Twenty-four-hour urinary cortisol levels before complete resection of non-small cell lung cancer and survival

Abstract

Cortisol has been speculated to play a role in survival from cancer. The objective of this study was to investigate the association between 24-h urinary cortisol levels and survival from non-small cell lung cancer. From June 1996 to April 1999, a total of 226 patients with resectable non-small cell lung cancer were enrolled. Urine samples were collected over a 24-h period before the curative resections. A total of 14 650 person-months (median = 71 months, range = 1–97 months) were accrued. Over the follow-up period, 56 deaths from all causes were identified through January 2004. Cox proportional hazards regression analyses were used to estimate the relative risk (RR) of death from all causes according to three categories of 24-h urinary cortisol levels. Multivariable RRs of death from all causes for individuals in the highest and intermediate categories, compared with the lowest category, were 1.09 (95% confidence interval = 0.54–2.21) and 1.17 (0.55–2.46), respectively (p for trend = 0.17). The present data do not support the hypothesis that 24-h urinary cortisol levels are associated with survival from non-small cell lung cancer.

Introduction

Lung cancer is the most common form of cancer and the most common cause of cancer-related deaths in the world [1,2]. In Japan, lung cancer is the leading cause of death from cancer among men and women, and the incidence of lung cancer has been increasing in recent years [3]. In 1999, the number of lung cancer deaths reached 39 380 deaths (22.4% of all cancer-related deaths) in men and 14 598 deaths (12.7% of all cancer-related deaths) in women [3]. Various individual characteristics, such as sex, pathologic stage, performance status, and comorbidity, have been shown to play a role in survival from lung cancer [4,5]. However, further clarification of the factors contributing to survival from lung cancer is needed.

The association between cortisol levels and survival from cancer has been hypothesized. The hypothalamic-pituitary-adrenal (HPA) axis is known to play a key role in the adaptation of the body to both physical and psychological stress [6]. Cancer patients repeatedly endure physical and psychological events that activate stress mechanisms, including the HPA axis [7]. Cortisol is the end product of HPA axis activation, and its 24-h accumulation in urine is a reliable measure of physiological or psychological stress [8]. Cortisol has been speculated to play a role in survival from cancer. Alterations in HPA axis functioning have been reported in cancer patients with elevated cortisol levels [9]. In animal models, cortisol may accelerate tumor growth via an immunosuppressive action [10] or by influencing metabolic processes [11]. In four studies, however, a statistically significant association between cortisol levels and an increased risk of death from all causes was not found among patients with renal cell carcinoma [12], lung cancer [13], or breast cancer [14,15]. Each of these previous studies had methodological limitations, including the absence of data on the 24-h urinary cortisol levels and failure to control sufficiently for potentially confounding variables, such as performance status [12–15], comorbidity (e.g. cardiovascular diseases or diabetes mellitus) [12–15], smoking status [12–15], and depression [12,13,15]. Thus, little evidence on the association between urinary cortisol levels and survival from lung cancer was available until now.

The objective of this study was to investigate in further detail the possible association between the 24-h urinary cortisol levels and survival from lung cancer by conducting a prospective cohort study in Japan. This is the first study to assess the 24-h urinary cortisol levels before the resection of lung cancer.

Material and methods

Study cohort

The design of this study has been reported in detail elsewhere [16,17]. Briefly, we invited patients with resectable non-small-cell lung cancer (NSCLC) who were scheduled for treatment in the Thoracic Oncology Division of the National Cancer Center Hospital East, Kashiwa, Japan, to participate in the study. The eligibility criteria were an age of 18 years or older; awareness of the diagnosis of cancer; ability to speak Japanese; predetermined standard surgical procedure (lobectomy or pneumonectomy with mediastinal lymph node dissection); no evidence of brain tumor on computerized tomography or magnetic resonance images of the head; no history of or current use of chemotherapy, immunotherapy, or radiation therapy; no active concomitant cancer; curative resection based on pathology reports indicating an International Union Against Cancer (UICC) disease stage of pT1 to pT3, pN0/1, pM0 [18]; and no other medical conditions. As a result, a total of 303 patients with NSCLC were enrolled in the study between June 1996 and April 1999. The study protocol was approved by the institutional review board of the National Cancer Center, Japan. Each patient was fully informed of the purpose of the study before written consent was obtained.

