Association Between Psychological Distress and Site‐Specific Cancer Incidence: A Systematic Review and Meta‐Analysis

ABSTRACT

Objective

Depression and anxiety are hypothesized to increase the risk for the development of chronic disease(s), including cancer. Several published observational studies and meta‐analyses have investigated the link between depression and anxiety in cancer but have reported conflicting results. To clarify this relationship, a meta‐analysis of observational studies reporting an association between depression, anxiety, and cancer incidence was performed.

Methods

Prospective and retrospective cohort studies investigating the relationship between depression, anxiety, and cancer incidence were included. A literature search of the PubMed was performed, with the final search conducted in March 2026. Egger's test and funnel plots were used to evaluate publication bias. Meta‐analysis was performed using a random‐effects model because high heterogeneity among the studies was expected.

Results

In the analysis including both retrospective and prospective studies, patients with depression or anxiety exhibited an elevated risk for lung, hepatobiliary‐pancreatic (HBP), hematological, and kidney, ureter, and bladder (KUB) cancers. In the analysis including only prospective studies, patients with depression or anxiety exhibited an elevated risk for female reproductive organ and lung cancer. Finally, in the analysis including only retrospective studies, patients with depression or anxiety exhibited an elevated risk for lung, HBP, hematological, KUB, and oral cancers, and a notably lower risk for female reproductive organ cancer.

Conclusions

Results of this study revealed that depression and anxiety were associated with an elevated risk for specific cancers. A future meta‐analysis including only prospective cohort studies with the same measures of depression or anxiety may further clarify this relationship.

1. Background

Depression and anxiety are some of the most common mental disorders, accounting for > 50% of the global disease burden associated with disability‐adjusted life years caused by mental disorders [1]. Depression and anxiety are hypothesized to increase the risk for the development of chronic disease(s), including cancer. Several mechanisms linking depression and anxiety to cancer have been proposed, such as dysregulation of the hypothalamic–pituitary–adrenal (HPA) axis and/or those influencing DNA repair [2]. Depression and anxiety may also increase the risk for cancer through lifestyle factors because individuals diagnosed with depression and anxiety are more likely to be obese, sedentary, consume alcohol, and/or smoke cigarettes [3].

Several published meta‐analyses based on longitudinal observation studies—including cohort and nested case–control designs in which depression ascertainment preceded cancer diagnosis—have addressed the link between depression and anxiety in cancer but have reported conflicting results. Some have demonstrated an increased risk for depression and anxiety [4, 5, 6], whereas others have found little to no risk [7, 8, 9, 10]. The inconsistency in findings across studies may be attributable to some studies combining different site‐specific cancers because different mechanisms underlie different tumor types; thus, depression and anxiety may influence various tumors in different ways. In addition, combining retrospective and prospective data may obscure the relationship between depression, anxiety, and cancer because retrospective data may be susceptible to recall or selection bias.

Most recently, van Tuijl et al. conducted an individual participant data (IPD) meta‐analysis using 18 prospective cohorts from the Psychosocial Factors and Cancer Incidence (PSY‐CA) consortium and reported that depression and anxiety were primarily associated with lung and smoking‐related cancers, whereas associations with most other cancer types were not observed after extensive adjustment for established risk factors [9]. Nevertheless, several questions remain. First, additional cohort studies have been published since the search period of the IPD meta‐analysis. Second, there are various studies including both prospective and retrospective cohort studies that were not part of the PSY‐CA consortium. Comparing findings across different study designs may help identify potential sources of heterogeneity and explain discrepancies reported in the literature.

