Serum Concentrations of Cytokines and Lung Cancer Survival in African Americans and Caucasians

Accumulating evidence suggests a role for inflammation in the development and progression of cancer. Our group recently identified a cytokine gene signature in lung tissue associated with lung cancer prognosis. Therefore, we hypothesized that concentrations of circulating cytokines in serum may be associated with lung cancer survival. Ten serum cytokines, namely, interleukin (IL)-1β, IL-4, IL-5, IL-6, IL-8, IL-10, IL-12, granulocyte macrophage colony-stimulating factor, interferon (IFN)-γ, and tumor necrosis factor-α, were assessed in 353 non–small cell lung cancer cases from a case-control study of lung cancer in the greater Baltimore, Maryland area. Cytokines were measured using an ultrasensitive electrochemiluminescence immunoassay. IL-6 serum concentrations (≥4.0 pg/mL) were associated with significantly poorer survival in both African Americans [hazard ratio (HR), 2.71; 95% confidence interval (CI), 1.26-5.80] and Caucasians (HR, 1.71; 95% CI, 1.22-2.40). IL-10 (HR, 2.62; 95% CI, 1.33-5.15) and IL-12 (HR, 1.98; 95% CI, 1.14-3.44) were associated with lung cancer survival only in African Americans. Some evidence for an association of tumor necrosis factor-α levels with survival in Caucasians was observed, although these results were not significant. These hypothesis-generating findings indicate that selected serum cytokine concentrations are associated with lung cancer survival, and indicate that further research is warranted to better understand the mechanistic underpinnings of these associations. (Cancer Epidemiol Biomarkers Prev 2009;18(1):215–22)

Lung cancer is a major cause of disease burden and loss of life in the United States and worldwide. More people die of lung cancer in the United States than from colon, breast, and prostate cancers combined. The American Cancer Society estimates that there will be 215,020 new cases of lung cancer and 161,840 deaths attributed to lung cancer in 2008. The majority of lung cancers are diagnosed at later stages, which results in high mortality; only 16% of lung cancers are diagnosed when the disease is still localized (1).

There is accumulating evidence that chronic inflammation is involved in the development and progression of cancer (2-4). Chronic inflammation can promote an environment that is conducive to carcinogenesis. In an inflammatory state, there is a high rate of cell turnover and the microenvironment is often highly oxidative and nitrosative, increasing the opportunities for DNA damage and mutation. The inflammatory response is regulated by cytokines, a class of signaling molecules that can have autocrine, paracrine, and endocrine effects. Although many cytokines can produce a variety of biological responses in target tissues, cytokines are often classified based on whether they initiate or maintain inflammation (proinflammatory, secreted by Th-1 T helper cells) or inhibit inflammation (anti-inflammatory, secreted by Th-2 T helper cells; ref. 2). Human lung cancer cells can also secrete inflammatory cytokines (5).

Whereas previous studies assessed the relationship of serum cytokine concentrations with lung cancer (6-10), few conducted multivariate survival analyses to rigorously examine the possible role of cytokines in lung cancer progression as measured by survival (11-18). None of these previous studies took into account the profoundly important role of cigarette smoking when examining the association of inflammatory markers with lung cancer survival. Cigarette smoking is the predominant risk factor for lung cancer and upon inhalation, smoke particulates and chemical irritants induce an immune response that can change cytokine concentrations (19, 20). Additionally, most of the previous studies investigated only a few cytokines in parallel in the same individuals. Another important gap in the evidence is that none of the previous studies reported on the association between circulating cytokine concentrations and lung cancer survival among African Americans, even though this ethnic group has the highest lung cancer incidence and mortality rates in the United States (21, 22).

Our group recently showed that elevated expression levels of interleukin (IL)-6, IL-8, and IL-10 in noncancerous lung tissue were associated with lymph node metastasis, whereas IL-8 and tumor necrosis factor-α (TNF-α) in tumor tissue were related to lung cancer prognosis in a population of patients with stage I adenocarcinoma (23). We sought to further examine the role that circulating cytokines may play in the latter, progression stage of lung carcinogenesis by examining the association of several proinflammatory and anti-inflammatory cytokines with lung cancer survival. In this prospective cohort study, we investigated the association of six proinflammatory (IL-1β, IL-8, IL-12, granulocyte macrophage colony-stimulating factor, IFN-γ, TNF-α) and four anti-inflammatory (IL-4, IL-5, IL-6, and IL-10) cytokines measured in serum with lung cancer survival, adjusting for possible confounders, among the cases from a lung cancer case-control study that included both African Americans and Caucasians.

