Total, bioavailable and free 25-hydroxyvitamin D are associated with the prognosis of patients with non-small cell lung cancer

To analyze the prognostic value of total, bioavailable and free 25-hydroxyvitamin D [25(OH)D] as well as vitamin D-binding protein (VDBP) in patients with non-small cell lung cancer (NSCLC). We prospectively collected and analyzed data for 395 patients diagnosed with NSCLC between January 2016 and December 2018 in two university-affiliated hospitals. Total and free 25(OH)D and VDBP were measured directly, and bioavailable 25(OH)D was calculated using a validated formula. Their prognostic values were evaluated by Cox proportional hazards model, and hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated. Patients with NSCLC had significantly lower levels of total, bioavailable, and free 25(OH)D and higher VDBP levels in comparison to healthy controls (all p < 0.001). In multivariate analyses, higher levels of total, bioavailable, and free 25(OH)D were independently associated better overall survival (OS) and progression-free survival (PFS). For OS, the adjusted HRs were 0.58 (95% CI, 0.40–0.87; p for trend = 0.008), 0.45 (95% CI, 0.30–0.67; p for trend < 0.001) and 0.49 (95% CI, 0.33–0.73; p for trend < 0.001) for the highest versus the lowest tertile of total, bioavailable and free 25(OH)D, respectively. The corresponding adjusted HRs for PFS were 0.61 (95% CI, 0.43–0.86; p for trend = 0.006), 0.56 (95% CI, 0.40–0.80; p for trend = 0.001) and 0.60 (95% CI, 0.42–0.85; p for trend = 0.004), respectively. However, VDBP was not associated with either OS or PFS. The current study suggested that total, bioavailable and free 25(OH)D may be reliable prognosis indicators in NSCLC patients, though the optimal 25(OH)D form for NSCLC prognosis remains to be assessed in future studies.


Introduction
Lung cancer is the second most common cancer with the highest mortality rate worldwide [1]. Non-small cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancer [2]. Although a wide range of treatments have been clinically available [3], the prognosis remains poor, with a 5-year survival rate of 19.8% in China [4]. It is of great clinical significance to explore valuable and adequate prognostic indicators to further guide individualized disease management in NSCLC.
Vitamin D, traditionally known as an essential nutrient, is a precursor of a potent steroid hormone that regulates a broad spectrum of physiological processes [5]. Increasing evidence indicates a significant anti-tumor effect of vitamin D through inhibiting cell proliferation, angiogenesis, and metastasis, while inducing apoptosis and differentiation [6]. However, vitamin D deficiency is evident globally and even more severe in cancer patients, including patients with lung cancer [7,8]. Low levels of circulating vitamin D concentrations have been associated with poorer survival in NSCLC, but the findings are inconsistent [9][10][11][12][13].
As the major vitamin D circulating form in the human organism, total serum or plasma 25-hydroxyvitamin D [25(OH)D] is currently routinely used in clinical practice to assess vitamin D status [14]. About 85-90% of total 25(OH)D circulates bound to vitamin D-binding protein (VDBP), and this proportion is recognized as the biologically inactive form. The non-VDBP-bound fraction, consisting primarily of albumin-bound 25(OH)D (10-15%) and less than 0.03% of total 25(OH)D in the free form, is considered as bioavailable 25(OH)D. The free and albumin (with a 1,000-fold weaker affinity for 25(OH)D than VDBP) forms are easier to be taken up by cell tissues, thereby exerting its biological actions [15]. Recent studies have reported that bioavailable or free 25(OH)D may be more representative indicators for vitamin D status as compared with total 25(OH)D [16,17]. Current studies on vitamin D and lung cancer have been focused on total 25(OH)D and, to our knowledge, the relationships between circulating bioavailable and free 25(OH)D levels and prognosis of NSCLC are yet to be assessed.
In addition, VDBP may influence cancer progression in addition to being a carrier for vitamin D metabolites [18]. Therefore, we aimed to investigate the associations of total, bioavailable, and free 25(OH)D and VDBP with survival in NSCLC patients.

