Results

Recruitment for this study began in December 2019. By the inclusion cut-off date of September 25, 2023, 74 patients had been included in the study. The date of final follow-up was May 11, 2025, and all patients had at least 12-month follow-up at this timepoint.

Figure 1 shows a detailed flowchart of the patient inclusion and blood sampling.

Fig. 1: Inclusion and analysis CONSORT diagram

Two patients did not have baseline blood samples taken, but contributed with later blood samples to the study and remained in the cohort.

Nine patients did not have blood sampling performed at the first follow-up visit following completion of RT. We used the blood sample taken at the end of treatment instead, which explains the short duration between the first and second blood samples for some SBRT patients.

Soluble PD-L1 analyses were not available for 6- and 9-month control blood samples.

Table 1 shows the patient characteristics. Two patients were treated with SBRT with 50 Gy in five fractions because of proximity to the main bronchi. All patients receiving long-course RT also received concomitant chemotherapy, with carboplatin + vinorelbine being most predominant.

Patient and treatment characteristics | Value, n (%)
Total number of patients | 68
Age, median (range and IQR) | 70 years (52–86 years, IQR 11 years)
Male sex | 36 (53%)
Histologic subtype
Adenocarcinoma | 37 (54%)
Squamous cell carcinoma | 25 (37%)
Othera | 6 (9%)
Disease stage
I | 32 (47%)
II | 3 (4%)
III | 33 (49%)
Radiation treatment
Long-course RT (24–33 fractions) | 35 (51%)
SBRT (3–5 fractions) | 33 (49%)
Concomitant chemotherapy (N = 35 (51%))
Cisplatin + vinorelbine | 6 (17%)
Carboplatin + vinorelbine | 26 (74%)
Vinorelbine monotherapy | 3 (9%)
WHO performance status
0 | 13 (19%)
1 | 40 (59%)
2 | 15 (22%)
Smoking status
Never | 1 (2%)
Former | 34 (50%)
Active | 33 (49%)
Smoking—number of pack years, median (IQR) | 41 pack years (IQR 18)
Body mass index, median (IQR) | 25 kg/m2 (IQR 6)
Charlson’s comorbidity index, median (range and IQR) | 6 (3–10, IQR 1)
Adjuvant Durvalumab following RT | 11 (16%)
Median number of series received (range and IQR) | 8 series (5–12, IQR 7)

Relapse pattern following curative RT for lung cancer

In the cohort, 26 out of the 68 patients (38%) experienced a relapse within 12-month follow-up. Fourteen of these relapses were verified with a biopsy; the remaining twelve were radiological relapses. Six of the relapsed patients died before 12-month follow-up.

Twenty-one out of 36 patients with stage II–III disease relapsed during the first 12 months (58%), whereas only 5 out of 32 stage I patients relapsed in the same time period (16%).

NKA and sPD-L1 analyses and risk of relapse within 12-month follow-up

We evaluated the association between NKA status (normal vs. reduced) at the different blood-sampling timepoints and the risk of relapse within 12 months of follow-up (Table 2). Twenty one of the 26 patients with a relapse within 12 months had baseline NKA measurements, and 2/3 of these were reduced at baseline. The Pearson chi2 test revealed a significantly higher risk of experiencing a relapse within 12 months after curative RT for patients with reduced NKA at baseline (p = 0.01).

Blood sampling timepoint | NKA levels—cutoff 250 pg/mL | Patients without relapse < 12 months follow-up, n (%) | Patients with relapse < 12 months follow-up, n (%) | Pearson chi2
Baseline | Normal | 22 (42%) | 7 (13%) | p = 0.01N = 53
Reduced | 10 (19%) | 14 (26%)
First control (5 weeks or 3 months) | Normal | 23 (50%) | 7 (15%) | p = 0.31N = 46
Reduced | 10 (22%) | 6 (13%)
Six months | Normal | 13 (39%) | 1 (3%) | p = 0.10N = 33
Reduced | 12 (36%) | 7 (21%)
Nine months | Normal | 11 (55%) | 0 (0%) | p = 0.19N = 20
Reduced | 7 (35%) | 2 (10%)

We also compared median sPD-L1 levels between patients who relapsed within 12 months of follow-up and those who did not, as shown in Table 3. There was a trend toward higher baseline sPD-L1 in patients who relapsed within 12 months of follow-up (p = 0.052), compared to those who did not relapse.