We carried out a prospective cohort study, comprising follow-up of lung cancer cases enrolled in an ongoing lung cancer-case control study in the greater Baltimore, Maryland area from 1998 to 2003. During this time period 475 lung cancer cases were enrolled and 448 provided serum samples. The study population accrual and eligibility criteria for the case-control study were described previously (24, 25). Briefly, all participants were self-reported African Americans or Caucasians residing in Metropolitan Baltimore or the Maryland Eastern Shore. All cases had histologically confirmed non–small cell lung cancer and were enrolled within 24 mo of diagnosis. Potential participants were excluded if they had known diagnosis of HIV, hepatitis C virus, or hepatitis B virus.

Institutional Review Board approval was obtained from all participating institutions and the NIH. The study protocol included informed consent for the collection of personal and medical information, and tissue specimens from each subject. Laboratory personnel were blinded to each participant's case-control status.

Concentrations of IL-1β, IL-4, IL-5, IL-6, IL-8, IL-10, IL-12p70, granulocyte macrophage colony-stimulating factor, IFNγ, and TNF-α were measured on 25 μL of serum using an ultrasensitive electrochemiluminescence immunoassay following instructions from the manufacturer (Meso Scale Discovery). Ultrasensitive 10-plex plates were custom-designed for the Meso Scale Discovery 6000 instrument. Controls for standard curves were included with each plate. Approximately 17% of the samples were duplicated and evenly distributed interplate and intraplate (29 intraplate and 33 interplate pairs). Reproducibility was evaluated using the Spearman correlation coefficient between duplicate samples (26). Samples with cytokine values less than the limit of detection were assigned a value of one half the limit of detection.

Date and cause of death were obtained from the National Death Index (NDI)⁴

using the NDI Plus search, which provides cause of death codes. The NDI Retrieval Program was used to search the NDI file to determine whether a particular NDI death record qualified as a possible record match. To qualify as a possible match, records had to satisfy an algorithm based on at least one of seven matching criteria, including the Social Security number, exact month of birth, and first and last name, according to instructions from the NDI. A person scored with a “no-match” was presumed alive.

The study end point was lung cancer death. Any mention of lung cancer as the cause of death or death due to another cancer within 2 y of the lung cancer diagnosis was treated as a lung cancer death. Follow-up for this study was through to December 31, 2005.

Serum cytokine concentrations may be influenced by a variety of factors; therefore, multiple factors known to be associated with lung cancer and survival were examined as potential confounders. At enrollment during a structured, in-person interview, participants answered multiple questions including an open-ended question on current medication use and targeted questions regarding previous physician diagnosis of select comorbidities (e.g., chronic bronchitis and emphysema) and lifetime tobacco use. A never smoker was defined as a person who had never smoked >100 cigarettes in his/her lifetime and a former smoker as a person who had quit smoking >1 y prior to the interview. Tumor stage was determined from available pathology or clinical reports and was categorized as stage I (noninvasive, nonmetastatic tumors) compared with >stage I (invasive tumors with positive lymph node and/or distant metastasis).

The influence of serum cytokine concentrations on lung cancer survival was determined by comparing the survival of those at or above the median to those below the median for each cytokine because the distributions of the continuous variables were skewed to the right (data not shown). Log-rank tests were first conducted to determine the extent to which the Kaplan-Meier survival curves varied by each binary cytokine variable. To estimate the adjusted hazard ratios (HR) of lung cancer death for each cytokine, we fit stage-adjusted (II-IV versus I) Cox regression models with delayed entry (or left truncation) to account for the fact that not all of the cases were enrolled at the time of diagnosis. Cytokines that were significantly associated with survival after adjustment for tumor stage were further investigated by adjusting for the following potential confounders: age, sex, current smoking status, pack-years, comorbidities, recent medications, tumor histology, and cancer treatments. The proportional hazards assumption was checked for each cytokine by (a) assessing the Kaplan-Meier curves for consistency during the follow-up period (e.g., see Fig. 1) and (b) including cytokine-time interaction terms in the model using a continuous time variable. If the β coefficient for the cytokine-time term was significant and/or if the survival curves were not consistent over time, a model that included a spline function was included to estimate the HR before and after a given time point. The time point for insertion of the spline function was where the survival curves' divergence/convergence occurred and where the β coefficient of the cytokine-time term was no longer significant. All analyses were done using SAS software, version 8 (SAS Institute Inc.). All reported P values were two-sided and the significance level was specified as P < 0.05.