Study population
The Soochow Lung Cancer Cohort (SLCC) study is an ongoing patient cohort conducted in the first affiliated hospital and the second affiliated hospital of Soochow University. This study was designed to investigate the long-term prognosis factors among lung cancer patients in Southeast China. As described previously [19], a total of 525 primary lung cancer patients aged ≥ 18 years were enrolled in the cohort between January 2016 and December 2018. For the current study, we excluded patients with small cell lung cancer (n = 62). We further excluded patients lost to follow-up (n = 29) and those with no sufficient data (n = 39). Finally, the current study included 395 NSCLC patients. To compare the vitamin D indices with controls, 50 healthy volunteers (25 men and 25 women with a mean age of 60 years) were enrolled by Suzhou Industrial Park Centers for Disease Control and Prevention in October 2020.

Data collection
An in-person interview was conducted by trained investigators using a structured questionnaire within 1 week after diagnosis to collect general information about age, gender, disease history [e.g., chronic obstructive pulmonary disease (COPD)], family history of cancer, body mass index (BMI), in addition to behavioral factors such as smoking status and alcohol consumption. Participants were asked to recall the frequency of alcohol consumption during the year prior to the baseline assessment. According to the obtained information, smoking status was categorized as never, former, or current smoker. Current smokers were defined as patients who had smoked continuously or accumulated for at least 6 months and continued to smoke during the baseline survey or those quitting smoking for less than 1 year. Former smokers were defined as those who had quit smoking for more than 1 year at the study entry. Furthermore, medical records were also reviewed to extract their clinical data, including cancer characteristics (TNM stage, histology, lesion and laterality), treatments (surgery, chemotherapy, radiotherapy and targeted therapy), and the levels of various blood analytes. Blood analytes included albumin, total cholesterol, neutrophils, lymphocytes, monocytes, carcinoembryonic antigen (CEA) and neuron-specific enolase (NSE). Neutrophil-tolymphocyte ratio (NLR) and lymphocyte-to-monocyte ratio (LMR) were calculated based on the collected data.

Determination of 25(OH)D fractions and VDBP
A fasting plasma sample was collected from each patient at the time of recruitment. All samples were numbered and stored at − 80 °C until further analysis. The plasma circulating total 25(OH)D (D 2 and D 3 ) levels were measured with the use of liquid chromatography-mass spectroscopy (LC-MS/MS) at Suzhou Harmony Health Medical Laboratory Co., Ltd. The mean intra-assay coefficient of variation (CV) was 1.7%. VDBP and free 25(OH)D were measured by commercial enzyme-linked immunosorbent (ELISA) assays (R&D Systems, Minneapolis, MN, USA and DIAsource ImmunoAssays, Louvainla-Neuve, Belgium, respectively). The mean CV values were 5.3% and 6.1% for VDBP and free 25(OH)D, respectively. Bioavailable 25(OH)D ([Bio D]) was calculated from measured free 25(OH)D, albumin, and affinity constant between 25(OH)D and albumin using the following formula [16]:

Follow-up
Follow-up of patient outcomes began at the date of enrollment until the last follow-up date (December 2020) or until the date of death. The survival outcomes were evaluated semiannually. Information sources included the hospital inpatient or outpatient records, patient or family telephonic contact, in addition to local death registration system. Overall survival (OS) was defined as the time from the date of enrollment to the date of death, and progression-free survival (PFS) was defined as the time from the date of enrollment to the date of confirmed progressive disease (PD) or death.