Exposure data

Urine samples were collected over a 24-h period (from 10:00 am to 10:00 am the following day) before the curative resections and stored at −80°C until further use. Assays were performed in triplicate. Urinary cortisol levels were measured using a commercial radioimmunoassay kit by TFB (Immnotech Ltd., Marseille, France). The reported intra-assay coefficients of variation for three different urinary cortisol levels was 4.2% to 7.6%, and the inter-assay coefficient was 3.9% to 5.4%. The sensitivity of the assay was reported to be 0.36 µg/dl. In our data, intra-assay coefficients of variation for three different urinary cortisol levels was 5.2% (n = 226), and test–retest reliability of correlation coefficients of urinary cortisol levels within a week was 0.91 (n = 25).

Some demographic and clinical characteristics, including age, sex, weight, height, pathologic stage, histology, type of surgery, performance status before curative resection, serum albumin levels, and past medical histories (myocardial infarction, hypertension, or diabetes mellitus), were obtained from the patient's medical charts, whereas educational level, marital status, and smoking status were determined in semi-structured interviews conducted by two psychiatrists in an interview room. History of depression between the diagnosis of lung cancer and surgery was assessed at 1 month after surgery by a trained psychiatrist using Structured Clinical Interviews for DSM-III-R (SCID) [16,19]. History of depression was assessed during the period from cancer diagnosis to surgery. The number of depressive episode items, according to the DSM-III-R results, was obtained by summing the number of the following nine items present: (a) depressive mood, (b) diminished interest or pleasure, (c) change in body weight or appetite, (d) insomnia or hypersomnia, (e) psychomotor agitation or retardation, (f) fatigue or loss of energy, (g) feelings of worthlessness or guilt, (h) loss of concentration or indecisiveness, and (i) suicidal thoughts. A diagnosis of depression was made when two or more items, including either depressive mood or diminished interest or pleasure, were present.

Follow-up

The survival of the subjects was followed up from June 1996 through to January 2004. Of the 303 subjects (189 men and 114 women) who had undergone surgery between June 1996 and April 1999 and were registered for the semi-structured interviews, curative resection was confirmed by the final pathological report of the disease stage in 262 (86.5%) subjects. At the time of the 1-month interview, 3 patients could not be contacted and 20 refused to participate in the study, 10 of them because of the psychological and physical burden, 4 because of the physical burden, and 3 for unknown reasons. Nine subjects with a past history of renal diseases were excluded from the analysis; in addition, four subjects were excluded because the 24-h urinary cortisol levels had not been obtained. As a result, a total of 226 (140 men and 86 women) patients were included in the final analysis. We observed little difference between patients who participated this study (n = 303) and those who remained for the analysis of patients (n = 226). The mean ages were 63.1 and 62.8 years, and the distribution of sex was 62.4% and 62.0%, respectively. Taken together, these findings indicate that the subjects we analyzed were representative of the total patients in the study.

The person-months of follow-up were counted for each subject from the date of study enrollment until the date of death or the end of the study period (January 2004), whichever occurred first. A total of 14 650 person-months (median = 71 months, range = 1–97 months) were accrued. A total of 2 subjects (0.8% of the analytic cohort) were lost to follow-up during the study period. Over the follow-up period, 56 deaths from all causes were identified.

Statistical analyses

The 24-h urinary cortisol levels were divided into three categories in a manner such that the total subjects were grouped as closely as possible into even-sized tertiles. Relative risk (RR) was computed as the number of deaths from all causes among the subjects in each 24-h urinary cortisol category divided by the number of deaths from all causes among the lowest 24-h urinary cortisol category. Cox proportional-hazards regression analyses were used to adjust for sex, age, and potentially confounding variables [20] using the SAS PHREG procedure on the SAS version 8.2 statistical software package (Cary, NC, USA). The assumption of proportional hazards was verified graphically. We repeated analysis after excluding the 11 deaths (9 men and 2 women) diagnosed after baseline (the date of entry) in the first year of follow-up. The p-values for testing the statistical significance of linear trends were calculated by treating 24-h urinary cortisol levels as continuous variables. All p-values were two-tailed.

