Prognostic implications of depression and inflammation in patients with metastatic lung cancer

Abstract

Background:

Lung cancer-related inflammation is associated with depression. Both elevated inflammation and depression are associated with worse survival. However, outcomes of patients with concomitant depression and elevated inflammation are not known.

Materials & methods:

Patients with metastatic lung cancer (n = 123) were evaluated for depression and inflammation. Kaplan–Meier plots and Cox proportional hazard models provided survival estimations.

Results:

Estimated survival was 515 days for the cohort and 323 days for patients with depression (hazard ratio: 1.12; 95% CI: 1.05–1.179), 356 days for patients with elevated inflammation (hazard ratio: 2.85, 95% CI: 1.856–4.388), and 307 days with both (χ² = 12.546; p < 0.001]).

Conclusion:

Depression and inflammation are independently associated with inferior survival. Survival worsened by inflammation is mediated by depression-a treatable risk factor.

Lay abstract

Depression and inflammation are common in patients with lung cancer. Inflammation is associated with depression, which helps explain why patients with lung cancer often develop depression. Both depression and inflammation lead to worse survival. However, the effect of concomitant depression and inflammation on survival has not been evaluated in this context. This study evaluated survival in 123 patients with metastatic lung cancer. Survival differed based on the singular or coupled presence of depression and inflammation. The survival rate for all patients was 1.4 years from the time of the survey. Patients with significant depression or inflammation survived less than 1 year from the time of the survey. Survival was even worse for patients with both depression and inflammation (approximately 0.84 years). Depression, in addition to inflammation, mediated worse survival in those patients. Depression and inflammation contribute to worse survival independently and collectively. Depression is a treatable risk factor for poor survival in patients with metastatic lung cancer.

Lung cancer is the most commonly diagnosed cancer worldwide and is associated with high rates of depression [1,2]. Although there are certainly many psychological reasons for this added misery (e.g., loss of life, stress, relationship changes, personal identity changes and effects on employment), there are also biological explanations that help explain the strong association and high prevalence of depressive symptoms [3]. Interestingly, many of the putative biological explanations are related to the effects of excessive inflammation, which is abundant in patients with metastatic lung cancer [4]. For example, hypothalamic-pituitary-adrenal axis disruption responsible for sleep dysregulation seen in depression is mediated by pro-inflammatory cytokines that alter cortisol feedback mechanisms [5]. As such, cancer and its treatments that induce inflammation appear to promulgate depressive symptoms at the same time [6]. However, patients who are otherwise healthy (without lung cancer) but depressed also exhibit elevated inflammation [7,8]. Although the directionality is not well established, the presence of inflammation in patients with lung cancer may serve as a risk factor for comorbid depression and have treatment implications. The relationship between inflammation and depression has been well documented in various lung cancer settings (e.g., early vs advanced stage, with and without treatment effects) using cross-sectional and longitudinal designs [9–11]. Our group previously reported this association using the readily available inflammatory marker C-reactive protein (CRP) [12].

The implications of experiencing depression alongside lung cancer are multitudinous (e.g., poor adherence to treatment, quality of life) and all highlight the need to address depressive symptoms early along the lung cancer trajectory [11,13]. Among the most convincing reasons for addressing concomitant depression early along the lung cancer trajectory are its survival implications given that depression is a modifiable prognostic risk factor [14,15]. Sullivan et al. found that longitudinal changes in depressive symptoms in patients with lung cancer were associated with differences in mortality [15]. Survival was 50% poorer in patients with either new-onset or persistent depression, but depression remission was associated with a similar mortality rate as never having had depression [15]. Estimated survival probabilities normalized to never having had depression for those patients who reported remittance of depression over the duration of the study. Similarly, Giese-Davis et al. found that women with metastatic breast cancer and depression had worse survival, but those patients whose depression remitted on follow-up demonstrated significantly improved survival independent of covariates [16]. However, none of these studies evaluated mechanisms of depression and cancer-related survival.