Therefore, we performed a systematic review and meta‐analysis of longitudinal cohort studies to quantify the association between depression and anxiety, and subsequent site‐specific cancer incidence. By incorporating recently published cohort studies, we sought to provide an updated synthesis of the available evidence while examining whether study design (prospective vs. retrospective) influenced the observed associations. Since retrospective and prospective designs may yield different estimates due to differential susceptibility to bias and reverse causation, we conducted analyses stratified by study design. In addition, we focused on cancer site–specific associations to avoid masking heterogeneity across tumor types. Our goal was to provide an updated and methodologically transparent synthesis of the epidemiologic evidence and to identify cancer sites for which associations appear most consistent across study designs.

2. Methods

A meta‐analysis of observational studies that reported an association between depression, anxiety, and the incidence of cancer was performed. This study was performed according to a pre‐specified protocol registered with the Open Science Framework (OSF registration: DOI 10.17605/OSF.IO/S3Y6R), and conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines. The PRISMA checklist is presented in Supporting Information S1: Table 1.

2.1. Literature Search Strategy and Selection Criteria

A systematic literature search of the PubMed database was performed, with the final search conducted in March 2026, using the following key words: (“cancer” OR “carcino*” OR “tumor” OR “leukemia” OR “lymphoma”) AND (“anxiety” OR “depress*”) AND (“longitudinal” OR “prospective”). Two independent investigators (SK and CP) reviewed the retrieved studies by examining the titles, abstracts, and full texts to determine which should be included or excluded. In addition, the Embase database was also searched for other relevant, potentially eligible studies; however, the studies retrieved overlapped with those found in the PubMed search. The search strategy is presented in Supporting Information S1: Figure S1.

The study selection criteria were established in accordance with the Population, Intervention, Comparison, Outcomes and Study (i.e., “PICOS”) strategy (Supporting Information S1: Table 2). Cohort studies that investigated the relationship between depression and cancer incidence and anxiety and cancer incidence were included. The included studies reported risk estimates (relative risk, hazard ratio [HR], and odds ratio) with corresponding 95% confidence interval (CI). Both prospective and retrospective cohort studies investigating measures or diagnostic methods for depression or anxiety were included.

Case‐control studies, studies lacking data necessary for performing a meta‐analysis (e.g., effect size, 95% confidence interval, and standard deviation), those that were not published in English, and those that had overlapping cohorts with other included studies were excluded. When multiple studies had overlapping cohorts, only 1 cohort study was selected for inclusion. Among the studies with overlapping cohorts, the study with the largest number of participants was included.

2.2. Data Extraction

For each included study, 2 investigators (STK and CIP) independently extracted the relevant data and discrepancies were resolved through consensus discussions with JIK and JIS. The following data were extracted from the included studies: first author; year of publication; study location (cohort); type of cancer; study design; measure(s) of depression or anxiety; adjusted variables; and effect size with corresponding 95% CI.

2.3. Statistical Analysis

Meta‐analysis was performed using a random‐effects model because high heterogeneity among the studies was expected. Risk estimates with corresponding 95% CIs are presented in tables and forest plots. Heterogeneity among the studies was measured using the Higgins' I ² statistic. In this study, I ² > 50% or p < 0.1 was considered to be indicative of high heterogeneity. In addition, Egger's test and funnel plots were used to evaluate publication bias, with p < 0.1 considered to be indicative of publication bias. Two investigators (STK and CIP) independently assessed the quality of the articles using the Newcastle–Ottawa Scale (NOS), which evaluates cohort selection, comparability, and outcome assessment [11]. The meta‐analysis was performed using R software (R Core Team; R Foundation for Statistical Computing, Vienna, Austria).

3. Results

The systematic literature search strategy, from database inception up to March 2026, and the study selection process are presented in Supporting Information S1: Figure 1. The initial literature search retrieved 7448 studies. After excluding duplicates, titles, abstracts, and full‐texts were screened, leaving 48 studies that fulfilled the inclusion criteria. All of the studies included were published in English.