In addition to sex and age, the following variables were considered as potential confounders a priori: body mass index in kg/m² (<18.4, 18.5 to 24.9,>25.0); education (college/university or higher or other); marital status (living with spouse or other); past histories (myocardial infarction, hypertension, or diabetes mellitus); cigarette smoking (never smoked, smoked in the past, or currently smoking); history of depression (presence or absence); pathologic stage (IA, IB, IIA–IIB, or IIIA); histology (adenocarcinoma, squamous carcinoma, large, or other); type of surgery (lobectomy or pneumonectomy); performance status before curative resection (0 or 1); serum albumin levels in g/100 ml (<3.4 or > 3.5).

Results

The 24-h urinary cortisol levels (µg/day) ranged from 6.1 to 119.0 (mean, 27.6; standard deviation, 19.1). Reported normal 24-h cortisol levels (µg/day) ranged from 11.2 to 80.3. The characteristics of the patients in the three categories (i.e. approximate tertiles) of 24-h urinary cortisol level (Table I) were compared. There was no significant association between the demographic/medical data, except for the gender distribution of the subjects, and the three categories of 24-h urinary cortisol levels. As compared with the sex distribution of the subjects with the lowest or intermediate urinary cortisol levels, the proportion of men in the group of subjects with the highest urinary cortisol levels was higher.

Table I.  Characteristics of the subjects according to twenty-four-hour urinary cortisol levels.

	Twenty-four-hour urinary cortisol levels (µg/day)*			p value
Characteristics	≤17.7	17.8–29.3	≥29.4	ANOVA or χ^2 test
No. of subjects	75	75	76
Demographic data
Age in years, mean±SD	63.0±9.2	64.1±11.3	61.3±11.0	0.26
Sex, man, %	50.7	64.0	71.1	0.033
Body mass index in kg/m^2≥25.0, %	13.3	21.3	18.4	0.60
Education, college/university or higher, %	17.3	14.7	26.3	0.17
Marital status, living with spouse, %	82.7	76.0	84.2	0.39
Past histories, %
Myocardial infarction	2.7	4.0	2.6	0.86
Hypertension	22.7	34.7	26.3	0.24
Diabetes mellitus	4.0	8.0	7.9	0.53
Smoking status, %				0.98
Current smoker	40.0	41.3	43.4
Ex-smoker	24.0	26.7	23.7
Never-smoker	36.0	32.0	32.9
History of depression, %§	10.7	10.7	10.5	0.83
Medical data
Pathologic stage, %†
IA	52.0	48.0	50.0	0.68
IB	28.0	22.7	21.1
IIA	2.7	4.0	5.3
IIB	13.3	21.3	15.8
IIIA	4.0	4.0	7.9
Histology, %				0.69
Adenocarcinoma	64.0	66.7	67.1
Squamous carcinoma	26.7	24.0	17.1
Large	4.0	2.7	6.6
Other	5.3	6.7	9.2
Type of surgery				0.16
Lobectomy	98.7	92.0	94.7
Pneumonectomy	1.3	8.0	5.3
Performance status before curative resection, %‡
0	66.7	69.3	69.7
1	33.3	30.3	30.3
Serum albumin in g/100 ml ≥ 3.5, %	8.0	9.3	5.3	0.63

*The 24-h urinary cortisol levels were divided into three categories in manner such that the total subjects grouped as closely as possible into even-sized tertiles.

†Defined by TNM classification: International Union Againt Cancer.

‡Defined by the Eastern Cooperlative Oncology Group.

§Defined by the Structured Clinical Interview for DSM-III-R (SCID).

