Prognostic value of tumor markers ProGRP, NSE and CYFRA 21-1 in patients with small cell lung cancer and chemotherapy-induced remission

2.1Study population and design

This prospective single-centre study enrolled patients from Thoraxklinik, University Hospital Heidelberg, a specialized lung cancer center. Eligible patients had pathologically confirmed small cell lung cancer at any Union for International Cancer Control (UICC) stage, an Eastern Cooperative Oncology Group Performance Status score of 0-2, a and planned treatment with at least four cycles of chemotherapy. Patients with unobtainable blood samples, a history of secondary malignancy, or were pregnant or breastfeeding were excluded. Patients received either platinum-based doublet chemotherapy as first-line treatment or an appropriate single-agent cytotoxic treatment for any subsequent lines of therapy, as directed by their physician. Blood samples were collected from each patient before treatment and after each treatment cycle for the analysis of biomarkers. Radiological assessment of treatment response was performed by CT after every two cycles of chemotherapy, with tumor response categorized as partial response (PR), stable disease (SD), progressive disease (PD), or ‘other’ based on the response evaluation criteria in solid tumors (RECIST v1.1 [19]). Follow-up visits after therapy occurred every 4-6 weeks, and OS or censoring of data was documented at the last known date available.

The study was approved by the ethics committee of the University of Heidelberg (MZmu-446/2012 and MZmu 067/2013) and was performed in accordance with the Declaration of Helsinki and Good Clinical Practice. All patients provided written informed consent prior to any study-related procedures.

2.2Blood samples and biomarker analyses

Venous blood samples were collected into serum-gel monovettes (7.5 ml, Sarstedt, Nürmbrecht, Germany) and allowed to clot for at least 30 minutes at room temperature. Samples were then centrifuged at 2000 x g at 10°C within 1 to 2 hours of venepuncture. The resulting serum was divided into 500μL aliquots and stored below –70°C until analysis. The samples were shipped on dry ice to the Institute of Laboratory Medicine at the German Heart Centre Munich for analysis. Samples were thawed and processed in batches. Measurements were performed using a panel of electrochemiluminescence immunoassays (ECLIA) for in vitro diagnostics biomarkers (ProGRP, NSE, CYFRA 21-1, CEA) on a cobas^® e 411 analyzer (Roche Diagnostic Solutions, Mannheim, Germany). All analyses were performed in a certified laboratory in accordance with the quality criteria outlined in the Guidelines of the German Federal Association. Laboratory data for patient samples and quality controls were electronically transferred for statistical evaluation. Clinical data were extracted from the patients’ medical records and recorded in electronic case report forms.

2.3Statistical analyses

All statistical analyses were performed using R (version 3.6.3; www.r-project.org). Patient demographics, disease characteristics and biomarker data were summarized for all patients using descriptive statistics. Biomarker dynamics were measured for all patients, while the correlation between OS and biomarker values was limited to the cohort of patients with advanced disease (UICC stages III and IV). The prognostic utility of the biomarkers was evaluated at three timepoints: (1) prior to first-line treatment in all patients with advanced disease, (2) after two cycles of treatment in patients who achieved a PR during first-line therapy, and (3) prior to second-line treatment in patients who either did not respond to first-line treatment or initially responded but subsequently experienced disease progression. Overall survival (OS) was defined as the time from the three timepoints described above to death from any cause.

The Kaplan-Meier method was used to estimate OS in subgroups defined by biomarker levels, with 95% confidence intervals (CIs) calculated for the median OS. Hazard ratios (HR) were calculated using Cox regression models. Biomarker cut-offs were optimized separately at each time point to obtain maximum risk differences (maximum HR between responder (defined as PR) and non-responder groups (defined as SD or PD), with all quantiles from 0.2 to 0.8 by steps of 0.05 tested to split the patients into two groups. These values were selected to avoid any group being inappropriately small, which could result in overfitting of the cut-off and the Cox regression models. The log-rank p-value was calculated for all possible cut-offs, and the quantile with the lowest value was chosen as the optimized cut-off. For analyses that involved combinations of biomarkers, the cutoffs determined for the individual biomarkers were used.

To determine whether the ProGRP pre-treatment values or the decrease from baseline to after the second treatment cycle were more relevant for OS, both categories were included in a single graph, and survival time was partitioned into quartiles. Individual cases with a very good prognosis (defined as OS above Q3), a medium prognosis (OS in Q1 to Q3) and a poor prognosis (OS below Q1) were depicted individually.

3.1Patient characteristics

In total, 232 patients were enrolled, including 12 patients with early-stage disease, 25 with stage IIIA, 48 with stage IIIB and 147 with stage IV disease. Detail baseline characteristics for the patients with advanced (stage III/IV) disease (n = 220) are displayed in Table 1. The majority of the patients with advanced disease received platinum-based or other doublet first-line chemotherapy and 12 received radiotherapy. At staging exams, 146 patients achieved PR (66.4%), 22 had SD (10.0%) and 19 had PD (8.6%). Twenty-seven patients discontinued the study before staging, five did not receive chemotherapy (best supportive care only) and one patient could not be evaluated. Ninety-five patients initiated a second-line therapy. Most patients with response to therapy received six cycles of chemotherapy while non-responsive patients switched to a second-line treatment early in their care.