Inflammation has long been evaluated as a prognostic marker in patients with lung and other solid malignancies [17,18]. Inflammatory indices may use single biomarkers such as CRP, IL-6 or combinations of inflammatory markers such as the CRP/albumin ratio or neutrophil to lymphocyte ratio or formulas that include other biometric data such as BMI, which is used in the Advanced Lung Cancer Inflammation Index [19–21]. In fact, many of the same prognostic markers for lung cancer are also markers of depression that have been evaluated in lung cancer and other settings [9]. The acute-phase reactants CRP and albumin are examples [18,22,23]. CRP has been the most extensively studied and has cut-points associated with depression (e.g., 1 mg/dl) [24,25].

The survival implications of concomitant depression and inflammation (measured by peripheral blood) have not been reported in any cancer setting. Inflammation may provide a mechanistic understanding of poor survival-related depression. That is, depression associated with inflammation may contribute cumulatively to adverse cancer-related survival [3]. The interaction between these prognostic factors may help us understand the survival implications of both factors individually and in combination. We hypothesized that depression and inflammation would be associated with worsened survival estimates as individual and additive collective risk factors.

Methods & materials

Study design

The study was approved by the Memorial Sloan Kettering Cancer Center Institutional Review Board in May 2018. Its design was a retrospective case-cohort. Surveys and blood work were collected over 12 months (May 2017–May 2018) in dedicated thoracic medical oncology clinics. Results were previously reported as a cross-sectional analysis of CRP and depression on 109 patients [12]. Survival data were collected November 19, 2019, and revealed 79 death events of 123 patients.

Participants

Study inclusion required a confirmed histologic diagnosis of stage IV lung cancer–non-small-cell lung cancer (NSCLC) including squamous cell carcinoma and nonsquamous cell carcinoma (e.g., adenocarcinoma and others) and small cell lung cancer (SCLC). Patients were receiving active anticancer treatments, were fluent English speakers and demonstrated an acceptable Eastern Cooperative Group performance status (≤2) [26]. Patients were required to have been receiving treatment for at least 1 month after a confirmed diagnosis for study inclusion. Of note, data were collected before the routine use of chemo-immunotherapy combinations so that treatments were divided between chemotherapy, immunotherapy and targeted therapies. Patients with additional recent cancer diagnoses (within 5 years) were excluded.

Procedure

Patients were approached in oncology clinics by a treating staff member (e.g., nurse practitioner and medical oncologist). They filled out questionnaires containing standardized survey questions either before the appointment or during infusion treatments. CRP laboratory values were obtained the same day that the questionnaires were completed. Survey results were reviewed with patients during the same visit as part of routine care. Referrals to available psychological services were discussed if clinically indicated.

Measures

Medical and demographic characteristics

Medical information was gathered from the electronic health record (EHR) and included disease type, treatment type, length of time with diagnosis, BMI and antidepressant medication use. Demographic information was gathered from the EHR and included age, race/ethnicity, marital status, deceased status (yes/no) and date of death.

Depression

Depression severity and criteria were measured by the Hospital Anxiety and Depression Scale (HADS-D). The HADS-D has demonstrated adequate psychometric properties in the lung cancer setting [27,28]. This 14-item measure was developed to identify clinically significant cases of depressive disorders among medically ill patients [27]. Therefore, physical symptoms are excluded from the HADS-D because they may potentially confound the effects of illness on depressive symptoms such as sleep, appetite disturbance and fatigue. HADS-D scores range from 0 to 21 points, and each question's score ranges from 0 to 3. HADS-D ≥8 was used as a cutoff in the present study. A cutoff of 8 on the HADS-D subscale is most commonly used to identify clinically significant depression, with an average sensitivity and specificity of 0.80 [27,29].

Inflammation

Laboratory data including CRP were collected from the date of survey administration and were run in a Clinical Laboratory Improvement Amendments (CLIA) certified lab [30]. A CRP value was obtained by turbidimetric immunoassay. CRP elevation is consistent with greater inflammation. Inter- and intra-assay coefficient of variation is reliably less than 5%. A cut-point of 1 mg/dL was chosen in the present study because of its use in evaluating antidepressant responses in previous studies [24,25].