3.1. Study Characteristics

The characteristics of the included studies are summarized in Table 1. Of the 48 studies from 11 countries, 11 were retrospective and 37 were prospective. Breast (n = 35 [72.9%]), prostate (n = 26 [54.2%]), lung and bronchial (n = 27 [56.3%]), and colorectal (n = 26 [54.2%]) cancers were the most common cancer types among the included studies. For the measures of depression or anxiety, the International Classification of Diseases, Eighth, Ninth, or Tenth Revisions (i.e., “ICD‐8, −9, or −10”) (n = 14 [29.2%]) and the Center for Epidemiological Studies‐Depression Scale (i.e., “CES‐D”) (n = 13 [27.1%]) were the most common measures among the included studies.

TABLE 1.

Summary of included studies.

	N (%)
Total number of studies	48
Retrospective cohort	11 (22.9%)
Prospective cohort	37 (77.1%)
Number of studies by cancer type
Thyroid	3 (6.3%)
Stomach	5 (10.4%)
Prostate	26 (54.2%)
Female reproductive organ	10 (20.8%)
Skin	7 (14.6%)
Lung and bronchus	27 (56.3%)
Hepatobiliary‐pancreatic	6 (12.5%)
Brain	3 (6.3%)
Breast	35 (72.9%)
Colorectal	26 (54.2%)
Esophagus	4 (8.3%)
Hematopoietic and lymphoid	5 (10.4%)
Kidney, urinary, bladder	6 (12.5%)
Oral	3 (6.3%)
Number of studies by measure of depression or anxiety
BAI	1 (2.1%)
BDI	1 (2.1%)
CCI	2 (4.2%)
CES‐D	13 (27.1%)
DSM‐III	3 (6.3%)
DSM‐IV	2 (4.2%)
Edinburgh post‐natal depression scale	2 (4.2%)
HADS	3 (6.3%)
GAD7	2 (4.2%)
GHQ‐12 or 30	2 (4.2%)
ICHPPC‐2	1 (2.1%)
HPL	1 (2.1%)
ICD‐8, 9, or 10	14 (29.2%)
ICHPPC‐2	1 (2.1%)
IDS	1 (2.1%)
MHI‐5	2 (4.2%)
MINI	1 (2.1%)
MMPI	1 (2.1%)
PHQ‐4, 9, or 10	6 (12.5%)
Physician diagnosis	2 (4.2%)
PSE	1 (2.1%)
SCL	1 (2.1%)
Burnam regression algorithm	1 (2.1%)

Abbreviations: BAI, Beck anxiety index; BDI, Beck depression index; CCI, Crown–Crisp phobic anxiety index; CES‐D, Center for Epidemiological Studies‐Depression Scale; DSM, Diagnostic and Statistical Manual of Mental Disorders; GAD7, Generalized Anxiety Disorder‐7; GHQ, General Health Questionnaire; HADS, Hospital Anxiety and Depression Scale; ICD, International Classification of Diseases; ICHPPC‐2, international classification of health problems in primary care; IDS, nventory of Depressive Symptomatology; MHI‐5, Mental Health Inventory 5; MINI, Mini‐International Neuropsychiatric Interview; MMPI, Minnesota Multiphasic Personality Inventory; PHQ, Patient Health Questionnaire; PSE, Present state examination; SCL, Symptom checklist.

3.2. Meta‐Analyses According to Cancer Type

Meta‐analysis results according to cancer type are summarized in Table 2 and Figure 1. Risk of lung cancer was elevated across all analyses. Forest plots of individual cancer types are presented in Supporting Information S1: Figure 3 (Figure S3a‐n). Meta‐analysis results and study characteristics of each individual study are presented in Supporting Information S1: Table 3 (Table S3a‐n).

TABLE 2.

Summary of meta‐analyses by cancer type.