3.2Biomarker dynamics

Biomarker levels were highly variable prior to first-line treatment as well as during and after treatment (Fig. 1). Pre-treatment (visit 1) levels ranged from 11 to 174,000 pg/mL for ProGRP, from 9.5 to 1160.4 ng/mL for NSE, from 0.6 to 71.5 ng/mL for CYFRA 21-1, and from 0.2 to 8134 ng/mL for CEA (Table 2). During treatment, median values of ProGRP decreased rapidly from 1405 pg/mL pre-treatment to 411 pg/mL after the first treatment cycle and 142 pg/ml after the second treatment cycle (first CT imaging) (Fig. 1 and Table 2). Subsequently, median levels remained stable until after completion of the first-line therapy, when they began to increase again. Similar dynamics were observed for NSE, for which median levels decreased from 76.9 ng/ml (visit 1) to 28.7 ng/ml (visit 2) and to 20.7 ng/ml (visit 3), remained stable and increased again from visit 7. Although at a lower level, similar dynamics were also seen for CYFRA 21-1, for which median levels decreased from 3.1 ng/ml (visit 1) to 1.9 ng/ml (visit 2) and to 1.7 ng/ml (visit 3), remained stable and increased slightly again starting from visit 7. Changes in CEA were similar but less pronounced (Fig. 1 and Table 2). From pre-treatment to after the first treatment cycle, the median percentage decrease in biomarker levels was ∼45% for ProGRP, ∼60% for NSE and ∼30% for CYFRA 21-1. Levels of CEA remained relatively stable from pre-treatment to after the first treatment cycle (Fig. 2A). After the second treatment cycle, median levels of ProGRP decreased further by ∼40%, while decreases in NSE were only around 20%, and both CEA and CYFRA 21-1 remained relatively stable (Fig. 2B).

Fig. 1

Distribution of tumor biomarker levels at pre-treatment (visit 1) and after each treatment cycle (visits 2-10) for (A) ProGRP, (B) NSE, (C) CYFRA 21-1 and (D) CEA. Plots show the median, interquartile range, minimum and maximum. Outlier points are defined as values 1.5 times the interquartile range.

3.3Correlation of single biomarkers with overall survival

Higher pre-treatment levels of ProGRP, NSE, and CYFRA 21-1 were associated with significantly worse OS, with hazard ratios ranging from 2.2 to 2.9 for median survival times. However, these analyses also included patients who had no or only limited response (PD and SD) to first-line therapy and thus were biased towards a more unfavourable prognosis (Fig. 3A-C). In a cohort limited to patients who achieved at least a PR during first-line therapy, high absolute biomarker levels measured at the time of the first CT imaging (after two treatment cycles) were also associated with significantly worse OS, with hazard ratios ranging from 1.9 for CYFRA 21-1 to 2.7 for ProGRP (Fig. 3D-F). For patients receiving second-line therapy after recurrence or disease progression after first-line therapy, absolute biomarker values measured at the visit prior to initiation of second-line therapy were also prognostic for OS (data not shown). Biomarkers were ranked at each timepoint based on the C-index and HR for median OS. The top-ranking biomarkers at each timepoint were as follows: CYFRA 21-1 (C-index 0.64 [95% CI, 0.60–0.68]; HR 2.9 [95% CI, 2.06–4.01) had the highest prognostic value when measured prior to first-line treatment, ProGRP (C-index 0.63 95% CI, 0.60–0.68]; HR 2.7 [95% CI, 1.84–3.98]) had the highest prognostic value after two treatment cycles in patients with PR after first-line therapy, and NSE (C-index 0.70 [95% CI, 0.64–0.75]; HR 3.6 [95% CI, 2.23–5.87]) had the highest prognostic value when measured prior to initiation of second-line therapy in patients with recurrence or progression after first-line therapy.

Fig. 2

Relative changes in tumor biomarker levels. Percent change in ProGRP, NSE, CYFRA 21-1 and CEA from (A) pre-treatment (visit 1) to after the first treatment cycle (visit 2) and (B) from after the first treatment cycle to after the second treatment cycle (visits 2 to 3).

Fig. 3

Correlation between tumor biomarker level and overall survival. Kaplan-Meier survival curves for patients in high-risk (red) and low-risk (green) subgroups defined by ProGRP, NSE and CYFRA 21-1 levels measured either pre-treatment (A-C) or after two treatment cycles (D-F) using optimized cutoffs.