CRP is an acute-phase reactant protein that is reflective of innate inflammation and its relationship with several symptoms such as pain, fatigue and sleep in addition to depression have been evaluated previously [31]. CRP is one of the biomarkers that are most consistently associated with depression, along with the pro-inflammatory cytokines, TNF-α, IL-6 and IL-1β [32–34]. CRP has a long half-life and therefore less diurnal/circadian fluctuation than inflammatory cytokines and provides a good measure of general inflammation [35,36]. On the basis of this previous literature, the current study used CRP as the primary indicator of inflammation in this sample of patients with lung cancer.

Statistical analysis

The primary outcome of this study was survival when comparing the impact of depression and inflammation defined by the aforementioned clinically established cutoff criteria. Descriptive analyses were performed to assess distributions and central tendency for demographic and clinical covariates. Associations with depression and inflammation criteria were assessed using independent samples t-tests and chi-square tests of association. Covariates and outcomes of interest (i.e., inflammation and depression) were analyzed for survival by Cox linear regression using a hierarchical linear regression format. Covariates were chosen a priori, along with those that were statistically significant. Survival analyses from the time of survey were evaluated using Kaplan–Meier curve analysis for depression and inflammation individual and also by intergroup analysis. In addition, Cox hierarchical linear regression models were formed using depression, inflammation, covariates, and an interaction term. Because CRP data are not normally distributed, CRP values in the Cox regression model were log transformed prior to data analysis; however, untransformed values are also reported for ease of interpretation. Tests of mediation on the relationship between inflammation and survival (depression as the mediator) were performed using casual steps approach [37]. Depression was chosen as the mediator because inflammation had the larger effect on survival. Mediation would be demonstrated if the following criteria were met: the independent variable was shown to significantly influence the dependent variable in the first regression equation, independent variable was shown to significantly influence the mediator in the second regression equation and mediator must significantly influence the dependent variable in third equation when the independent variable and mediator are entered as predictors. To confirm mediation, a Sobel test was subsequently performed. Statistical procedures were performed using the SPSS version 24 software (SPSS, IL, USA; 2013), and all statistical tests were two-tailed with a 0.05 significance level.

Results

Cohort characteristics

Of 170 potential participants, survey information was returned by 123 individuals (70% response rate). Cohort characteristics are presented in Table 1. Key demographic data revealed an average age of 66.3 years old of predominately white females with adenocarcinoma, and the majority were receiving first-line treatment. The average amount of time since lung cancer diagnosis was 15.2 months (SD 12.8).

Table 1. . Clinical and demographic characteristics of the sample.

