Discussion

The use of blood based biomarkers in lung cancer has been widely studied in recent years; however, not many studies concern the use of biomarkers in patients treated with RT. Given that a substantial number of patients with localized lung cancer are treated primarily with RT, we saw a demand for the present study, which to our knowledge is the first concerning methylated ctDNA, sPD-L1, and NKA in patients with localized or locally advanced NSCLC treated with curative RT.

This study shows the statistically significant discriminative potential of NKA, sPD-L1, and methylated ctDNA predicting the risk of relapse during 12-month follow-up after curative RT for lung cancer.

NKA measured at baseline was prognostic by itself, meaning that patients with a baseline value of NKA below 250 pg/mL had a statistically significant higher risk of experiencing a relapse within a 12-month follow-up period. This is in line with the findings of Choi et al. [23], who found a significant correlation between reduced baseline NKA levels and shorter progression-free survival (PFS) in a cohort of 34 patients with advanced NSCLC treated with immune checkpoint inhibitors (ICIs) (median PFS 37 vs. 78 days, respectively; log-rank test p = 0.003). They likewise found that reduced baseline NKA levels predicted a higher risk of progressive disease (PD) at the follow-up response assessments, thus predicting a worse outcome of the treatment. Worth mentioning is that this study used a higher cutoff of 1200 pg/mL NKA, dividing patients into the normal and reduced groups, obtained through an ROC analysis of their data, instead of the cutoff of 250 pg/mL recommended by the manufacturer and used in our study.

Having a positive ctDNA status, i.e., two or more ctDNA markers above the LoB at baseline, was also statistically significantly prognostic of a higher risk of experiencing a relapse during the 12-month follow-up. This was also demonstrated in a study involving 132 patients with advanced lung cancer by Frank et al. [24], who showed a significantly longer OS in patients who had undetectable ctDNA at baseline (median 578 vs. 296 days, respectively; HR 0.66 [0.44–0.98]. This study used a more extensive, tumor-informed approach where every patient underwent a tissue biopsy, which was tested with the TSO500 HT gene panel (Illumina). The identified mutations were then used to design a personalized ctDNA assay, measured in plasma. In contrast, our study used a less labor-intensive, tumor-agnostic approach that may be easier to implement in clinical practice.

We also found that having positive ctDNA status at the 6-month blood sampling was significantly prognostic of developing a relapse within the first 12 months. However, due to the small numbers in this analysis (N = 36 and only 8 relapses), we tend to regard this as a random result, although the same association was found in another study examining mutations in ctDNA after treatment of lung cancer with chemo-RT [25]. In this study, the detection of ctDNA 4.5 months after treatment was significantly associated with higher odds for tumor recurrence (OR 5.4 [1.1–31]), where ctDNA was detectable in 16 patients out of 39.

The baseline sPD-L1 value was borderline significant by itself, meaning that patients with values above the median may have a higher risk of experiencing a relapse before 12-month follow-up compared to patients with values below the median. Use of a more optimal cutoff for sPD-L1 could potentially have resulted in a more desirable outcome; however, this limit is not known. A study by Mazzaschi and colleagues illustrates the application of an alternative cutoff and an alternative outcome [26]. Using the same immunoassay as in our study, they quantified sPD-L1 in baseline samples from 109 patients with advanced lung cancer. A cutoff of 113 pg/mL, defined by CART tree analysis, was applied. They found that patients with sPD-L1 levels above this cutoff exhibited a significantly shorter median PFS (3.8 months) than those with sPD-L1 levels below the cutoff (11.9 months—HR 2.55 [1.50–4.32]). Like us, they also hypothesized that NK cells could be of prognostic value in lung cancer. While their results showed that patients with low NK-cell count had shorter PFS (median PFS 3.8 months vs. 10.2 months, respectively, p < 0.001), our study demonstrated that lung cancer patients with low NKA experienced shorter PFS. The two different approaches preclude a direct comparison of the results. NK-cell count has been shown to be a relevant biomarker in some conditions; however, as this cell type can be affected by a systemic immunosuppressive environment, they may be present, but not functional. The methodology used in our study aimed to mitigate the impact of such bias. The findings from both studies, however, point to NK cells as a prognostic biomarker in lung cancer, highlighting their possible utility in guiding clinical decision-making.

The ROC analysis of the four methylated ctDNA loci also revealed some discriminative potential; however, the highest AUC values required the addition of NKA and sPD-L1 in the ROC analysis.

Most of the patients with reduced baseline NKA, increased sPD-L1, and/or presence of methylated ctDNA were in stage 3 and received long-course RT. Compared to the patients treated with SBRT, they had extensive disease with larger primary tumors and/or involvement of regional lymph nodes. These factors might have increased the shedding of ctDNA and sPD-L1 to the circulation, and increased the risk of developing reduced NKA [27]. The stage III patients received chemo-RT, a less-effective treatment than SBRT, and have a higher risk of relapse due to a more advanced disease with a larger disease burden. Still, it may seem that having a reduced baseline NKA level or positive methylated ctDNA status could be a warning that intensified treatment and closer follow-up are indicated for stage III patients. This could mean receiving adjuvant immunotherapy if eligible and/or doing follow-up CT scans at shorter intervals.

The values of these biomarkers for stage I lung cancer seem modest and require further investigation, as none of the relapsing patients were positive for ctDNA, and only two out of seven patients with reduced NKA at baseline experienced a relapse. The ambiguous ability of ctDNA to distinguish relapsing patients from non-relapsing patients was also shown in a subgroup analysis of 13 patients with NSCLC treated with curative radiotherapy or chemoradiotherapy in the Danish SUPE_R trial [28]. In this subgroup analysis, having detectable mutated ctDNA either at an early follow-up < 4.5 months following treatment or at a later follow-up > 4.5 months after treatment was not associated with higher risk of recurrence within 24 months after RT. Although only a subgroup analysis with a small sample size, this supports the need for larger studies with more patients and longer follow-up to fully establish the role of ctDNA and other biomarkers in detection of relapse and minimal residual disease in stage I lung cancer patients.

Several patients had PET/CT scans made during their follow-up program due to enrollment in other studies, but we did not see a higher relapse rate among these patients compared to the patients followed with CT scans, so we do not expect this to have influenced our results.

At the first control, between 5 weeks and 3 months after the treatment, most patients had regained normal levels of the biomarkers of interest, i.e., ctDNA was below the cutoff; levels of sPD-L1 were below the median of the baseline samples, and NKA was no longer reduced. This is in line with the above-mentioned study examining biomarkers after RT for lung cancer, where blood sampled on average 1.6 months after RT did not discriminate patients with later relapse [25]. This might be explained by the cancer cells being eliminated and not shedding measurable amounts of ctDNA shortly after completion of RT. With more months passing, the growing tumor might again shed measurable amounts of ctDNA, increased sPD-L1, or cause reduced NKA. However, our study did not have enough relapses between 6 and 12 months following RT to allow meaningful analyses of these possible patterns.

The strengths of this study include the detailed panel of blood analyses being performed. The analyses are tumor-agnostic and not very resource-demanding, allowing easy implementation in clinical practice. The cohort was also well described and had complete radiologic and clinical follow-up in the period of interest.

The limitations of this study include the limited follow-up period of 12 months following RT, the number of patients with missing samples, and the fact that the cohort is a combination of patients treated with SBRT and long-course RT. Nevertheless, this study presents interesting data suggesting the prognostic capability of the combination of ctDNA, sPD-L1, and NKA.