The pathologic stage of the patients was strongly associated with the risk of death from all causes: the sex- and age-adjusted RRs, with reference to stage IA (95% confidence interval [CI]) were 1.05 (0.48–2.33), 3.22 (1.70–6.11), and 4.98 (2.02–12.26) in patients with stage IB, IIA–IIB, and IIIA cancers, respectively. This finding remained basically unchanged after multivariate adjustment (Table II).

Table II.  Results of Cox's proportional-hazards regression analysis for pathologic stage and survival from lung cancer*.

	Pathologic stage
	IA	IB	IIA–IIB	IIIA
No. of subjects	113	54	47	12
Median follow-up time (range), months	74 (9–91)	68 (11–97)	66 (1–89)	38 (3–87)
No. of deaths from all causes	17	10	22	7
Sex-, age-adjusted RR1	1.00 (referent)	1.05 (0.48-2.33)	3.22 (1.70–6.11)	4.98 (2.02–12.26)
Multivariable adjusted RR1	1.00 (referent)	0.77 (0.33-1.79)	2.91 (1.34–6.31)	2.89 (0.95–8.78)

*The multivariate relative risk (RR) has been adjusted for age (continuous variables); sex; body mass index in kg/m2 (<18.4, 18.5–24.9, or >25.0); education (college/university or higher or other); marital status (living with spouse or other); past histories (myocardial infarction, hypertension, or diabetes mellitus); cigarette smoking (never smoked, smoked in the past, or currently smoking); history of depression (presence or absence); histology (adenocarcinoma, squamous carcinoma, large, or other); type of surgery (lobectomy or pneumonectomy); performance status before curative resection (0 or 1); serum albumin levels in g/100 ml (<3.4 or >3.5); 24-h urinary cortisol levels in µg/day (<17.7, 17.8–29.3, or >29.4). RR1 denotes the RR with all cases of deaths.

A multivariable Cox proportional-hazards regression analysis showed no significant association between the 24-h urinary cortisol levels and the risk of death from all causes (Table III). This finding remained basically unchanged when we excluded the subjects identified with death from all-cause during the first one year of follow-up. An additional stratified analysis according to pathologic stage (IA or IB–IIIA) was also conducted, but a significant association between the 24-h urinary cortisol levels and survival from lung cancer was not found for either stratification (Table IV). We also conducted a stratified analysis according to sex. A multivariable Cox proportional-hazards regression analysis showed no significant association between the 24-h urinary cortisol levels and the risk of death from all causes in either men or women. In men, the multivariate RR with reference to the lowest category (95% CI) of urinary cortisol levels was 1.31 (0.59–2.93) and that for the intermediate and high categories was 1.30 (0.58–2.92) (p for linear trend = 0.14). In women, the multivariate RRs with reference to the lowest category (95% CI) of urinary cortisol levels was 1.13 (0.17–6.50) and that for the intermediate and high categories was 0.34 (0.12–1.86) (p for linear trend = 0.10). Further stratified analyses according to age at diagnosis, body mass index, education, marital status, past histories (myocardial infarction, hypertension, or diabetes mellitus), history of depression, pathologic stage, histology, type of surgery, performance status before curative resection, and serum albumin levels were also conducted. However, the lack of association between the 24-h urinary cortisol levels and survival from lung cancer was not remarkably modified by these variables (data not shown).

Table III.  Results of Cox's proportional-hazards regression analysis for twenty-four-hour urinary cortisol levels and survival from lung cancer*.

	Twenty-four-hour urinary cortisol levels (µg/day)
	≤17.7	17.8–29.3	≥29.4	p for linear trend
No. of subjects	75	75	76
Median follow-up time (range), months	71 (11–89)	71 (5–90)	72 (1–97)
No. of deaths from all causes	17	22	17
Sex-, age-adjusted RR1	1.00 (referent)	1.17 (0.62–2.22)	0.87 (0.44–1.71)	0.58
Multivariable adjusted RR1	1.00 (referent)	1.09 (0.54–2.21)	1.17 (0.55–2.46)	0.17
Multivariable adjusted RR2	1.00 (referent)	0.76 (0.34–1.73)	1.00 (0.45–2.22)	0.75

