The utilization pattern of serum tumor markers in lung cancer patients: A population‐based retrospective descriptive study

Abstract Background The trends in usage of tumor markers, including CEA, SCC, NSE, Cyfra21‐1, and ProGRP, in Chinese lung cancer patients in the real‐world setting are not fully investigated. Methods A retrospective descriptive study was conducted using the database of Qilu Hospital of Shandong University, China between January 2013 and December 2017, involving patients primarily diagnosed with NSCLC or SCLC. Utilization trends by first discharge year, utilization rates within different durations before and after first discharge date, and combined utilization patterns of multiple tumor markers were analyzed. Results The utilization of all these tumor markers showed increased from 2013 to 2017. CEA, Cyfra21‐1, and NSE were the most frequently detected, which increased slightly from around 50% in 2013 to around 78% in 2017 in NSCLC and from around 70% in 2013 to around 92% in 2017 in SCLC. CEA, Cyfra21‐1, and NSE were the most commonly measured within 3 months before first diagnosis with approximately 65% in NSCLC and 80% in SCLC, and ProGRP had the lowest utilization (around 30%). CEA, NSE, and Cyfra21‐1 had the highest utilization rates after first diagnosis with both around 80% in NSCLC or SCLC. Combined usage of five tumor markers was ranked the first pattern in combined utilization. Conclusions This study suggests CEA, Cyfra21‐1, and NSE are the most frequently detected before or after first diagnosis of NSCLC or SCLC. However, SCC and ProGRP tests appeared to have relatively low usages. The utilization pattern was consistent with recommendations of guideline, but underutilization still existed.


| INTRODUC TI ON
Lung cancer is one of the most common cancers globally, accounting for an estimated 2.09 million new cases and 1.76 million deaths in 2018, ranking first among all cancer types regardless of incidence and mortality. 1 Based on data from 339 cancer registries in China, the age-standardized incidence rate of lung cancer was 36.71 per 100 000 and the age-standardized mortality rate for lung cancer was 28.49 per 100 000 in China. 2 It is estimated that lung cancer mortality in China may increase by approximately 40% between 2015 and 2030. 3 Tumor markers are biomarkers found in blood, urine, or body tissues that can be elevated by the presence of one or more types of cancer. 4 They contribute usefully to patient management including to guide treatment decisions, to verify the effect of treatment, to predict the chance of recovery, to predict or watch for recurrence, to diagnosis of specific tumor types, and to screen for common cancers on a population basis. 5 Commonly used primary lung cancer markers that are currently recommended by the American Association for Clinical Biochemistry and the European Group on Tumor Markers include carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), cytokeratin fragment (Cyfra21-1), pro-gastrin-releasing peptide (ProGRP), and squamous cell carcinoma antigen (SCC). 6 The combined detection of these tumor markers can improve the sensitivity and specificity of assessments in clinical practice. They are selectively detected in auxiliary diagnosis, efficacy monitoring, and follow-up. 7 According to the National Academy of Clinical Biochemistry Guidelines for the use of Tumor Markers in Lung Cancer, recommended markers are Cyfra21-1 and CEA before therapy and Cyfra21-1 and/or CEA in post-therapy follow-up in patients with adenocarcinoma, squamous cell carcinoma, and large cell carcinoma. In SCLC patients, recommended markers are NSE and ProGRP before therapy and NSE and/or ProGRP in post-therapy follow-up. 8 One recent study described a retrospective analysis to evaluate the frequency of serum tumor marker use in patients with advanced solid tumors. 9 They found that a high rate of serum tumor marker testing use, and carbohydrate antigen 19-9 (CA199) and CEA were the most commonly overused tests. 9 Another study assessed real-world patterns of epidermal growth factor receptor (EGFR) testing and associated treatment and outcomes among non-small cell lung cancer (NSCLC) population. 10 They found that relatively low rate of NSCLC patients received EGFR testing and significant disparities in testing were observed by different patient characters. 10 However, relevant evidence on usage of lung cancer-related tumor markers was lack in Chinese lung cancer patients in routine clinical practice, and whether this pattern was consistent with guideline recommendations for tumor markers in lung cancer is unknown. In the current study, we explored the trends in usage of lung cancer-related tumor marker testing in patients with NSCLC and small cell lung cancer (SCLC) in routine clinical practice.

| Study design and setting
We conducted a descriptive study in Shandong province, China using clinical Laboratory Information System (LIS) database and Hospital Information System (HIS) database in Qilu Hospital of Shandong University. Q ilu Hospital is a Grade A comprehensive hospital located in Shandong province and established in 1890, which was the top 20 hospitals in China. In 2018, the Qilu Hospital had approximately 3 800 000 annual outpatient and emergency visits, 210 000 annual hospitalizations, and 97 000 annual procedures.

| Data source
Data were obtained from laboratory and hospital information systems, a real-time system that electronically captures administrative data, clinical data, and laboratory data on patients visiting the Qilu hospital. The systems store routinely collected healthcare data from 2008 to current, which includes demographic data, hospitalization data, laboratory data, prescription data, procedure data, and imaging data. Each patient attended to the hospital was allocated a unique identifying number, which could link individual records across multiple systems.
The HIS database stores hospital discharge data, which contains some demographic characteristics of hospitalized patients, principal conditions, major medical procedures, pathology diagnosis, and hospitalization outcomes. Medical diagnostic information has been coded according to the International Classification of Disease, Tenth Revision (ICD-10). Clinical laboratory database is held and maintained by clinical laboratory department, which contain biochemical, hematology, microbiology, virology, and serology data. Laboratory database can be record-linked to HIS database for each patient encounter using unique patient identifier. Data were anonymized for the purposes of research that follows national healthcare big data standards, safety, and service management approach. This project was approved by Qilu hospital committee on research medical ethics.

