was carried out for this briefing in accordance with the interim process and methods statement. This briefing includes the most relevant or best available published evidence relating to the clinical effectiveness of the technology. Further information about how the evidence for this briefing was selected is available on request by contacting mibs@nice.org.uk.

Published evidence

Seven studies including 2,106 participants are summarised in this briefing. All are prospective studies, 2 are randomised controlled trials (RCTs; n=819). Two studies recruited participants internationally, including within the UK, and 5 studies were done in Asia.

Table 4 summarises the clinical evidence as well as its strengths and limitations.

Overall assessment of the evidence

Although diagnostic test accuracies varied from study to study, concordances between plasma and tissue tests were generally above 80%. Additionally, clinical outcomes and treatment efficacy were similar for people diagnosed with epidermal growth factor receptor (EGFR) mutation-positive disease by either tissue or plasma. The digital droplet polymerase chain reaction (ddPCR) and Cobas v2 had slightly better diagnostic test accuracies relative to the other tests.

One of the studies included (Jenkins et al. 2017) describes the findings from 2 RCTs; it had a robust study design and a relatively large sample population. There was a lack of evidence available on populations with disease who had a negative-mutation status when tested using tissue EGFR, but a positive mutation when tested using plasma EGFR. This is because negative status in tissue samples was an exclusion criteria for many of the studies. In 1 of the included studies, subgroup analysis revealed significant differences in progression-free survival (PFS) based on plasma mutation status for people whose disease has tissue-positive status. They found that people whose disease has plasma-positive and tissue-negative status may have lower PFS than people with disease with a negative status in both plasma and tissue.

Most of the studies did not differentiate between the multiple types of EGFR‑sensitising mutations; they instead grouped participants by positive- or negative-mutation status. However, individual mutations might have different sensitivities and specificities and may lead to different clinical outcomes. The reference standard in these studies tended to be a tissue EGFR mutation test; it is generally agreed that this test is not a 'gold-standard' test. Tumour shrinkage or next-generation sequencing (NGS) are also used as reference standards. Five of the 7 studies did not report the time between plasma and tissue testing; it is strongly supported that the time interval should be minimal and that using stored samples is not ideal for diagnostic accuracy studies.

Table 4 Summary of evidence

Jenkins et al. (2017)

Study size, design and location

551 adults with advanced NSCLC in 2 phase II, single-arm, open-label, multicentre clinical trials (AURA extension and AURA2) recruited in 41 locations across Europe, Asia, Canada and the US.

Intervention and comparator(s)

Index test: Roche Cobas v2 (plasma test).

Comparator: Roche Cobas v2 (tissue test) and NGS on an Illumina MiSeq.

Key outcomes

Positive and negative agreements between plasma and tissue tests for detection of T790M mutation were 61% and 79% respectively. Comparing the plasma test with NGS showed positive and negative agreements of higher than 90%. The ORR was 64% in people with T790M mutation-positive assessed by both tissue and plasma tests. People whose disease had positive-tumour and negative-plasma status had higher ORR (70%) than people with both tumour- and plasma-positive statuses.

Strengths and limitations

Inclusion in the sample was based on existing tissue testing results; there are no results for plasma testing alone. NGS was used as the comparator; this is considered to be the gold-standard comparator for EGFR mutation testing in research. Plasma testing was done prospectively, but tissue results were retrospective.

Kim et al. (2017)

Study size, design and location

102 adults with EGFR-mutated NSCLC in a single-centre, prospective, observational, case-control study, South Korea.

Intervention and comparator(s)

Index test: Panagene PanaMutyper R EGFR (plasma test).

Comparator: Panagene PanaMutyper R EGFR (tissue test).

Key outcomes

The agreement between matched tissue and plasma samples for specific mutations was 80.4% for Ex19del, 90.2% for L858R and 56.3% for T790M. At 4 weeks after EGFR‑TKI treatment, detection of mutations in plasma predicted lower ORR and PFS. Plasma EGFR testing could detect the presence of T790M (indicating EGFR‑TKI resistance) on average 103 days before tumour progression detected by CT imaging.

Strengths and limitations

Results for each mutation were reported separately. Plasma testing conducted prospectively. Case-control study design is not recommended for studies of diagnostic tests.

Li et al. (2017)

Study size, design and location

109 adults with metastatic NSCLC in a single-centre, prospective, observational study in China.

Intervention and comparator(s)

Index test: AmoyDx SuperARMS EGFR mutation detection kit (plasma test).

Comparator: AmoyDx EGFR29 mutation detection kit (tissue test).

Key outcomes

EGFR mutations were detected in 45.9% of the plasma samples and in 56.9% of the matched tumour tissue samples. The overall concordance between matched plasma and tissue tests was 89.9%. Sensitivity, specificity, PPV and NPV for plasma EGFR mutation detection were 82.0%, 100%, 100% and 81.4% respectively. The ORR for people with plasma-positive mutations on first generation EGFR‑TKIs was 65.7%, which was comparable to the tissue-positive ORR (64.3%).

Strengths and limitations

PFS and OS are not reported here, however they will be published in a future publication by the authors. The interval of time between tissue and blood testing was relatively small (14 days).

