This is one of the first studies that aims to evaluate the sensitivity of the cobas test in a clinical setting and investigates the role of microscopic dissection of small biopsies to overcome its false negative results. This study revealed that the EGFR mutation detection failure of cobas occurred in about 25% of EGFR mutated patients in a clinical setting. Moreover, we demonstrated that the false negative rate of cobas went up to 50% when the tumour content ratio was below 10% and that dissection on a slide could improve the test sensitivity.
Now that EGFR-TKI therapy dramatically improved the prognosis for EGFR mutated NSCLC [4, 10-12], the detection of such mutations is a critical step when managing lung cancer patients. Recent evidence has shown the superiority of osimeritinib over first- and second-generation EGFR-TKIs and is an approved first line treatment for patients with EGFR-mutated NSCLC [4]. Cobas ver2 is the only osimertinib companion diagnostic tool to identify EGFR mutations. Although it is considered to be as precise as conventional methods [13], there are only a few studies that compare the sensitivity of these methods [14]. One study compared the detection abilities of EGFR mutations between cobas and PCR clamp with 15 re-biopsied samples and reported consistent detection accuracy between the two assays except for one sample that tested positive with PCR clamp but negative with cobas [14]. In our study, we found a 25% false negative rate with cobas compared with that of PCR clamp. This lower sensitivity in cobas was consistent with that revealed in the report [14], and, moreover, our study successfully calculated the high false negative rate of cobas with a large population.
This low positive rate of cobas was not reported at the correlation tests of EGFR mutation identification performed previously to the approval of the cobas method [15]. We hypothesized that the different experimental conditions in our study for clinical setting and in laboratory research for sensitivity test could be one of the important factors of this false negative results. As we analysed tumour samples taken for usual clinical practice, our study included samples with low tumour content ratio. With the trait of cobas, these samples with limited tumour cellularity potentially cause EGFR mutation detection failure. The cobas method is an in vitro diagnostic tool that applies an allele-specific PCR method. Matching failure of primers at annealing tends to occur in samples with a lower rate of EGFR mutations over total EGFR and this determines the sensitivity of the cobas method for EGFR mutation identification. Furthermore, the PCR clamp, we used as a control for EGFR mutation identification, is a major laboratory developed test with the sensitivity for EGFR mutations as high as 1%. This high sensitivity of PCR clamp was confirmed by masking wild type EGFR with specific nucleic acids during the PCR process [11]. However, as PCR clamp was independently developed in a major laboratory centre, the problem of uniformisation between laboratories should be considered, while the consistency of the cobas method is guaranteed as an in vitro diagnostic. These specific processes in PCR and non-commercial optimisation of PCR clamp could affect both high sensitivity and false positive rates in EGFR mutation identification, especially when the response to EGFR-TKI therapy is considered. In our cohort, EGFR mutation was determined by PCR clamp at inclusion, and patients had an equivalent EGFR-TKI response to previous studies [10, 11]. Moreover, both cobas-positive and -negative patients showed no difference in the response. From these results, we concluded that our result revealed the high false negative rate of the cobas method instead of high false positive rate of the PCR clamp.
Our study also suggests the association between a low tumour content ratio and high false negative rates of EGFR identification in a small biopsy cohort in the cobas method. Although AUC of the ROC curve could only demonstrate a weak relationship, importantly, when the tumour content ratio was below 10%, the false negative rate was as high as 50%. Therefore, we focused on the microscopic tumour dissection on a slide to confirm high tumour cellularity of the samples and improve its low sensitivity. Indeed, the manufacturer of cobas encourages tumour dissection. However, the dissection is not routinely performed in clinical settings, especially on small biopsies. This might be due to a cost-performance effect or the limitation from pathologist manpower at the hospital, but we have to acknowledge that the evidence of dissection in conjunction with cobas is lacking. Here, we revealed that 24% of the cobas negative samples showed positive EGFR mutations after performing tumour dissection on a slide with bronchoscopic biopsied samples. Especially in a subgroup with tumour content ratio below 10%, one-third of the EGFR mutation false negative samples with the cobas method at first turned to be positive at the re-test after the dissection. This study provides some of the first evidence for the importance of maintaining high tumour content ratio when testing for EGFR mutations and encourages microscopic tumour dissection with small biopsies.
Our study did have some technical limitation that were important to consider during our analysis and interpretation. Firstly, we could only confirm EGFR mutation with PCR clamp and cobas performed at the same time in 97 samples out of the 132 included patients. With the rest 35 samples, PCR clamp was conducted prior to the test with cobas method at the beginning of this study. In these samples, the best part of tissue samples for the genomic testing were already taken at the time of the PCR clamp testing, and samples with worse quality could have been used for the cobas testing as compared to those used for the initial PCR clamp tests, especially the cobas testing after the tumour dissection that was conducted later in our study. Although this could somewhat contribute to the high false negative rate in the cobas tests, the false negative rate was consistent even in samples where the sample slides for cobas method were made at the same time as those collected for PCR clamp at the beginning of our study. We also demonstrated the recovery of EGFR mutation detection after the dissection for samples whose slides were made after the cobas test showed a negative result. Based on these results, we conclude that the false negative rate of cobas observed in our study reliably represents the clinical setting and that time of sampling does not significantly alter the different diagnostic outcomes.
Secondly, we observed that 16% of the samples retested with an improved PCR clamp at the beginning of the study were negative for EGFR mutation. The long retention period from the time the samples were collected to the time of the EGFR mutation test may have contributed to this result. Indeed, the Japanese Society of Pathology reports that DNA deteriorates in old samples, showing a decreased Q-value of DNA over time [9]. The varying sample collection times for testing EGFR mutations may, therefore, also be another factor contributing to observed negative result. However, our cohort showed consistent characteristics including response to the TKIs, with those of a previous study [10, 11], suggesting the validity of our study results for representing the sensitivity of cobas and the utility of the dissection in clinical settings.
Other limitation is that the number of the false negative results with cobas was so small that the association between the tumour content ratio and cobas sensitivity was not well described. Further, the precise histology including patterns of invasive adenocarcinoma were not available preventing us from investigating the association between tumour histology and EGFR mutation detection with cobas. Moreover, this is a retrospective analysis with samples collected more than ten years earlier. There is also a possibility of confounding factors we could not evaluate and further studies would be needed to address these challenges. However, we emphasize the importance of this study as one of the first to point out the high false negative rate in the realistic use of the cobas method and importance of dissection even in a small sample for maintaining its sensitivity.