Table 1 provides a comparison of all cases enrolled from 1998 to 2003, those who had cytokine measurements, and those who were included in the survival analyses. Cytokines were measured in 74% (n = 353) of cases enrolled during this period. The cases with cytokine measurements were representative of the cases enrolled during this period with respect to all of the possible confounding variables assessed. Nineteen subjects were excluded from the subsequent analyses because they had missing information; 18 had missing stage information and 1 had missing pack-year information. Staging data were incomplete due to unavailable records (surgery at other hospitals, or patient did not go to surgery). The analytic cohort (n = 334) remained representative of the cases enrolled during the same calendar years.

A summary of the Spearman correlation coefficients for duplicate assays, the average limit of detection for each cytokine, the number of samples below the limit of detection, and the coefficients of variation for all of the lung cancer cases is provided in Table 2. The variation in measured cytokine concentrations was higher at lower concentrations as indicated by the higher coefficients of variation in the low categories. Spearman correlations between duplicate samples both intraplate and interplate were significant for all cytokines other than granulocyte macrophage colony-stimulating factor. Correlation coefficients were >0.5 for all cytokines intraplate, and for IL-10, IL-8, IL-6, IL-5, and IL-4 interplate (Table 2).

The time from diagnosis to end point (death or censoring) ranged from 1.2 months to 89.9 months, with a median of 30.7 months. During the study period, 211 (60%) individuals died of lung cancer. Five individuals had another cancer listed as the cause of death on their death certificate, but because the participants died within two years of lung cancer diagnosis they were included in the analysis as lung cancer deaths. Of the 160 individuals who were censored, 16 (3 African Americans and 13 Caucasians) died of other causes.

Based on the log-rank tests for differences in the Kaplan-Meier survival curves, only high concentrations of IL-6 in both race groups, and IL-12 in African Americans, were significantly associated with worse survival (Table 3 and Fig. 1). After adjusting for stage (II-IV versus I), higher concentrations of IL-6 in African Americans and Caucasians, IL-10 and IL-12 in African Americans, and TNF-α in Caucasians were associated with worse survival. The associations for IL-6 in Caucasians and IL-12 in African Americans were consistent over the follow-up period (i.e., no proportional hazards assumption violation), whereas the associations for IL-6 and IL-10 in African Americans and TNFα in Caucasians diminished over time (i.e., cytokine-time β coefficient was negative and P < 0.05).

Additional Cox regression models were then used to evaluate the association of serum concentrations of IL-6, IL-10, IL-12, and TNFα with survival after further adjustment for the other possible confounders (Table 4). Spline functions that dichotomized follow-up time at ≤18 and >18 months were included in the adjusted models when assessing IL-6 and IL-10 among African Americans and TNF-α among Caucasians because of the proportional hazards assumption violations. Among African Americans, higher concentrations of IL-6 and IL-10 were each associated with at least a 2-fold increased risk of lung cancer death in the first 18 months. These associations were robust to adjustment for tumor stage, treatment, smoking status, comorbidities, and recent medications. After 18 months of follow-up, higher concentrations of IL-10 were associated with better survival; caution should be taken, however, when interpreting these results due to limited sample size. Additionally, concentrations of IL-12 above the median were associated with worse prognosis after adjusting for stage (HR, 1.98; 95% confidence interval, 1.14-3.44). This association was slightly attenuated only after adjusting for recent medication use (HR, 1.71; 95% confidence interval, 0.97-3.00). Among Caucasians, having high concentrations of IL-6 (HR, 1.71; 95% confidence interval, 1.22-2.40) and TNF-α (HR, 1.51; 95% confidence interval, 0.97-2.35) in the first 18 months of follow-up was associated with worse prognosis after adjusting for stage; these findings were only slightly affected by further adjustments. Results were similar when adjusting for stage categorized as I, II, III, and IV (data not shown).

Inflammation has been shown to have a role in cancer initiation and promotion (2-4, 27), and may thus be an important contributor to carcinogenesis across many anatomic sites. The carcinogenic influence of inflammation may also extend to cancer progression. Thus, this study was done to investigate the potential role of circulating cytokine concentrations in lung cancer survival. In this study, we observed that IL-6 was associated with worse lung cancer prognosis in both African Americans and Caucasians. IL-10 and IL-12 were associated with lung cancer survival in African Americans, whereas TNF-α was associated with survival in Caucasians.