Covariates	Total (n = 123)	Meets depression criteria		Difference	Meets inflammation criteria		t
		Yes	No	t	Yes	No
	M (SD)	M (SD)	M (SD)		M (SD)	M (SD)
Age (years)	66.3 (9.3)	66.1 (9.2)	66.3 (9.3)	0.112; p = 0.91	67.2	65.6	-0.96; p = 0.34
Time with disease (months)	15.2 (12.8)	15.1 (12.6)	15.2 (12.9)	0.57; p = 0.96	14.8	15.2	0.19; p = 0.85
CRP (mg/dl)	2.1 (3.3), median 0.82	3.5 (3.7)	1.6 (3.1)	2.352; p = 0.01	4.1	0.35	-6.76; p < 0.001
Depression score (HADS-D)	5.2 (3.8)	10.9 (2.4)	3.5 (2.2)	-15.5; p < 0.001	6.7	4.1	-3.84; p < 0.001
		Yes	No		Yes	No
Meets screening criteria	N (%)	N (%)	N (%)	χ2	N (%)	N (%)	χ2
– Depression (HADS-D≥8)	28 (23%)	–	–		21	7	12.5; p < 0.001
– CRP ≥1 mg/dl, n = 120	55 (46%)	21 (38%)	34 (62%)	12.5; p < 0.001	–	–
Gender				χ2			χ2
– Male	44 (36%)	14 (32%)	30 (68%)	3.194; p = 0.07	23 (47%)	20 (53%)	t = 1.582; p = 0.21
– Female	79 (64%)	14 (18%)	65 (82%)		32 (42%)	45 (58%)
Disease type				F			F
– Adenocarcinoma	92 (75%)	18 (20%)	74 (80%)	2.138; p = 0.54	38 (42%)	52 (58%)	2.704; p = 0.44
– Squamous cell	6 (5%)	2 (33%)	4 (67%)		4 (33%)	2 (67%)
– Small cell lung	19 (15%)	6 (32%)	13 (68%)		11 (58%)	8 (42%)
– Unspecified	6 (4.9%)	2 (33%)	4 (67%)		2 (40%)	3 (60%)
Treatment type				F			F
– Chemotherapy	47 (38%)	15 (32%)	32 (68%)	7.591; p = 0.02	28 (39%)	18 (61%)	15.97; p < 0.001
– Immunotherapy	37 (30%)	9 (24%)	28 (76%)		18 (50%)	18 (50%)
– Targeted therapy	26 (21%)	1 (0.04%)	25 (99.6%)		3 (12%)	22 (88%)
– Missing	13 (11%)	–	–
Line of treatment				F			F
– First	53 (43%)	8 (15%)	45 (85%)	4.052; p = 0.13	22 (42%)	31 (68%)	1.166; p = 0.56
– Second	39 (32%)	12 (31%)	27 (69%)		19 (51%)	18 (49%)
– Third or beyond	22 (18%)	7 (32%)	15 (68%)		11 (55%)	10 (45%)
– Missing	9 (7%)
Race/ethnicity				χ2			χ2
– Nonwhite	18 (14.6%)	6 (33%)	12 (67%)	4.894; p = 0.18	10 (56%)	8 (44%)	0.806; p = 0.40
– White	105 (85%)	22 (21%)	83 (79%)		45 (44%)	57 (56%)
Married				χ2			χ2
– Yes	86 (70%)	18 (21%)	68 (79%)	0.547; p = 0.46	35 (42%)	48 (58%)	1.456; p = 0.23
– No	37 (30%)	10 (27%)	27 (73%)		20 (54%)	17 (46%)
Antidepressant				χ2			χ2
– Yes	26 (21%)	9 (35%)	17 (65%)	2.634; p = 0.11	10 (42%)	14 (58%)	0.210; p = 0.65
– No	97 (79%)	19 (20%)	78 (80%)		45 (47%)	51 (53%)

HADS-D: Hospital Anxiety and Depression Scale.

The mean depression score on the HADS-D was 5.2 with 28 participants (22.8%) meeting depression screening criteria (i.e., HADS-D ≥8). The average CRP was 2.1 mg/dl (SD 3.3) with 55 participants (46%) meeting the inflammation threshold of CRP ≥1 mg/dL. CRP was higher in patients meeting depression screening criteria (3.5 mg/dl vs 1.6 mg/dl) (t = 2.35; p = 0.01), and depression scores were higher in those patients with elevated inflammation (HADS-D 6.7 vs 4.1) (t = -3.84; p < 0.001) (Table 1). Of the demographic variables, only treatment type (namely, chemotherapy) was associated with depression (F = 7.59; p = 0.02) and inflammation (F = 15.97 p < 0.001).

Subcohort groups were created based on meeting threshold criteria for depression (HADS-D ≥8) and inflammation (CRP ≥1 mg/dl) and are defined as follows:

• Group 0 (no inflammation or depression): n = 61 (50%);

• Group 1 (inflammation but no depression): n = 34 (27%);

• Group 2 (depression but no inflammation: n = 7 (6%);

• Group 3 (both depression and inflammation): n = 21 (17%).