*The multivariate relative risk (RR) has been adjusted for age (continuous variables); sex; body mass index in kg/m2 (<18.4, 18.5–24.9, or >25.0); education (college/university or higher or other); marital status (living with spouse or other); past histories (myocardial infarction, hypertension, or diabetes mellitus); cigarette smoking (never smoked, smoked in the past, or currently smoking); history of depression (presence or absence); pathologic stage (IA, IB, IIA-IIB, or IIIA); histology (adenocarcinoma, squamous carcinoma, large, or other); type of surgery (lobectomy or pneumonectomy); performance status before curative resection (0 or 1); serum albumin levels in g/100 ml (<3.4 or >3.5). RR1 denotes the RR with all cases of deaths. RR2 denotes the RR with cases of death who died in the first 1 year of follow-up excluded from the analysis.

Table IV.  Results of Cox's proportional-hazards regression analysis for twenty-four-hour urinary cortisol levels and survival from lung cancer according to pathologic stage*.

	Twenty-four-hour urinary cortisol levels (µg/day)
	≤17.7	17.8–29.3	≥29.4	p for linear trend
	Pathologic stage IA (n = 113)
No. of subjects	39	36	38
Median follow-up time (range), months	73 (29–83)	75 (21–89)	80 (9–91)
No. of deaths from all causes	17	22	17
Multivariable adjusted RR1	1.00 (referent)	1.91 (0.38–9.77)	0.81 (0.16–4.10)	0.41
	Pathologic stage IB to IIIA (n = 113)
No. of subjects	36	39	38
Median follow-up time (range), months	66 (11–89)	65 (5–90)	67 (1–97)
No. of deaths from all causes	17	22	17
Multivariable adjusted RR1	1.00 (referent)	1.20 (0.50–2.88)	1.05 (0.42–2.62)	0.09

*The multivariate relative risk (RR) has been adjusted for age (continuous variables); sex; body mass index in kg/m2 (<18.4, 18.5–24.9, or >25.0); education (college/university or higher or other); marital status (living with spouse or other); past histories (myocardial infarction, hypertension, or diabetes mellitus); cigarette smoking (never smoked, smoked in the past, or currently smoking); history of depression (presence or absence); histology (adenocarcinoma, squamous carcinoma, large, or other); type of surgery (lobectomy or pneumonectomy); performance status before curative resection (0 or 1); serum albumin levels in g/100 ml (<3.4 or >3.5). RR1 denotes the RR with all cases of deaths.

The patients were also grouped into evenly sized quartiles of 24-h urinary cortisol level. The findings, however, remained basically unchanged; the multivariate RRs with reference to the lowest category (<14.4 µg/day) were 2.34 (0.99–5.50), 0.59 (0.22–1.60), and 1.99 (0.81–4.92) for subjects in the second (14.5 to 22.6 µg/day), third (22.7 to 33.6 µg/day), and fourth (>33.7 µg/day) highest categories, respectively (p for linear trend = 0.18).

Discussion

In this prospective cohort study in Japan, no association was found between pre-surgical 24-h urinary cortisol levels and survival from lung cancer. The study suggested that 24-h urinary cortisol levels might not play an important role in survival from lung cancer. Our results were consistent with those of four earlier studies that examined the association between cortisol levels and death from all causes [12–15]. In detail, Walop et al. [12] followed up 119 men and women with lung cancer for about 3 years (longest follow-up period) and documented 90 deaths from all causes. Serum cortisol levels were assessed at 10:00 am. No significant association was observed between the cortisol levels and survival. Rasmuson et al. [13] followed up 211 men and women with renal cell carcinoma for 102 months (median follow-up time) and documented 155 deaths from all causes. Serum cortisol levels were assessed between 08:00 am and 11:00 am. Compared with the lowest cortisol category, a stronger, but not significant, association was found for patients with high cortisol levels (multivariate relative risk, 1.35; 95% confidence interval, 0.92–1.96). Sephton et al. [14] followed up 104 women with metastatic breast cancer for 7 years and documented 71 deaths from all causes. Salivary cortisol levels were assessed four times at 08:00 am, 12:00 noon, 17:00 pm, and 21:00 pm. No significant association was observed between the cortisol levels and survival. Osborne et al. [15] followed up 61 women with early breast cancer for 6.1 to 7.9 years and documented 14 deaths from breast cancer. Plasma cortisol levels were assessed at 07:30 am to 09:30 am. No significant association was observed between the cortisol levels and survival from cancer.