The role of inflammation and immunity in tumor biology is complex. When the immune response is functioning normally, inflammation is self-limiting. The production of proinflammatory or Th-1 cytokines is followed by anti-inflammatory or Th-2 cytokines (2, 27). In the case of chronic inflammatory diseases, the balance between Th-1 and Th-2 cytokines is disrupted, and increased inflammation results in greater production of oxygen and/or nitrogen radicals. These radicals may damage both epithelial and stromal cells (4). Meanwhile, recruitment of inflammatory cells may function to inhibit tumor growth (2, 28). Furthermore, Th-2–dominant cytokine profiles have been correlated with enhanced tumor promotion and progression (27), and tumor cells that produce immunosuppressive (Th-2) cytokines may escape host tumor response (29).

Consistent with previously reported correlations of Th-2 cytokines with tumor promotion, our study observed an inverse association between lung cancer survival and IL-6 in Caucasians over the entire follow-up period and between lung cancer survival and IL-6 and IL-10 in African Americans within the first 18 months post-diagnosis. Both IL-6 and IL-10 are considered Th-2 cytokines (30), although IL-6 was reported to produce both Th-1 and Th-2 responses (27). In addition, our laboratory previously noted that elevated IL-6 and IL-10 mRNA in normal lung tissue were associated with lymph node metastasis (23).

Our observation that serum concentrations of IL-6 were associated with lung cancer prognosis in both African Americans and Caucasians was consistent with previous reports (7, 11, 12, 15). One study noted that the association of IL-6 with lung cancer survival was an independent prognostic factor but only within the first 3 years of follow-up (12). Our median follow-up was only 31 months (20 for African Americans, 34 for Caucasians); therefore, the association of IL-6 with survival should be evaluated again after a longer follow-up period to determine if and when the association between IL-6 and lung cancer survival diminishes. Furthermore, additional studies are needed to investigate our finding that the association between IL-6 concentrations and survival diminishes more rapidly among African Americans than Caucasians. We also observed that IL-6 concentrations were higher in later-stage lung tumors (data not shown), in agreement with other studies (6, 8, 11, 12, 15, 31).

Several studies suggest possible biological mechanisms for increased IL-6 in serum from cancer patients. The tumor cells themselves might have been a source of IL-6; a recent study examined the expression of cytokines from 31 lung cancer cell lines and reported that 55% of the lines expressed IL-6 (5). In addition, results from several studies indicated that IL-6 may function in angiogenesis (5, 32, 33). Furthermore, cyclooxygenase-2 expression in lung tumors was shown to induce IL-6 (32) and was associated with worse lung cancer prognosis (34).

In this study, elevated serum IL-10 was associated with worse survival early in follow-up and better survival later in follow-up among African Americans. The latter finding should be interpreted with caution because of the limited sample size; only 42 (25 low and 17 high IL-10) African Americans survived beyond 18 months to be included in this calculation. In previous studies, elevated serum IL-10 levels were predictive of poor survival (14, 16) even after adjusting for additional covariates, including serum IL-8 and IL-2 levels (14). However, another report did not observe this association (35). IL-10 levels have been shown to correlate with both tumor size and stage (11, 16); therefore, racial differences in tumor characteristics might partially explain why no association was observed with IL-10 among Caucasians. The African American cases in this study tended to have more advanced disease compared with Caucasians. Further, based upon a comparison of the controls from the parent study, which were not included in the present report, Caucasians had higher serum IL-10 levels than African Americans (data not shown). IL-10 is proposed to have several functionalities, which could result in differing effects on lung tumors. Secretion of IL-10 by cancer and immune cells may result in the suppression of cell-mediated immunity, allowing tumors to proliferate and escape immune surveillance, but this immunosuppression may also inhibit angiogenesis, thereby inhibiting tumor growth (36).

Associations of Th-1 cytokines with lung cancer survival were also observed. Elevated IL-12 in African Americans and TNF-α in Caucasians were associated with worse survival. Another report observed an association of higher IL-12 levels with improved survival in an Italian population (17) in contrast to our observations. IL-12 was shown to stimulate antitumor responses in tumor models (37), be associated with antiangiogenic activity (38), and induce tumor-associated macrophage proinflammatory profiles as antitumor responses (39). The reason for the association of higher levels of IL-12 with worse survival only in African Americans in our study is unclear.

Consistent with our previous TNF-α mRNA expression results (23), TNF-α serum concentration was associated with worsened lung cancer prognosis. However, other prospective epidemiologic studies have not found serum concentration to be associated with lung cancer survival (7, 11, 12, 18). The reason for the inconsistency is unclear, but may be due to differences in case population (i.e., distribution of stage or histology). Multiple in vitro and animal studies have shown TNF-α to have antiproliferative qualities (40-44).