Survival data

Kaplan–Meier analysis

At analysis, 79 death events had occurred. The overall estimated mean survival was 515.4 days (SE 35.3) and median survival was 384 days (SE 52.0). The estimated mean survival for patients with depression (HADS-D ≥8) was 323.6 days (SE 56.0) versus 558.0 days (SE 38.9) for participants without depression (log rank χ²: 7.788; p = 0.005). The estimated mean survival for participants with inflammation (CRP ≥1 mg/dl) was 356.5 days (SE 41.1) versus 622.2 days (SE 46.9) for participants without inflammation (SE 24.2; log rank χ²: 12.546; p < 0.001) (Figure 1).

Figure 1. . Kaplan–Meier curve survival analyses of (A) patients with and without depression.

Defined as scoring ≥8 on the Hospital Anxiety and Depression Scale and (B) patients with and without inflammation-defined as C-reactive protein ≥1 mg/dl.

Participants were placed into groups (defined earlier) based on meeting criteria for depression (HADS-D ≥8) and inflammation (CRP ≥1 mg/dl) and assessed for survival based on Kaplan–Meier method (log rank χ²: 16.251; p = 0.001) (Figure 2). Pair-wise survival estimates are as follows:

• Group 0 (no inflammation or depression): estimated mean survival of 647.7 (SE 48.9);

• Group 1 (inflammation but no depression): estimated mean survival of 376.2 days (SE 51.5; median survival of 272 days; SE 68.2);

• Group 2 (depression but no inflammation): estimated mean survival of 392.3 days (SE 110.4; median survival of 234 days; SE 83.3);

• Group 3 (both depression and inflammation): estimated mean survival of 307.1 days (SE 60.6; median survival of 192 days; SE 16.8).

Figure 2. . Kaplan–Meier Curve analysis showing survival functions of four groups of patients: group 0, no inflammation and no depression; group 1, inflammation and no depression; group 2, depression and no inflammation; and group 3, both inflammation and depression.

Inflammation is defined as a C-reactive protein ≥1 mg/dl. Depression is defined as a score of ≥8 on the Hospital Anxiety and Depression Scale.

Pair-wise comparisons were significant for differences in survival between groups 0 and 1 (log rank chi-square = 10.147; p = 0.001) and between groups 0 and 3 (chi square 13.819; p < 0.001); however, group 3 was not significantly different from group 1 (log rank χ² = 0.677; p = 0.411) or group 2 (log rank χ²: 0.734; p = 0.39).

Cox regression

Depressive symptoms were associated with worse overall survival when controlling for age, sex and BMI. Overall, participants with depressive symptoms were more likely to die than participants with few depressive symptoms (hazard ratio [HR]: 1.115; 95% CI: 1.05–1.179) when controlling for age, sex and BMI (Model 1, Table 2) (change χ² = 19.725; p = 0.001). Elevated BMI was protective for survival with an HR of 0.928 (95% CI: 0.879–0.980). However, adding inflammation to the model incurred an even higher risk of death with an HR of 2.854 (95% CI 1.856–4.388; change χ² = 24.570; p < 0.001) (Model 2, Table 2). An interaction term was created for depression and inflammation, but it was not significantly association with survival (HR: 1.005; 95% CI: 0.898–1.126; change χ² = 0.009; p < 0.93) (model 3, Table 2).

Table 2. . Cox proportional hazards regression comparing the effects of depression and inflammation on overall survival.