Statistical analysis
The 25(OH)D fractions (total, bioavailable, and free) and VDBP levels were categorized into three groups by the tertiles of the variables, and the Spearman correlation test was used to test their pairwise associations. One-way analysis of variance (ANOVA) with Tukey's Honestly Significant Difference (HSD) pairwise comparisons were used to evaluate the differences of the 25(OH)D fractions and VDBP levels among normal group, NSCLC survivors, and NSCLC non-survivors. The receiver operating characteristic (ROC) curve was used to determine the optimal cutoff value of the NLR and LMR. The chi-square test or Fisher's exact test was applied to explore differences in categorical variables. The Kaplan-Meier method with log-rank test was used to construct survival curves. Univariate and multivariate Cox proportional hazards regression models were used to identify variables associated with OS and PFS. Hazard ratios (HR) and 95% confidence intervals (CI) were then calculated for OS and PFS. Baseline variables that were considered clinically relevant or that showed a univariate relationship with outcome were entered into multivariate analyses. In order to reduce the impact of lifestyle alteration for the outcomes, additional sensitivity analysis was performed after excluding patients who were deceased in the first 6 months of the follow-up. p values for the linear trends were calculated by entering the median values for each tertile of 25(OH)D or VDBP into the models as a continuous variable. Further stratified multivariate analyses were performed according to age, gender, smoking, TNM stage, histology, lesion and surgery, while we also recognized that these analyses are subject to limited statistical power. Interactions were evaluated using the Wald test. Time-dependent ROC curves for the prognostic values of 25(OH)D fractions and VDBP were estimated using the R package survivalROC. All analyses were performed by IBM SPSS 25.0 and R software version 3.6.3. Two-sided p value < 0.05 was considered statistically significant.

Baseline clinical characteristics
Of the 395 NSCLC patients included in the study, the mean age was 63.0 ± 10.5 years, and 63.8% of the participants were men, 49.4% were former/current smokers and 65.1% were classified as stage IIIB-IV (Table 1). The associations between baseline clinical characteristics and total 25(OH) D are summarized in Table 1. Patients with higher total 25(OH)D levels had higher levels of albumin and LMR, and were more likely to be collected for the blood samples in spring and autumn and to receive surgery or chemotherapy. Meanwhile, patients with higher bioavailable 25(OH)D levels had higher levels of BMI, albumin and LMR. Additionally, both bioavailable and free 25(OH)D were associated with the season of blood-drawing (Supplementary Table 1).

Plasma levels of 25(OH)D Fractions and VDBP in NSCLC patients
Plasma levels of total, bioavailable, and free 25(OH)D and VDBP in control group and NSCLC patient group are presented in Fig. 1. Mean total 25(OH)D concentration was 20.78 ± 7.81 ng/mL for all NSCLC patients. According to the reference ranges of clinical practice guideline [20], 37% of the patients were considered to be 25(OH)D insufficient (20 to 30 ng/mL) and 50.1% were deficient (< 20 ng/ mL). Compared with healthy controls, NSCLC patients had significantly lower levels of total, bioavailable and free 25(OH)D (25.56 ± 8.75 vs. 20.78 ± 7.81 ng/mL, 5.82 ± 1.06 vs. 2.87 ± 1.35 ng/mL, and 16.13 ± 1.82 vs. 7.69 ± 3.41 pg/ mL, respectively, all p < 0.001), while had higher VDBP levels (175.69 ± 35.24 vs. 204.93 ± 50.47 μg/mL, p < 0.001). Meanwhile, NSCLC survivors had significantly higher total, bioavailable and free 25(OH)D levels than non-survivors (all p < 0.001). However, no significant differences in VDBP levels were observed between survivors and non-survivors. All blood samples of the healthy controls were collected in Autumn. In order to minimize the influence of seasonal differences, the levels of 25(OH)D fractions and VDBP for the healthy controls were further compared with the patients enrolled in autumn, and the results were similar to the original findings involving all patients ( Supplementary Fig. 1). We then examined the relationships among vitamin D fractions. As shown in Table 2, total 25(OH)D had a moderate correlation with bioavailable 25(OH)D (r s = 0.60, p < 0.001) and free 25(OH)D (r s = 0.58, p < 0.001), while a weak correlation with VDBP (r s = 0.24, p < 0.001). Notably, there was a strong correlation between bioavailable and free 25(OH) D (r s = 0.96, p < 0.001). However, we did not find any association between VDBP and bioavailable or free 25(OH)D.

