A new reality of molecular surveillance for resected lung cancer

Compared with the gradual and tedious progress for metastatic tumors over the last 2 decades, the new perioperative paradigm for resectable non-small-cell lung cancer (NSCLC) emerged astonishingly fast, having within 2 years incorporated all essential features of modern cancer medicine: next-generation sequencing (NGS) at initial diagnosis, use of immune-checkpoint inhibitors (ICIs) upfront, and targeted drugs for tumors with actionable driver mutations (1). The importance of these developments is potentiated by ongoing and upcoming lung cancer screening programs in most countries of the world, which are poised to make resectable NSCLC the main mode of lung cancer presentation globally (2). In the near future, we expect the majority of lung cancer patients to have early disease and achieve long-term survival based on the management principles set forth today, however one key piece of the puzzle has been lacking in the curative setting so far: the ability for non-invasive monitoring of residual disease.

With a recent publication in Clinical Cancer Research, Zhang et al. nicely demonstrate how this once distant dream is now becoming a new reality (3). Blood-based minimal residual disease (MRD) assays have for decades been the privilege of malignant hematology, as blood cancers typically feature neoplastic cells in the circulation contrary to solid tumors, in which circulating tumor cells are very rare (4). Therefore, among solid tumors including lung cancer surveillance for recurrence after surgery has relied on radiological studies, mostly computed tomography of the chest and abdomen performed in regular intervals, but evidence about an impact on survival remains mixed and somewhat controversial (5). However, owing to several recent technological breakthroughs, we can now track cell-free (cf) circulating tumor deoxyribonucleic Acid (ctDNA) in human plasma with a sensitivity beyond 0,01% minimal detectable variant allelic frequency (VAF), which brings most NSCLC patients, even those with early-stage disease, within the realm of our monitoring capabilities (6). Zhang et al. were globally among the first to recognize this huge potential and launched a pivotal prospective study, whose initial (7) and currently updated results (3) together contribute valuable experience and provide new data about how to best make use of this novel tool (Table 1).

Table 1. Key practical issues for MRD monitoring after curative surgery of NSCLC according to the study by Zhang et al. (3).

Testing strategy	Longitudinal monitoring (surveillance)
Landmark sample	1 month (±14 days) after surgery or after the last cycle of adjuvant chemotherapy; for patients starting adjuvant therapy: the first sample at least 1 week after surgery
Blood sample quantity	20 mL (2× 10 mL Streck tubes)
Sequencing controls	Matched tumor tissue; paired white blood cells; extensive bioinformatic filtering of detected variants
Frequency of subsequent sampling	Every 3–6 months
Duration of monitoring	>3 years, complemented by serial brain MRI
Positive predictive value	92.8% (higher for more advanced initial stage); main cause for false positives is non-persistent MRD positivity
Negative predictive value	93.2% (independent of initial tumor stage); main cause for false negatives is brain-only relapse
Peak likelihood of a positive result	18 months, bell-shaped curve
Median lead time	5.2 months

MRD, minimal residual disease; MRI, magnetic resonance imaging; NSCLC, non-small-cell lung cancer.

One first principle is the need for longitudinal monitoring, which is essential in order to maximize detection of NSCLC patients with residual cancer after surgery (7). As currently available methods may still miss some tumors with very minute ctDNA traces, the sensitivity of ctDNA detection in stage I–III disease by Zhang et al. was relatively low with 36.4% (95/261 positive samples) before vs. 8.8% (23/261) after surgery (7). Consequently, the negative predictive value (NPV) for subsequent relapse of a single negative liquid biopsy at the landmark time-point 1 month after surgery was only 76.5%, but could be increased to 93.2% if the test results of subsequent samples collected every 3–6 months were also taken into account (3). It should be stressed here, that this risk of false negatives holds true for all contemporary ctDNA assays, since even PhasedSeq with a minimal detectable VAF <0.00001% has suffered from patients relapsing after resection of NSCLC despite undetectable ctDNA in landmark testing (8). Of note, false negative liquid biopsy results can occasionally also occur with serial monitoring, but these were rare, affecting only 13/192 (6.7%) of MRD negative patients in the updated report of Zhang et al. with a median follow-up exceeding 3.5 years (43.4 months), and mainly caused by brain-only recurrence (7/13) or lack of testing in the last 6 months preceding relapse (4/13) (3). A similar lower sensitivity of ctDNA testing in case of disease progression confined to the central nervous system has also been demonstrated for metastatic NSCLC (9). Hence, the main reasons for failure of liquid biopsy monitoring to identify subsequent disease relapse are either logistical or anatomical, which explains why this phenomenon showed no association with the disease stage at initial diagnosis in the study of Zhang et al. (3). In practical terms, this means liquid biopsy monitoring of NSCLC after curative surgery should be performed regularly and complemented with surveillance using brain magnetic resonance imaging (MRI) in order to improve performance. Overall, in the study of Zhang et al., 948 postoperative samples were collected from 261 patients regularly every 3–6 months over the course of 4.5 years (from March 2019 until September 2023).