Model 0	Survival			CI
	Regression coefficient	HR	p	Lower	Upper
Age	01 (0.01)	1.005	0.71	0.980	1.029
Sex	-0.32 (0.25)	0.727	0.73	0.446	1.186
BMI	-0.06 (0.03)	0.940	0.02	0.890	0.992
	χ2 change: 6.376; p = 0.10
Model 1	Survival			CI
	Regression coefficient	HR	p	Lower	Upper
Age	0.00 (0.01)	1.000	0.98	0.97	1.025
Sex	-0.30 (0.25)	0.739	0.23	0.451	1.210
BMI	-0.07 (.03)	0.928	0.007**	0.879	0.980
Depression (HADS-D)	0.11 (0.03)	1.115	<0.001***	1.055	1.179
	χ2 change: 13.349; p < 0.001
Model 2: adding inflammation	Survival			CI
	Regression coefficient	HR	p	Lower	Upper
Age	0.01 (0.01)	1.007	0.60	0.982	1.032
Sex	-0.14 (0.25)	0.870	0.58	0.544	1.421
BMI	-0.09 (0.03)	0.910	0.001	0.862	0.962
Depression (HADS-D)	0.08 (0.03)	1.080	0.007**	1.022	1.142
Inflammation (CRP)	1.05 (0.22)	2.854	<.001***	1.856	4.388
	Chi Square Change 24.381; p < 0.001
Model 3: depression + inflammation interaction	Survival			CI
	Regression coefficient	HR	p	Lower	Upper
Age	0.01 (0.01)	1.007	0.60	0.982	1.033
Sex	-0.14 (.25)	0.868	0.57	0.530	1.422
BMI	-0.09 (0.03)	0.91	0.001**	0.862	0.962
Depression (HADS-D)	0.08 (0.03)	1.079	0.016*	1.014	1.147
Inflammation (CRP)	1.02 (0.37)	2.775	0.006**	1.339	5.754
Depression*inflammation	0.01 (0.06)	1.005	0.93	0.898	1.126
	χ2 change: 0.009; p = 0.93

HADS-D: Hospital Anxiety and Depression Scale; HR: Hazard ratio.

Mediation analysis

The relationship between inflammation and survival was shown to be significantly mediated by depression while controlling for age, sex and BMI (Table 3; Figure 3). Analytic assumptions were carefully checked to confirm there was no violation due to potential multicollinearity between depression and inflammation given the observed significant relationship between depression and inflammation as noted earlier. All criteria for mediation were met: CRP significantly influenced survival (HR 3.2; p < 0.001) in the first regression equation (step 1); CRP significantly influenced depression in the second regression (β = 0.34; p < 0.001) (step 2); and depression (mediator) significantly influenced survival (HR: 1.08; p = 0.007), while the contribution of inflammation to reduced survival decreased from HR 3.14 to 2.85 with both predictors in the model (step 3). Partial mediation was confirmed by a Sobel test (2.17, SE 0.07; p = 0.03).

Table 3. . Regression analysis testing the mediation hypothesis in three subsequent steps.

	Step 1: path C
	Survival			CI
	Regression coefficient	HR	p	Lower	Upper
Controls
Age	01 (0.01)	1.008	0.52	0.983	1.034
Sex	-0.13 (0.25)	0.881	0.61	0.539	1.440
BMI	-0.08 (0.03)	0.921	0.003	0.873	0.973
IV
CRP (log	1.14 (0.22)	3.140	<.001	2.06	4.786
	χ2: 37.223; p < 0.001
	Step 2: path A
	Depression			CI
	Regression coefficient	B	p	Lower	Upper
Controls
Age	0.00 (0.04)	0.00	0.98	-0.070	0.073
Sex	0.29 (0.71)	0.04	0.69	-1.132	1.706
BMI	-0.07 (0.03)	0.00	0.97	-0.133	0.139
IV
CRP (log)	2.02 (0.54)	0.34	<0.001***	0.947	3.086
	Adjusted R2 = 0.085; F = 3.742; p = .007
	Step 3: path C’ and B
	Survival			CI
	Regression coefficient	HR	p	Lower	Upper
Controls
Age	0.01 (0.01)	1.007	0.60	0.982	1.032
Sex	-0.14 (.25)	0.870	0.58	0.533	1.421
BMI	-0.09 (0.03)	0.910	0.001	0.862	0.962
IV
CRP (log)	1.05 (0.22)	2.854	<0.001	1.856	4.388
Mediating variable
Depression	0.08 (0.03)	1.080	0.007	1.022	1.142
	χ2: 44.106; p < 0.001

CRP: C-reactive protein; HR: Hazard ratio; IV: Independent variable.

Figure 3. . Mediation model.

Depression as a mediator of inflammation on survival.