Impacts of 25(OH)D fractions and VDBP on survival
During a median 32 months (range: 1 to 60 months) of follow-up period, 179 (45.3%) patient deaths occurred. The Kaplan-Meier analysis showed that NSCLC patients with higher total 25(OH)D levels had better OS (p < 0.001, log-rank test) and PFS (p < 0.001, log-rank test) ( Fig. 2A,  B). Similarly, higher bioavailable and free 25(OH)D levels were associated with significantly increased OS (p < 0.001 for bioavailable and p = 0.002 for free 25(OH)D, log-rank  test) and PFS (p < 0.001 for bioavailable and p = 0.001 for free 25(OH)D, log-rank test) (Fig. 2C-F). However, there were no significant associations between VDBP and survival outcomes (p = 0.610 for OS and p = 0.470 for PFS, log-rank test; Fig. 2G-H). Then, we further explored the association between 25(OH)D fractions or VDBP and survival through Cox regression analyses. Univariate analyses suggested that gender, smoking, drinking, COPD, BMI, season of blooddrawing, TNM stage, histology, surgery, CEA, NSE, albumin, total cholesterol, NLR and LMR were significant prognostic factors for OS (all p < 0.2; Supplementary Table 2). And in addition to the factors mentioned above, radiotherapy and targeted therapy were also significant prognostic factors for PFS (all p < 0.2; Supplementary Table 2). Furthermore, age was considered as an important factor due to its clinical relevance.
All three 25(OH)D factions were identified as independent prognostic indicators for survival after adjustment for factors mentioned above. The multivariable-adjusted HRs for the highest versus the lowest tertile of total 25(OH)D levels were 0.58 (95% CI, 0.40-0.87; p for trend = 0.008) for OS and 0.61 (95% CI, 0.43-0.86; p for trend = 0.006) for PFS. Meanwhile, the multivariable-adjusted HRs for the highest versus the lowest tertile of bioavailable 25(OH)D levels were 0.45 (95% CI, 0.30-0.67; p for trend < 0.001) for OS and 0.56 (95% CI, 0.40-0.80; p for trend = 0.001) for PFS. And the multivariable-adjusted HRs for the highest versus the lowest tertile of free 25(OH)D levels were 0.49 (95% CI, 0.33-0.73; p for trend < 0.001) for OS and 0.60 (95% CI, 0.42-0.85; p for trend = 0.004) for PFS However, VDBP were not significantly associated with either OS or PFS (both p for trend > 0.05; Table 3). Lifestyle alteration (e.g., reduced outdoor activities) after diagnosis may lead to a decrease in vitamin D levels (especially so for the patients with very poor health status), thus affecting the outcomes. Therefore, in order to reduce the impact of lifestyle alteration, an additional sensitivity analysis was performed through excluding patients who died in the first 6 months of follow-up. It showed that high levels of three 25(OH)D factions were still independently associated with improved survival (Supplementary Table 3).
We found that the inverse association of total, bioavailable, and free 25(OH)D with OS or PFS rate remained in most of the predefined subgroups, despite reduced statistical power in these analyses. And there were no statistical interactions between the three 25(OH)D fractions and any of the above clinical characteristics (All p for interaction > 0.05; Supplementary Tables 4-6).

Prognostic performance of 25(OH)D fractions and VDBP
As shown in Fig. 3, we compared the prognostic performance of the total, bioavailable, free 25(OH)D and VDBP. The time-dependent ROC curves showed that bioavailable and total 25(OH)D obtained similar AUCs in dynamic trends within the follow-up time. In addition, the time-dependent