	Retrospective and prospective studies				Prospective studies only				Retrospective studies only
	Random effects (95% CI)	p	I 2% (p‐value)	Egger's p	Random effects (95% CI)	p	I 2% (p‐value)	Egger's p	Random effects (95% CI)	p	I 2% (p‐value)	Egger's p
Thyroid	There are only three retrospective studies								1.34 (0.81–2.23)	0.25	58 (0.09)	0.24
Stomach	1.26 (0.96–1.66)	0.09	65 (0.02)	0.20	2.38 (0.29–19.65)	0.42	NA		1.26 (0.96–1.66)	0.12	73 (0.01)	0.17
Prostate	1.02 (0.95–1.08)	0.65	71 (< 0.01)	0.56	1.00 (0.97–1.03)	0.99	11 (0.30)	0.19	1.40 (0.88–2.25)	0.16	93 (< 0.01)	< 0.0001
Female reproductive organ	1.03 (0.93–1.15)	0.52	50 (0.01)	0.22	1.16 (1.09–1.24)	< 0.01	0 (0.44)	0.23	0.87 (0.78–0.98)	0.02	0 (0.71)	0.74
Skin	1.01 (0.89–1.15)	0.85	49 (0.05)	0.32	1.10 (0.79–1.51)	0.58	5 (0.37)	0.34	0.99 (0.86–1.15)	0.92	68 (0.01)	0.02
Lung	1.20 (1.13–1.28)	< 0.01	71 (< 0.01)	0.61	1.15 (1.10–1.21)	< 0.01	30 (0.07)	0.81	1.38 (1.20–1.58)	< 0.01	78 (< 0.01)	0.05
HBP	1.40 (1.08–1.80)	0.01	82 (< 0.01)	0.36	1.65 (0.21–12.69)	0.63	NA		1.39 (1.08–1.81)	0.01	84 (< 0.01)	0.30
Brain	There are only three retrospective studies								1.82 (0.79–4.21)	0.16	84 (< 0.01)	0.01
Breast	1.01 (0.96–1.07)	0.60	73 (< 0.01)	0.21	0.99 (0.94–1.03)	0.52	52 (< 0.01)	0.31	1.18 (0.95–1.47)	0.13	92 (< 0.01)	0.11
Colorectal	1.02 (0.97–1.08)	0.46	53 (< 0.01)	0.81	0.99 (0.96–1.03)	0.74	6 (0.37)	0.60	1.12 (0.96–1.30)	0.15	77 (< 0.01)	0.17
Esophagus	There are only four retrospective studies								1.49 (0.81–2.75)	0.20	87 (< 0.01)	0.48
Hematological	1.15 (1.04–1.27)	0.01	0 (0.69)	0.80	1.29 (0.45–3.76)	0.63	53 (0.14)	0.14	1.16 (1.05–1.28)	< 0.01	0 (0.68)	0.98
KUB	1.69 (1.32–2.16)	0.00	88 (< 0.01)	< 0.0001	2.89 (0.66–12.59)	0.16	NA		1.67 (1.30–2.14)	< 0.01	89 (< 0.01)	< 0.0001
Oral	There are only three retrospective studies								1.89 (1.29–2.64)	< 0.01	80 (< 0.01)	0.33

Note: Bold values indicate statistically significant items.

Abbreviations: CI, confidence interval; HBP, hepatobiliary and pancreatic cancers; KUB, kidney, ureter, and bladder cancers; NA, not available.

FIGURE 1.

Differences of effect size and 95% confidence interval among the different cancer type.

Between‐study heterogeneity was high for several cancers in retrospective studies, and only for hematological cancer in prospective studies. Publication bias was present in prostate (Egger's p < 0.0001), lung (Egger's p = 0.05), skin (Egger's p = 0.02), brain (Egger's p = 0.01), and KUB (Egger's p < 0.0001) cancers in the analysis of retrospective studies, and none in the analysis of prospective studies. Funnel plots for the studies investigating each cancer type are presented in Supporting Information S1: Figure 2 (Figure S2a‐n).