**p < 0.01; ***p < 0.001.

Discussion

Depression and inflammation were independently associated with worse survival in patients with metastatic lung cancer. Depression predicted inferior survival even when controlling for inflammation and also mediated the association between inflammation and poor survival. Patients who reported depression and exhibited significant inflammation demonstrated the lowest survival rates. Similar findings were reported by Cohen and colleagues, who described changes in cortisol slope (a nonperipheral blood measure inflammation), and depression independently predicted worse survival in patients with metastatic renal cell carcinoma [38]. Although it is encouraging that their study led to similar results, our study is the first to report this association using a plasma biomarker of peripheral inflammation (i.e., CRP), which is much more easily obtained from patients, interpreted clinically and incorporated into practice with reasonable turn-around time and cost. Depression is an actionable prognostic risk factor that not only leads to psychological suffering but may interfere with survival gains made with modern anticancer therapeutics.

Depression continues to be underrecognized and underdiagnosed in cancer clinics despite ongoing recommendations to screen for distress and other psychological maladies associated with cancer [39]. Therefore, an increased emphasis should be placed on factors that predict for depression in oncology settings and can inform the psychosocial management of cancer patients beyond a dichotomous psychosocial triage assessment [40,41]. Most patients with elevated distress scores will decline additional care [42]. Patients with the most severe depression are also more likely to decline psychosocial care as apathy and lack of executive functioning are hallmarks of depression [43,44]. Inflammation is not only a risk factor for depression but for even worse survival when combined with depression. Therefore, a patient with both depression and elevated inflammation risks inferior survival and would benefit to an even greater extent from in-depth questioning and motivational interviewing to accept psychosocial care.

The presence of depression and inflammation on survival may be additive but is not likely to be multiplicative as their interaction term was not close to significant. Although the analysis was not powered to be sensitive to an interaction effect, there was not a trend leading to the suggestion of an interaction effect. Perhaps, a multiplicative effect would be observed in a setting without anticancer treatments or other factors that mitigate the survival effects of depression and inflammation. Any additive deleterious effect may have been mitigated by other protective factors (e.g., cancer and depression treatments, social supports, individual resilience). For example, patients with higher inflammation and more aggressive disease may undergo more aggressive therapies with more frequent changes in therapy that are not directly accounted for in this analysis. The mechanisms by which depression and inflammation lead to inferior survival overlap, hence the mediation effect, and are also distinct because they are both independently associated with worse survival. A cumulative (additive) effect of depression and inflammation on survival was observed on the Kaplan–Meier plot but did not reach a statistical significance. Given that the majority of patients with depression also had elevated inflammation, pair-wise analysis was limited due to a low number of patients in subgroup 2 (depression and no inflammation).

There are some putative mechanisms by which depression and inflammation or immune dysregulation could interact to lead to inferior survival. For example, a study of depression in patients with hepatobiliary carcinoma found that NK cell depletion mediated the relationship between depression and worsened survival and that there may be a relationship between depression and lymphovascular invasion [45]. Another study found that inflammation was associated with certain depressive symptoms (e.g., fatigue, poor appetite and insomnia) but that other cognitive symptoms not associated with inflammation (e.g., loss of pleasure or anhedonia) also had specific survival implications [46]. The current study reveals that depression and inflammation have related and independent effects on survival through different mechanisms without further suggesting exactly what those mechanisms are. Thus, further research is warranted because these answers may lead to improved depression management strategies in patients with lung cancer and other metastatic cancers.

It should be noted that CRP reflects inflammation that is part of the innate immune system as opposed to the antibody producing adaptive or acquired immune system. These systems are related but represent different immune strategies that are both activated in the presence of cancer [47]. The association with depression and prognosis is more thoroughly described with the innate immune system (i.e., generalized inflammation) but has been observed in association with adaptive immune system activation as well [48]. The relationship between these two immune pathways is being further investigated with relation to cancer initiation and promotion [49]; its relationship with cancer-related prognosis, depression and other symptoms also deserves further investigation.