On the other hand, the positive predictive value (PPV) of liquid biopsies with detectable ctDNA was always high (91.3% in landmark vs. 92.8% in longitudinal testing), and increased with more advanced preoperative tumor stage (III vs. II vs. I) due to the higher a priori likelihood of relapse according to the Bayes’ theorem (3). Interestingly, detectable ctDNA without subsequent relapse was also observed, but these cases were very rare (5/69 or 7.2% of patients) and all showed non-persistent MRD positivity only, which means that a positive result was followed by negative subsequent samples. This temporary false positivity may have partly been due to technical reasons despite the rigorous sequencing controls using both matched tumor tissue and matched white blood cell (WBC) DNA for every patient, as well as meticulous filtering out of variants with very few supporting reads (<2 for hotspot/<4 for other mutations) and/or presence in the patients’ germline controls or public single-nucleotide polymorphism or WBC DNA databases. However, the presence of residual cancer cannot be excluded in cases with temporary MRD positivity, as 4 patients in the study of Zhang et al. had a lead time (i.e., time between the first positive ctDNA sample and radiologic relapse) exceeding 30 months (3). Since latency can be that long, future MRD monitoring programs should extend over several years, even though the likelihood of a positive MRD sample is highest approximately 18 months after surgery and declines afterwards. The median lead time observed by Zhang et al. for all relapsed patients was 5.2 months, which is longer than that reported using less sensitive, conventional ctDNA assays to detect disease progression earlier compared to imaging studies in the metastatic setting, e.g., approximately 3 months in anaplastic lymphoma kinase (ALK)-driven NSCLC (10). In multivariable analysis, the duration of lead time correlated with elevated plasma ctDNA levels [>13 haploid genome equivalents per milliliter (hGE/mL)] and higher maximum tumor VAF (≥0.55%), which were generally associated with short-term recurrence within 6 months, in contrast to a wide range of 1–35 months for the time to relapse in patients with less pronounced molecular findings. Combining single-nucleotide variants (SNVs), which were exclusively interrogated in the study of Zhang et al. (3), with copy number alterations, fragmentomic or epigenetic features and other ctDNA or circulating tumor ribonucleic acid (ctRNA) characteristics may therefore improve segregation of patients with low detectable ctDNA levels and imminent vs. delayed relapse, as it could also mitigate the challenge of landmark-negative progressors (11,12). Furthermore, exosome evaluation can provide additional information, for example through the monitoring the levels of specific long non-coding RNAs (lncRNAs), which have also been associated with relapse in a pivotal study (13). Such multiparametric assays could prove particularly useful for the monitoring of tumors with low tumor mutational burden and hence fewer SNVs, like the ALK⁺ NSCLC, especially in the presence of high-risk molecular features, like the more potent EML4 ALK variant 3 and/or TP53 mutations, which are associated with a higher abundance of copy number alterations and epigenetic markers in the ctDNA (14-16). From a practical standpoint, MRD positivity after definitive local treatment is very important because it can be predictive for the subsequent benefit from adjuvant or consolidation therapy for both resectable stage II/III and inoperable stage III tumors, as the current study of Zhang et al. along with several other previous analyses have demonstrated (3,8,17). How could we take advantage of this?

For regulatory approval, public reimbursement and wide adoption, interventional prospective studies are essential, but unfortunately such results are not available yet. Therefore, ctDNA surveillance after surgery for lung cancer at present remains experimental and feasible only within the context of clinical trials. In order to facilitate progress, the United States Food and Drug Administration (FDA) has published detailed guidance for the industry outlining key methodological requirements (18). Based on the technical status-quo with several commercially available MRD methods currently, and the established poor prognosis of MRD positive patients (19,20), treatment escalation studies are reasonable and could be realized in the postoperative setting as intensification of adjuvant therapy in case of detectable MRD, inasmuch as several novel drugs, like antibody drug conjugates and multispecific antibodies, emerge as promising for potential adoption in interventional arms (21,22). Similar treatment escalation studies could also be performed in the neoadjuvant setting, e.g., with intensification of further systemic therapy for patients with insufficient response and persistent MRD after two preoperative cycles, as MRD negativity before surgery appears to be sine qua non for pathologic complete remission (23,24). On the other hand, deescalation studies remain problematic due to the limited sensitivity of current assays and low NPV of negative landmark results, as already outlined. However, as the study of Zhang et al. elegantly showed, serial longitudinal testing can compensate for the sensitivity deficit of landmark assays and facilitate a very good NPV >90% even with contemporary methods (3). This important insight creates the chance for deescalation in patients who remain MRD negative during surveillance after surgery vs. preemptive (delayed) adjuvant treatment in case of MRD conversion in the postoperative follow-up. Based on this novel concept, Zhang et al. continue to pioneer the field by launching the first trial of adjuvant therapy-free strategy for stage IB to IIIA NSCLC patients after radical resection based on longitudinal undetectable MRD (CTONG 2201; NCT05457049) (25). By relying on the insights gained through the observational prospective study under discussion (3,7), this innovative new trial represents one additional reason why the current work by Zhang et al. has unique importance in the field and becomes instrumental for making non-invasive molecular monitoring of NSCLC under curative treatment a new reality.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.