| Study population
All patients aged 18 years or older visited the Qilu hospital between January 1, 2013, and December 31, 2017, were eligible for inclusion.
Patients may have had multiple visits to the hospital during the study period. All visits were included.

| Study subjects
Study subjects were those with a primary diagnosis of lung cancer between January 2013 and December 2017. They were identified from HIS database with coded C34.0, C34.1, C34.2, C34.3, C34.8, C34.9 according to ICD-10. 11 Patients were categorized into NSCLC and SCLC patients based on pathology records. NSCLC patients were further classified into adenocarcinoma, squamous cell carcinoma, large cell carcinoma, and undifferentiated NSCLC according to pathology records.

| Variables
Administrative data, clinical data, and laboratory data were extracted from data source. Age at diagnosis was defined as from the date of first discharge from hospital with primary diagnosis of NSCLC or SCLC to the date of birth. First discharge year was defined as the year of first discharge from hospital with primary diagnosis of NSCLC or SCLC. Length of stay (LOS) was defined as from the date of patient admission to the date of patient discharge from the hospital in each hospitalization. Observation window of tumor marker utilization was defined according to cutoffs of observation windows recommended by guideline. 7 The utilization rate of tumor marker in one observation window was calculated as numbers of patients with one tumor marker testing divided by numbers of patients who were in follow-up during this observation window. The frequency of combined testing of tumor markers was observed in three periods, which include the whole study period, the period of before first diagnosis of NSCLC or SCLC, and the period of after first diagnosis of NSCLC or SCLC.

| Statistical methods
Data were summarized as mean (SD) for continuous variables and number of subjects (percentage) for categorical variables. Patientrelated demographic and clinical characteristics, utilization distribution of each tumor marker, and combined utilization distribution of multiple tumor markers were demonstrated by standard descriptive statistics. Utilization frequencies and rates within different observation windows were expressed by standard descriptive statistics. All analyses were carried out using R 3.5.1.

| RE SULTS
The lung cancer cohort consisted of 3443 NSCLC patients and 489 SCLC patients. Among NSCLC and SCLC patients, the majority were male, aged 45-64 years old, with only once hospitalization, and with LOS from 8 to 14 days (Table 1).
In NSCLC and SCLC patients, tumor marker utilization showed increased trends from 2013 to 2017 ( Figure 1). CEA, Cyfra21-1, and NSE had similar utilization rates and increased slightly from around 50% in 2013 to around 78% in 2017 in NSCLC patients ( Figure 1A) and from around 70% in 2013 to around 92% in 2017 in SCLC patients ( Figure 1B). Also, the utilization rate of ProGRP increased dramatically from 19% in 2015 to 69% in 2017 for NSCLC ( Figure 1A) and from 25% in 2015 to 82% in 2017 for SCLC ( Figure 1B). SCC utilization rates were from 43% in 2013 to 70% in 2017 in NSCLC ( Figure 1A) and from 64% to 78% in SCLC ( Figure 1B) Figure 2).
The utilization rate and test times of tumor markers before 3 months before first discharge of NSCLC and SCLC patients were displayed as Table 2 and 3, respectively. Generally, both the utilization rate and test times of tumor markers were increasing from 2013 to 2017, and SCLC patients had higher utilization rate and more test times than NSCLC patients. CEA, Cyfra21-1, and NSE were obviously more utilized than ProGRP and SCC in both NSCLC and SCLC patients.
The usage of CEA, Cyfra21-1, and NSE tests was similar within 3 months before first discharge with diagnosed different subtypes of NSCLC, and all utilization rates were approximately 60% ( Figure 3).
ProGRP test had the lowest utilization rate before first diagnosis of NSCLC. In addition, compared with NSCLC, the utilization rates within 3 months before first discharge with diagnosed SCLC were relatively high with around 75% for CEA, Cyfra21-1, and NSE, 64.6% for SCC, and 30.7% for ProGRP (Figure 3).
Over 70% of NSCLC and 60% of SCLC patients had become lost to follow-up after 3 months of first discharge from hospital with diagnosed NSCLC or SCLC (Appendix Figure S1). Among patients with follow-up, the utilization rates of each tumor marker in different follow-up durations were relatively stable ( Figure 4). CEA, NSE, and  Figure S2B) and 503 (29%) in SCLC (Appendix Figure S2D).

F I G U R E 2
Test times for each tumor marker among patients with NSCLC or SCLC during the study period was greatly higher than that of SCLC, the above situation might be more common in NSCLC patients. This speculation could also be verified by the follow-up results, in which we found that although the untested percentage of NSCLC patients was higher than that of SCLC patients in the first two discharges, the difference gradually decreased to non-significance in the following follow-ups.  Studies in patients with good follow-up rates are required to investigate the utilization pattern. Also, treatment information was not available for this study due to data access limitation. The lack of information has an influence on full description of the utilization pattern of these tumor markers.
In conclusion, we explored the utilization patterns of lung cancer-related serum tumor markers in Chinese NSCLC or SCLC pa-