4. Discussion

In this systematic review and meta‐analysis of 48 cohort studies, depression and/or anxiety were associated with increased incidence of several site‐specific cancers, most consistently lung cancer. Importantly, effect estimates differed by study design: prospective studies showed more limited and less heterogeneous associations, whereas retrospective studies yielded broader patterns of associations with substantial heterogeneity and evidence of publication bias for several cancer sites. These findings suggest that methodological differences—including exposure ascertainment, confounding control, and susceptibility to bias—likely contribute to previously inconsistent results in the literature.

Our results revealed that depression and anxiety were associated with an elevated risk for specific cancers. Our findings are consistent with those of previous meta‐analyses revealing that the associations differed among various cancer types [5, 6, 8, 9]. A comparison of the findings from previous meta‐analyses and the present study according to cancer type is provided in Supporting Information S1: Table 5. While the overall pattern was generally consistent across studies, particularly for lung cancer, additional associations were observed in the present study, which may reflect differences in study design, confounding adjustment, and inclusion criteria. Study heterogeneity was high for most cancers in the analysis of retrospective studies but only for hematological cancer in the analysis of prospective studies. In addition, while there was publication bias for 5 cancer types in the retrospective studies, Egger's test did not detect statistically significant evidence of publication bias for any of the cancers in the prospective studies.

Notably, the risk for lung cancer significantly increased among patients with depression or anxiety across all analyses (i.e., retrospective and prospective, prospective only, and retrospective only), which is consistent with previous meta‐analyses that reported similar associations. In post‐hoc subgroup analyses, studies that controlled for smoking (HR 1.14, p < 0.001) and those that did not (HR 1.44, p < 0.001) demonstrated a significantly elevated risk for lung cancer, indicating that smoking was a major mediator in the relationship between depression or anxiety and lung cancer, although additional mechanisms are likely to be involved. For example, a prospective study found that lifetime pack‐years smoked accounted for 38% of the overall association between depressive symptoms and lung cancer risk [12]. While depression is positively associated with smoking, a well‐established risk factor for lung cancer, other pathways, such as dysregulation of the HPA axis and increased oxidative stress leading to DNA damage, may also contribute to tumorigenesis in individuals diagnosed with depression or anxiety.

Our findings should be interpreted in the context of the IPD meta‐analysis by van Tuijl et al., which reported significant associations primarily for lung and smoking‐related cancers. In addition to lung cancer, our study identified associations for hepatobiliary‐pancreatic, hematological, kidney/ureter/bladder, and female reproductive cancers. These differences may be explained by several methodological factors. First, unlike the IPD meta‐analysis, the present review synthesized evidence from both prospective and retrospective cohort studies. Notably, retrospective studies in our review showed substantially greater heterogeneity and more frequent evidence of publication bias than prospective studies. Second, adjustment for important confounding variables varied considerably across the included studies, particularly with respect to smoking, alcohol use, body mass index, and socioeconomic status. Therefore, the additional associations observed in our study should be interpreted cautiously until confirmed by future large‐scale prospective studies with comprehensive adjustment for established cancer risk factors.

Because combining prospective and retrospective studies in the same analysis can confound the results, we separated the analysis into three components: retrospective and prospective; prospective only; and retrospective only. Interestingly, female reproductive organ cancer, which exhibited no association in the combined analysis, exhibited an increased risk in the analysis of prospective studies and a decreased risk in the analysis of retrospective studies. In the combined analysis, the I² value was 50%, which indicated moderate heterogeneity, whereas in the prospective‐only and retrospective‐only analyses, the risk for heterogeneity was low. When we performed a post‐hoc subgroup analysis to determine why the results were opposite, body mass index, which was controlled for in 5 of 6 prospective studies and none of the retrospective studies, was determined to be a source of heterogeneity among studies investigating female reproductive organ cancer (test of moderators [coefficient 2], QM [degrees of freedom (df) = 1] = 16.6391; p < 0.0001). In addition, smoking status, which, again, was controlled for in 5 of 6 prospective studies and none of the retrospective studies, was determined to be another source of heterogeneity (test of moderators [coefficient 2], QM [df = 1] = 6.7870; p = 0.0092). Smoking and obesity are considered to be important risk factors for the development of various female reproductive cancers [13, 14, 15]. In our meta‐analysis, controlling for these variables had a significant influence on the overall results.