Discussion
In this study, we evaluated the potential importance of total, bioavailable and free 25(OH)D as well as VDBP as pre-treatment prognostic indicators for OS and PFS among patients with NSCLC. We found that all three 25(OH)D fractions had discrimination for long-term survival in NSCLC patients, as high levels of total, bioavailable and free 25(OH) D each were associated with better OS and PFS. But we did not find a superior prognostic performance in bioavailable or free 25(OH)D as compared to total 25(OH)D. Moreover, no significant association was found between VDBP and survival in NSCLC. Vitamin D deficiency is common in cancer patients, including patients with lung cancer. Herein, we found that 50.1% of the patients had vitamin D deficiency [25(OH)D level < 20 ng/mL] and 37.0% had insufficiency (20-30 ng/ mL), suggesting a severe vitamin D deficiency status in the current NSCLC cohort, which is consistent with previous studies [11,21]. In addition, compared with healthy controls, total 25(OH)D was lower in NSCLC patients, among whom survivors had a higher level of total 25(OH)D than non-survivors, indicating a latent impact of vitamin D on the disease prognosis of NSCLC.
Various mechanisms may underlie the influence of vitamin D on the progression and prognosis of lung cancer. Vitamin D can inhibit lung cancer tumor growth, migration and proliferation, and promote apoptosis by downregulating Histidine-rich calcium-binding protein (HRC) [22]. As an active metabolite of vitamin D, 1,25-dihydroxyvitamin D3 [1,25(OH)2D3] induces E-cadherin and decreases EMTrelated molecules SNAIL, ZEB1, and vimentin in NSCLC cells, thus preventing NSCLC progression [23]. Besides, vitamin D deficiency may increase inflammation and proliferation of pre-malignant cells thereby enhancing progression of squamous lesions [24]. By contrast, vitamin D treatment in the EGFR mutant NSCLC cells may lead to an inhibition of clonogenic growth in a dose-dependent manner [25]. And vitamin D or vitamin D derivatives play a powerful synergic role when used in combination with chemotherapy drugs, promoting the cytotoxic effects of drugs and reducing drug resistance on NSCLC cells compared to using the free single drug [26,27].
Given the close affinity of vitamin D and NSCLC, the prognostic impact of vitamin D levels on NSCLC has been investigated previously, but the results are controversial. Several cohort studies addressed that higher pre-treatment circulating 25(OH)D showed an association with improved survival of patients with early-or advanced-stage NSCLC [9][10][11]. In contrast, two other studies demonstrated that 25(OH)D levels at the time of initial diagnosis or before treatment had no effect on survival in advanced NSCLC [12,13]. In addition, a nested case-control study suggested that higher 25(OH)D was associated with significantly better survival among patients with lung adenocarcinoma but not among those with squamous cell carcinoma [28].The inconsistent findings indicate that the role of vitamin D in NSCLC prognosis still deserves further exploration.
Recent studies have highlighted bioavailable or free 25(OH)D as a more meaningful marker of vitamin D function [16,17]. In the current study, besides total 25(OH)D, we measured bioavailable and free 25(OH)D levels and investigated their relationships with NSCLC prognosis for the first time. We found that plasma concentrations of bioavailable and free 25(OH)D both were significantly decreased as compared to healthy controls. Additionally, total 25(OH)D was positively associated with both bioavailable and free 25(OH) D, which may partly explain the decrement of bioavailable and free 25(OH)D. We also found that total, bioavailable, and free 25(OH)D levels were inversely related to both OS and PFS, indicating all three 25(OH)D forms as independent predictors for survival among NSCLC patients. Besides, the AUCs for bioavailable and free 25(OH)D were similar to that for total 25(OH)D in dynamic trend, thus we could not distinguish if bioavailable and free 25(OH)D were superior to total 25(OH)D in predicting NSCLC prognosis.
Our results did not completely correspond with finding for other cancers in previous studies. Chen et al. reported that plasma bioavailable 25(OH)D was independently associated with PFS and OS, while total 25(OH)D was independently associated with PFS but not OS in diffuse large B-cell lymphoma [29]. Yang et al. showed that higher bioavailable and free 25(OH)D, rather than total 25(OH)D, were associated with better OS in I-III stage colorectal cancer, though only free 25(OH)D was an independent prognostic factor for OS [30]. Similarly, Fang et al. suggested that bioavailable 25(OH)D was a preferable biomarker over total 25(OH)D for the prognosis of hepatocellular carcinoma [31]. By contrast, Yuan et al. demonstrated that bioavailable 25(OH)D levels were not more strongly associated with risk of advanced and lethal prostate cancer than total 25(OH)D levels [32]. There are several potential reasons for the inconsistent findings for different cancers. First, the measurements for 25(OH)D fractions varied across these studies, and there is no uniform standard measurement or cutoff value for bioavailable and free 25(OH)D at present. Second, circulating vitamin D concentration may be influenced by various factors such as liver function, kidney