Blood sampling timepoint | Soluble PD-L1 levels—pg/mL | Did not relapse < 12 months follow-up | Did relapse < 12 months follow-up | Total, N | Wilcoxon rank sum
Baseline | Patients, n | 40 patients | 26 patients | 66 patients | p = 0.05
Median (IQR) | 69.5 (34.5) | 85 (54) | 75.5 (40)
First control | Patients, n | 39 patients | 20 patients | 59 patients | p = 0.31
Median (IQR) | 75 (29) | 82.5 (62.5) | 76 (36)

ctDNA status and risk of relapse within 12-month follow-up

Twenty-seven patients had positive baseline samples. This dropped to eight patients at the first control sampling (two of those were not positive at baseline). Fourteen of the twenty-one relapsing stage II–III patients had positive ctDNA status in the last blood sample before the relapse. In contrast, none of the five stage I relapsing patients had any positive ctDNA measurements.

The association between ctDNA status (negative vs. positive) at the different blood-sampling timepoints and the risk of relapse within 12 months of follow-up was also evaluated (Table 4). Patients with a positive ctDNA status at baseline (p = 0.03) or at 6 months (p < 0.01) had a significantly higher risk of relapse within 12 months.

Blood sampling timepoint | ctDNA status | Patients without relapse < 12 months follow-up, n (%) | Patients with relapse < 12 months follow-up, n (%) | Pearson chi2
Baseline | Negative | 28 (42%) | 11 (17%) | p = 0.03N = 66
Positive | 12 (18%) | 15 (23%)
First control (5 weeks or 3 months) | Negative | 35 (59%) | 16 (27%) | p = 0.70N = 59
Positive | 5 (8%) | 3 (5%)
Six months | Negative | 26 (72%) | 2 (6%) | p < 0.01N = 36
Positive | 2 (6%) | 6 (17%)
Nine months | Negative | 20 (80%) | 2 (8%) | p = 0.33N = 25
Positive | 2 (8%) | 1 (4%)

Baseline ROC analyses for the risk of relapse within the first 12 months following treatment

We performed ROC analyses of the four methylated ctDNA markers, sPD-L1 values, and NKA at baseline prior to RT to assess their ability to discriminate patients who relapsed within 12 months of follow-up. The analyses were done in a stepwise fashion (Fig. 2).

Fig. 2: Stepwise ROC analysis of baseline biomarkers and the risk of relapse within 12-month follow-up

The discriminative value for each of the four methylated ctDNA markers at baseline was modest, with MCIDAS AUC value 0.65 [95% CI 0.51–0.78], SP9 AUC 0.66 [0.53–0.79], HOXA9 AUC 0.68 [0.54–0.82], and TFAP2B AUC 0.68 [0.56–0.80]. The discriminative value for the binary NKA marker had an equivalent AUC value of 0.68 [95% CI 0.55–0.81]. The continuous sPD-L1 marker had a likewise modest AUC of 0.64 [95% CI 0.50–0.78].

When combining two markers, AUC rose to 0.77 [95% CI 0.63–0.90] for the combination of HOXA9 and NKA. The addition of extra markers consistently increased the AUC. As seen in Fig. 2, all four methylated ctDNA markers contributed a little to the increase in AUC. However, with five or six markers, there was no difference in the AUC whether MCIDAS was included or not—the AUC was 0.86 [95% CI 0.74–0.98] in both cases. Using DeLongs test to compare the AUC of 1 marker (TFAP2B) with the AUC of 6 markers, the obtained p value was 0.068.

Between 6- and 12-month follow-up, only 8 out of 36 sampled patients had a relapse, so the data material was too small to allow meaningful ROC analyses.

Figure 3 shows a swimmer plot with timing and results of blood samples, relapse, and death for each of the patients treated with long-course RT. See Online Resource A2 for the same plot with the patients treated with SBRT. The time of relapse is the date of the CT scan revealing the relapse. As the patients often had the blood sampling performed during the hospital visit where they received their CT-scan results, this explains why their relapse came a short time before the final blood sampling.

Fig. 3: Swimmer plot with patients receiving long-course RT. NKA Natural Killer-cell activity. PD-L1 Programmed death-ligand 1. ctDNA Circulating tumor DNA