In the analysis of retrospective studies, study heterogeneity was high for all cancer types except female reproductive organ and hematological cancers, whereas in the analysis of prospective studies, study heterogeneity was high only for hematological cancer. The high study heterogeneity for hematological cancer in the analysis of prospective studies may have been due to the small number of such studies (n = 2). The difference in the frequency of heterogeneity between the 2 study types was expected because prospective studies have fewer potential sources of bias and confounding factors than retrospective studies.

5. Clinical Implications

The present findings suggest that depression and anxiety may be associated with the incidence of selected cancers, particularly lung cancer, although causality cannot be inferred from the available evidence. These findings highlight the importance of considering psychological distress as one component of comprehensive health assessment, particularly because depression and anxiety are closely linked to modifiable health behaviors such as smoking, alcohol consumption, physical inactivity, and obesity, which may contribute to cancer risk. Further research is needed to clarify whether effective identification and management of psychological distress, together with interventions targeting unhealthy behaviors, may contribute to reducing long‐term cancer risk. In addition, future prospective studies with standardized assessments of psychological distress and adequate control for confounding factors will be essential to clarify the mechanisms underlying these associations.

6. Limitations

This study has several limitations. First, exposure definitions for depression and anxiety varied widely across studies, ranging from clinical diagnoses based on ICD codes to symptom scales such as CES‐D, limiting comparability and precluding dose–response analyses. Second, residual confounding may be present, particularly for smoking, alcohol use, physical activity, and socioeconomic factors, which were inconsistently measured and adjusted for across studies. Third, reverse causation remains a concern, especially in retrospective studies and in analyses without lag periods excluding early cancer events. Fourth, the number of studies for several cancer sites was small, limiting statistical power and precision and increasing the likelihood of small‐study effects. Fifth, the observational nature of included studies limits causal inference, and the pooled estimates should be interpreted as associations rather than evidence of causality. Finally, although we added follow‐up duration for individual studies where available, the reporting of follow‐up was heterogeneous across studies (mean, median, estimated from person‐years, or not reported), which precluded a formal meta‐regression according to follow‐up duration. Despite these limitations, our results contribute valuable evidence to the relatively sparse literature addressing the associations among depression, anxiety, and cancer incidence.

7. Conclusion

In conclusion, this meta‐analysis suggests that psychological distress, including depression and anxiety, is associated with an increased risk for specific cancers, most notably lung cancer. Future large‐scale prospective studies using standardized psychological measures, along with investigations of the underlying biological mechanisms, are warranted to further elucidate these associations and the pathways through which they may operate.

Author Contributions

J.I.K. and J.I.S. are joint corresponding authors. S.T.K., S.J.K., J.I.K., and J.I.S. conceived and designed the study. S.T.K. and C.I.P. performed the article search and extracted the data. S.T.K., J.Y.K., J.I.K., and J.I.S. analyzed and interpreted the data. Funding acquisition and overall study supervision were managed by S.J.K., J.I.K., J.I.S. All authors contributed to the writing and revision of the manuscript and approved the final draft. J.I.S. is the guarantor. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Funding

This work was supported by the National Research Foundation of Korea (NRF) grants funded by the Korean Government (RS‐2025‐23963393 to J.I.K. and RS‐2023‐00209077 to S.J.K.), and the Yonsei Fellowship, funded by Lee Youn Jae (to J.I.S.).

Ethics Statement

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Analytic Code Availability

The analytic code associated with this study are available on request from the corresponding author, [J.I.S.].

Supporting information

Supporting Information S1

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author, [J.I.S.].