diseases, estrogens and genetic background [15], thus 25(OH)D fractions measured at a single time point in these studies may not comprehensively represent the overall vitamin D status of the enrolled patients. And finally, total 25(OH)D levels may indeed differ in distinct cancers. The mean level of total 25(OH)D in Fang et al. study was much higher than the levels in Chen et al. and in our present study (35.8 ng/mL vs. 16.0-20.8 ng/mL) [29,31]. It is possible that total 25(OH)D with a small increment provides more benefit for patients with vitamin D deficiency (< 20 ng/mL), but no meaningful impact on patients with sufficient vitamin D (> 30 ng/mL). Taken together, high level of vitamin D may have a positive association with improved prognosis in various cancers. However, it still should be more prudent about the conclusion that bioavailable or free 25(OH) D could be a better biomarker for the estimation of vitamin D status or the prediction of cancer prognosis.
In addition to different forms of vitamin D, another important biomarker for vitamin D status is VDBP. We found that VDBP levels were significantly higher in NSCLC patients compared with healthy controls. As VDBP is bound to most of vitamin D, the increase of its levels may reduce the free or bioavailable fractions of vitamin D. In other words, the bioavailability of vitamin D may be limited even though circulating vitamin D is sufficient. Previous studies showed an inverse association between VDBP and free or bioavailable 25(OH)D [32,33], though the association was not confirmed in the current study. Furthermore, VDBP per se has multiple biological activities beyond vitamin D transporter, including extracellular actin scavenger and enhancement of neutrophil chemotactic activity [18]. Particularly, the VDBP-derived macrophage activating factor (MAF) demonstrated potent inhibition of both proliferation and migration of tumor cells [34], and induced tumor cells apoptosis through stimulating macrophages [35]. To date, two studies have examined the relationship between VDBP level and lung cancer prognosis. One UK study including 148 NSCLC patients showed that patients with the highest quartile of VDBP (> 430.2 μg/mL) tended to have better prognosis in comparison to those with the lowest quartile of VDBP (< 199.4 μg/mL) [36]. The other study conducted in 500 Finnish male lung cancer cases demonstrated that circulating concentration of VDBP was not associated with lung cancer specific survival [28]. In the current study, no significant association was found between VDBP level and NSCLC survival. Of note, VDBP in the European studies are much higher than that in our study (291-337 μg/mL vs. 204.9 μg/mL), which might be partially attributed to the differences in race and genetic polymorphisms [16]. Whether a sufficiently high level of VDBP (such as > 430.2 μg/mL) may have a positive effect on the progression of lung cancer remains to be further investigated.
Several limitations to our study should be acknowledged. First, the sample size was relatively small, thus the findings of the present study may need to be generalized with caution, and further studies with larger sample sizes are needed to precisely validate these results. Second, the bioavailable 25(OH)D level was calculated based on the total 25(OH) D level, albumin level, and VDBP level in the plasma. The direct measurement of the bioavailable 25(OH)D level was established recently [37] and could be used in future studies. Thirdly, the concentrations of 25(OH)D and VDBP were measured only once but not dynamically, and it remains unknown whether temporal changes in 25(OH)D and VDBP have significant effect on the prognosis of NSCLC patients. Finally, other factors affecting vitamin D levels, such as physical activity, sunlight exposure and vitamin D supplementation, were not evaluated in our study. But season of blood-drawing was included as a potential confounder factor, which may reflect sunlight exposure partially. Nevertheless, this study also has several strengths, including the prospective design to minimize the potential for reverse causality. Furthermore, to the best of our knowledge, this is the first study to explore of the relationship between bioavailable and free 25(OH)D and NSCLC survival.

Conclusion
In summary, total, bioavailable, and free 25(OH)D may serve as reliable biomarkers in predicting the OS and PFS of NSCLC patients. And further studies are needed to validate our findings in larger and different patient populations.
Funding This work was supported by the project of State key laboratory of radiation medicine and protection (GZK1202007); the second affiliate hospital of Soochow university science and technology innovation team project funding (XKTJ-TD202009); and the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).
Data availability All datasets generated for this manuscript can be obtained from the corresponding author on reasonable request.

Conflict of interest
The authors declare no conflict of interest.
Ethical approval This study was approved by the Research Ethics Committee of Soochow University (Approval No. ESCU-2015-0002). All procedures performed in studies involving human participants were in accordance with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed consent Written informed consent was obtained from all participants.