Our group recently observed that elevated expression levels of IL-8 noncancerous lung tissue were associated with lymph node metastasis, whereas IL-8 in tumor tissue was related to lung cancer prognosis in a population of patients with stage I adenocarcinoma (23). Serum IL-8 was not found to be a predictor of prognosis in this current study for either race. These results may be attributable to differences in local tissues versus circulating in serum. Consistent with the results of this report, Orditura et al. (14) found that elevated serum IL-8 levels were not associated with poorer lung cancer prognosis after adjusting for other covariates, including IL-10 and IL-2.

Inconsistent results by race (e.g., IL-10, IL-12, and TNF-α) may have been due to differences in genetics, tumor characteristics, or other exposures. Racial variation was observed among controls for many of the cytokines, including IL-10 and TNF-α (data not shown). Cytokine concentrations can be modulated by polymorphisms (45-48) and many of the allele frequencies for these polymorphisms vary by race (45, 46, 49-52). Additionally, differences could have been due to variation in tumor characteristics (e.g., stage or histology). In this study Caucasians had less advanced tumors (χ² P = 0.01) and were more likely to have adenocarcinomas (χ² P = 0.02) than African Americans. Variation in concentration and prognostic ability might also have been influenced by racial differences in other exposures, including infection or cigarette smoking. Given the higher incidence and mortality of lung cancer among African Americans (21), these observations may provide useful avenues for future study.

A limitation of this study is the variability of the cytokine assays and skewed distribution of cytokine concentrations. As noted, variation was higher in the low concentration ranges of cytokine assays. This variability unlikely altered the results shown. Spearman correlations between duplicate samples were consistent with good reproducibility for IL-6, IL-10, TNF-α, and IL-12. Further, all analyses were done using categorization by median cytokine levels. Assuming possible misclassification is nondifferential, results are likely biased towards the null.

Other limitations need to be addressed. Cytokines were measured only once in this study and may be influenced by illnesses (other than lung cancer) or anti-inflammatory medications. Thus, it is not possible to determine how cytokines change over time and if there is an optimal window of time to evaluate cytokines to maximize their prognostic ability. Our findings remained consistent after adjustment for other comorbidities and medication use as well as other possible confounders; however, prospective studies with multiple serial measures are needed. In addition, due to the limited availability of information on tumor histology (surgical sections were unavailable from cases who did not go to surgery and specification of only non–small cell lung cancer on pathology reports), we were unable to examine specific histologic subtypes or tumor size rigorously. Finally, treatment information was not complete for this population and is likely a modifier of cytokine concentrations and survival outcomes. In 1999, however, Martin et al. (12) investigated whether or not the inclusion of patients with surgically removed early-stage cancers influenced their findings. Although the median survival times were attenuated, after accounting for treatment, patients with elevated IL-6 still had a significantly worse prognosis. Therefore, it is reasonable to believe that better assessment of treatment would not alter the findings of the current study.

A major strength of the present study is that it was large enough to make meaningful comparisons in both African Americans and Caucasians. None of the previous studies were conducted in African Americans. Thus, we not only provided the initial investigation of these cytokines and lung cancer prognosis specifically among African Americans, but also provided a comparison between the two races.

The results of the present study suggest that high serum concentrations of IL-6, IL-10, and IL-12 in African Americans and IL-6 and TNF-α in Caucasians are associated with lung cancer prognosis. The associations of IL-10, IL-12, and TNF-α were only observed in one population; therefore, these results should be considered preliminary. The similar results in both African Americans and Caucasians for IL-6 suggest that this cytokine is possibly equally important in both groups. Moreover, the observation of IL-6 association with survival in both populations strengthens the evidence for a role of IL-6 in lung cancer prognosis. Our study represents a promising line of inquiry, but future prospective studies are needed to clarify whether circulating cytokine concentrations contribute to lung carcinogenesis and/or can be used to assist in the treatment of lung cancer.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Karen MacPherson for bibliographic assistance; Raymond Jones, John Cottrell, Donna Perlmutter, and Mark J. Krasna (currently at St. Joseph Hospital in Towson) at the University of Maryland and the Surgery and Pathology Departments at University of Maryland Medical System, Baltimore VA Medical Center, Sinai Hospital, Bon Secours Hospital, Harbor Hospital, and Johns Hopkins Hospital for their contributions to this study; and Leonidas D. Leondaridis from Advance Medical Systems Consultants for his work with the NDI data. Cytokine assays were done by Helen Rager at the Clinical Services Program, under the direction of Dr. William C. Kopp, at SAIC-Frederick, Inc. (http://www.ncifcrf.gov/research/csp/). We also thank Lauren Richey for her work on staging at Johns Hopkins University.