Ma et al. (2016)

Study size, design and location

219 adults with advanced NSCLC in a single-centre, prospective, observational study in China.

Intervention and comparator(s)

Index test: AmoyDx EGFR29 mutation detection kit (plasma test).

Comparator: AmoyDx EGFR29 mutation detection kit (tissue test).

Key outcomes

The overall concordance rate between matched plasma and tissue samples was 82%. The overall sensitivity and specificity were 60% and 97% respectively. No significant difference in median PFS was observed between people with positive mutation status in plasma or tissue testing (10.88 months versus 9.89 months, p=0.411). For people with stage III NSCLC, the sensitivity, specificity and concordance were 50%, 100% and 81.3% respectively. For stage IV, the sensitivity, specificity and concordance were 60%, 96% and 81% respectively.

Strengths and limitations

People with different identified mutations were grouped together as EGFR mutation-positive, however, results by stage of cancer were presented separately.

Thress et al. (2015)

Study size, design and location

Plasma and tumour samples were obtained from patients with NSCLC enrolled in a multicentre open-label phase 1 trial (NCT01802632). 38 plasma samples were used for an initial assessment of all technologies, an additional 72 plasma and matched tumour samples were used in further investigations of Cobas v2.

Intervention and comparator(s)

Index test: Roche Cobas v2, Qiagen Therascreen, Bio‑Rad ddPCR.

Comparator: Roche Cobas v2 (tissue test).

Key outcomes

All 3 plasma EGFR mutation tests had high sensitivity (78–90%) and specificity (100%) for EGFR‑TKI sensitising mutations. The digital platform (ddPCR) had marginally higher values than the analogue tests (Cobas v2 and Therascreen) when considering all mutations (sensitising and resistance). In further assessments for the T790M mutation, only Cobas v2 was assessed. Sensitivity and specificity were 73% and 67%. Concordance between the Cobas and a digital technology not considered in this briefing was >90%. ORR for people with T790M detected in plasma or tissue was comparable (59% and 61% respectively).

Strengths and limitations

At the time of this study, the Bio‑Rad ddPCR test did not detect ex19del mutation, therefore, the overall sensitivity value for this test does not account for ex19del. The sensitivity values for the other tests do account for ex19del. Bio‑Rad ddPCR EGFR mutation testing can now detect ex19del mutations.

Wu et al. (2017)

Study size, design and location

709 people with advanced NSCLC recruited from 2 phase III RCTs (LUX‑Lung 3 [LL3] and LUX‑Lung 6 [LL6] trials), in China, South Korea and Thailand.

Intervention and comparator(s)

Index test: Qiagen Therascreen (Plasma sample).

Comparator: Qiagen Therascreen (tumour tissue sample).

Key outcomes

EGFR mutation detection rates were 28.6% in serum and 60.5% in plasma. People with plasma mutations had shorter PFS and OS than those without. There was no evidence to suggest that the treatment effect (afatinib versus chemotherapy) was different for either group (all subjects were tumour mutation status positive).

Strengths and limitations

The comparator used was tissue testing. Samples were taken from 2 different trials, 1 using serum (LL3) and the other, plasma (LL6). Serum may have a lower overall detection rate accounting for the difference in detection. However, these trials used different DNA extraction kits and methodologies on different populations. Future results from these trials are forthcoming and will update on the results of this study.

Zhou et al. (2017)

Study size, design and location

306 people with advanced NSCLC having osimertinib in a phase II, open-label, single-arm study (AURA17 trial) recruited in Australia, China and South Korea.

Intervention and comparator(s)

Index test: AmoyDx SuperARMS EGFR T790M mutation detection kit, Roche Cobas v2 and an in‑house ddPCR (the Innovation Centre China, AstraZeneca).

Comparator: Roche Cobas v2 (tissue test).

Key outcomes

For the 3 tests, PPA for T790M was 42% to 56% and NPA was 73% to 83%. ddPCR had the highest PPA and Roche Cobas had the highest NPA.

Using the Cobas plasma test as a reference, OPA was higher than 80% for both SuperARMS and ddPCR. For people with positive T790 statuses in both tumour and plasma tests, the ORR with osimertinib was consistent across the 3 plasma tests (56% to 64%).

Strengths and limitations

Despite being a conference proceeding, the abstract contained detailed methodological information.

The PPA found here was lower than previously published values from other AURA trials. The authors suggest this could be associated with the lower prevalence of people with extra-thoracic disease in AURA17. People with extra-thoracic disease are more likely to have a higher disease burden resulting in a higher chance of T790M detection in plasma.

Abbreviations: ddPCR, digital droplet PCR; EGFR, epidermal growth factor receptor; EGFR‑TKI, EGFR tyrosine kinase inhibitor; NGS, next-generation sequencing; NPA, negative percentage agreement; NPV, negative predictive value; NSCLC, non-small-cell lung cancer; OPA, overall percentage agreement; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; PPA, positive percentage agreement; PPV, positive predictive value; RCT, randomised controlled trial.

Recent and ongoing studies