Our study has several methodological advantages, compared with previous studies on cortisol levels and survival from cancer. First, since the urine samples were collected over a 24-h period, the cortisol data were free from diurnal rhythms. This is the first study to assess the association between 24-h cortisol levels and survival from cancer. Second, we sufficiently controlled for potentially confounding variables, such as pathologic stage, performance status, comorbidity (cardiovascular diseases or diabetes mellitus), smoking status, and depression. In particular, various studies have reported that pathologic stage [12] and depression are associated with cortisol levels [22] as well as cancer prognosis [4,23]. In our data, the 24-h cortisol levels were slightly associated with pathologic stage in a positive manner. The mean values (standard deviation) for pathologic stages IA, IB, IIA to IIB, and IIIA were 27.6 µg/day (19.5), 24.7 µg/day (15.4), 28.4 µg/day (18.2), and 37.8 µg/day (29.3), respectively. In addition, the pathologic stage of the lung cancer was strongly associated with the risk of death from all causes (see Table II). On the other hand, the mean values (SD) for patients with or without a history of depression were 29.8 µg/day (23.4) and 27.4 µg/day (18.5), respectively, and a history of depression was not associated with the risk of death from all causes (data not shown). We observed little difference in the point estimates of RRs whether or not multivariable adjustments were made for pathologic stage, depression, and other variables (Table III). The RRs did not exhibit any significant variations when the analyses were stratified according to these covariates. These findings indicate that the observed lack of association between the 24-h urinary cortisol levels and the risk of death from all causes was not substantially confounded or modified by pathologic stage, depression, or other variables in the present study.

There were several limitations to this study. First, there was some sampling bias, because the results were obtained from only one institution, a teaching cancer center hospital in Japan. Moreover, some participants of this study refused to participate for reasons such as psychological and/or physical burden. Therefore, caution must be exercised before generalizing the results of our study in regard to the association between the 24-h urinary cortisol levels and survival from lung cancer. Second, the number of deaths from all causes was modest to small (56 deaths from all causes). Therefore, our study may not have sufficient statistical power to detect slight increases or decreases in the risk of death from all causes associated with the urinary cortisol level. Thus, our follow-up period and the number of deaths from all causes may have been insufficient, and the possible association between the 24-h urinary cortisol level and survival should be further investigated. Third, our subjects consisted of lung cancer patients who had been selected to undergo curative resections and were mostly early-stage patients. Patients with advanced lung cancer are likely to suffer from greater degrees of physical and psychological stress than early-stage lung cancer patients. Thus, the possible association between cortisol levels and survival in patients with advanced lung cancer remains unclear. Lastly, abnormal diurnal cortisol rhythms may compromise tumor resistance and have been implicated in tumor growth, in both animal models and in vitro studies [24,25]. Only one previous study has focused on the association between diurnal cortisol rhythms and survival from cancer, and that study found a statistically significant association between flattened or abnormal diurnal cortisol rhythms and an increased risk of death from all causes among women with metastatic breast cancer [14]. We were unable to examine this issue because our study did not assess diurnal cortisol rhythms.

In conclusion, the present data do not support the hypothesis that 24-h urinary cortisol levels may be associated with survival from non-small cell lung cancer.

This study was supported by the awardee of a Research Resident Fellowship from the Foundation for the Promotion of Cancer Research (Japan) for the 3rd Term Comprehensive Control Research for Cancer. The authors wish to thank Y. Kojima, N. Taguchi, and R. Katayama of the Psycho-Oncology Division, National Cancer Center Research Institute East, Kashiwa, Japan, for their research assistance.