These data may help researchers refine questions regarding depression treatments in the lung cancer setting with excessive inflammation. Past results of depression treatments on survival effects have been mixed, but cancer-related inflammation may play a role deciphering why a treatment works or does not work. For example, some studies have shown that the survival implications of depression may only be relevant for certain clinical categories (e.g., younger patients with early-stage disease) [50]. Also, improvement in depressive symptoms after depression intervention trials (SMaRT Oncology-2 and -3 trials) did not lead to a difference in survival among patients with various solid tumors [51], nor was improvement in depression responsible for improved survival in a sample of lung cancer patients who were treated with early palliative care [52]. These studies demonstrate that the association between depression and survival in cancer populations is complicated, making it difficult to detect changes in survival after depression interventions [53]. Accounting for the presence of inflammation and perhaps changes in inflammation throughout a course of antidepressant treatment may further reveal why some populations demonstrate particular survival benefit but others do not.

Clinicians should be highly attentive to patients who are depressed and demonstrate excessive inflammation. In these cases, it would be appropriate to (pro)actively pursue depression management along the cancer disease trajectory to ensure the best possible outcomes. Inflammation may be associated with not only depression but depression treatment refractoriness [54]. Encouragingly, antidepression treatments that target inflammation directly or downstream effects of inflammation are currently being elucidated and will certainly be highly relevant for lung and other cancer patients [55]. For example, reduction of inflammation through anticytokine agents such as the IL-6 inhibitor tocilizumab or NSAIDs ameliorate depressive symptoms [56–58]. In addition, antidepressant medications that specifically address the downstream effects of inflammation on the CNS that lead to depression (i.e., reduction of neurotransmitters dopamine and serotonin) show promise but should be evaluated in relevant cancer population given its applicability [59,60].

Conclusion

Overall, this study contributes to the relationship between inflammation and depression in patients with cancer and their unique survival implications in particular. The study was limited to its single institution setting with a relatively small sample size, especially a limited number of depressed patients without inflammation. Also, the mediation analysis was limited by absence of a temporal design, so the directionality of the relationship between depression and inflammation on survival cannot be directly inferred. However, the homogenous sample of patients with lung cancer limits speculation regarding variation depending on cancer type and increases external validity for patients with lung cancer as the most frequently diagnosed cancer worldwide. Another element of this study that many not be recognized specifically as a strength is the use of CRP as the predictive inflammatory biomarker. This lends itself to reproducibility and external validity because this acute-phase reactant is relatively inexpensive, provides an accurate and stable picture of general inflammation and its results are fairly straightforward to interpret. A meta-analysis of CRP and depression by Horn and colleagues found that the association was still significant after controlling for other variables that increase inflammation (age, sex, socioeconomic status, BMI, medication and other substance use and comorbidities) [61]. However, many of these inflammation-inducing factors are also associated with depression (e.g., smoking, obesity and other medical comorbidity), and its implications for depression treatment are still valid regardless of which factors may be contributing to CRP elevation as reported in a follow up editorial by Felger and colleagues [60].

In summary, clinically significant depression and inflammation are independently and collectively associated with inferior survival in patients with metastatic lung cancer. These factors carry prognostic weight, can be measured easily and are clinically important actionable items that may change patients' trajectory with lung cancer if addressed and acted on early enough. Further research is warranted to understand these specific mechanisms and evaluate the interaction between inflammation and depression in cancer settings.

Summary points.

• Lung cancer–related inflammation is associated with depression and partially explains high rates of depression in patients with metastatic lung cancer.

• C-reactive protein (CRP) appears to be a reliable marker for identifying inflammation-related depression in this population, and a cut-point of 1 mg/dl is associated with inferior survival.

• Cancer-related inflammation as measured by CRP could help identify cases of depression given that the identification and treatment of depression in the cancer center remains challenging.

• Both depression and elevated inflammation independently predict worse survival outcomes.

• The combination of depression and elevated inflammation (CRP >1 mg/dl) is associated with even greater decrement in survival estimation.

• The relationship between inflammation and poor survival is mediated by depression.

• Depression is a modifiable risk factor for poor survival and should be addressed early in the lung cancer trajectory to potentially modify outcomes.

• Future research is warranted in this area of outcomes in cancer-related inflammation and depression.