The prognostic value of absolute baseline biomarker values versus relative change from pre-treatment to after the second treatment cycle for OS were also evaluated. The majority of patients classified as having a very good prognosis (OS above quartile 3) had low pre-treatment ProGRP levels (<200 pg/ml) and a large decrease after two treatment cycles (>50%) (Fig. 4). In contrast, most patients with a poor prognosis (OS below quartile 1) had high pre-treatment ProGRP levels (>200 pg/ml) or a less pronounced decrease after two treatment cycles (<50%). For patients with medium prognosis (Q1 to Q3) there was no clear pattern in ProGRP levels, either pre- or post-treatment.

Fig. 4

Scatter plot for overall survival in relation to pre-treatment ProGRP values and change over time. Overall survival times are partitioned in quartiles and patients with very good prognosis (above Q3, green dots), medium prognosis (between Q1 and Q3, yellow dots) and poor prognosis (below Q1, red dots) are shown. X-axis indicates ProGRP values pre-treatment. Y-axis indicates percent change in ProGRP from pre-treatment to after two treatment cycles.

3.4Correlation of a triple biomarker combination with overall survival

As ProGRP, NSE and CYFRA 21-1 all showed prognostic relevance, the combination of the three biomarkers was compared with the individual markers (Fig. 5). For the values obtained before treatment, the triple biomarker combination exhibited a higher prognostic value than the best single marker CYFRA 21-1 (C-index: 0.65 vs 0.64; HR: 3.2 vs 2.9). Similarly, when measured after two treatment cycles, the triple biomarker combination outperformed the best single marker ProGRP (C-index: 0.66 vs 0.63; HR: 3.7 vs 2.7). Finally, for the values obtained prior to initiation of second-line therapy, the triple biomarker combination was equivalent to the best single marker NSE (C-index: 0.70 vs 0.70; HR: 3.6 vs 3.8).

Fig. 5

Three-marker combination of ProGRP, NSE and CYFRA 21-1 for the estimation of overall survival at three timepoints. Kaplan-Meier survival curves for patients in high-risk (red) and low-risk (green) subgroups defined by a combination of ProGRP, NSE and CYFRA 21-1 levels using optimized cutoffs. Tumor biomarkers were measured (A) pre-treatment in all patients, (B) after two treatment cycles in patients with a PR to first-line treatment, and (C) prior to second line treatment. For each graph, the starting timepoint reflects the time of blood drawing. Figure legend indicates the optimized cutoffs used for each timepoint.

3.5Monitoring progressive disease during surveillance

The utility of biomarkers to detect recurrence or progressive disease after successful first-line therapy was illustrated by evaluating biomarker dynamics in relation to the appearance of clinical events in individual cases. As shown in the following examples (Fig. 6A-C), ProGRP, NSE and CYFRA 21-1 values all indicated a clear biochemical response during therapy and sensitively and promptly detected new clinical events during the surveillance phase.

Fig. 6

Longitudinal tumor biomarker courses for individual patients. (A) 67-year-old male smoker with stage IV SCLC treated with four cycles of a combination of carboplatin, etoposide and ipilimumab followed by two cycles of maintenance ipilimumab plus radiation therapy of the brain 30 Gy (1). After disease progression, second-line therapy included six cycles of topotecan combined with 30 Gy radiation of neck vertebra (2) and a further irradiation of the brain (28 Gy) and neck vertebra (24 Gy) (3). The patient showed further tumor progression and died 547 days after treatment start. ProGRP, NSE and CYFRA 21-1 decreased dramatically at the beginning of first-line therapy but increased at the time of tumor progression. During second-line treatment temporary tumor control was mirrored by a moderate decrease in ProGRP. At the time of further progression, all biomarkers and particularly ProGRP increased again. (B) 57-year-old female smoker with stage IV SCLC treated with six cycles of first-line carboplatin and etoposide with an additional PCI of 30 Gy between days 18 to 31. Along with radiological response to therapy, levels of ProGRP, NSE and CYFRA 21-1 reduced considerably and continuously. Subsequent early increases in ProGRP indicated progressive disease with a lead time of about 60 days. Second-line treatment with topotecan was only applied for two cycles as the patient further progressed paralleled by a further increase of ProGRP. (C) 58-year-old female smoker SCLC stage IV treated with six cycles of a first-line combination of cisplatin and etoposide followed by 30 Gy PCI (1). All biomarkers were strongly elevated pre-treatment, particularly ProGRP and NSE, and reduced quickly and completely during successful therapy when radiological examinations showed partial remission. After completion of primary therapy all markers rapidly increased again and anticipated the later radiological progression by around 60 days. Second-line therapy consisted of two cycles of carboplatin and etoposide, complemented by simultaneous irradiations of the brain (24 Gy) and the bones (30 Gy; 2). Thereafter chemotherapy was changed to three cycles of vincristine and etoposide. Effective tumor control was accompanied by decreasing and stable biomarker levels during the treatment. Shortly after completion of treatment, biomarker values increased again. While radiological examinations still indicated stable disease, the patient deteriorated and died 535 days after treatment initiation.