The impact of anxiety and depression on hematologic malignancy outcomes: a systematic review and meta-analysis.

Introduction
Cancer is the second leading cause of morbidity and mortality globally, resulting in a significant economic and public health burden [1]. Hematologic malignancies (HM) are a significant component of this global health challenge, ranking among the most prevalent cancer types worldwide. In 2022, estimations showed 1.3 million new diagnoses of hematologic malignancies globally, which accounted for approximately 730,000 deaths [2]. These malignancies originate from hematopoietic cells residing in the bone marrow, peripheral blood, lymph nodes, and spleen [3]. The 5th WHO Classification of Hematolymphoid Neoplasms categorized hematologic malignancies into two distinct subtypes: myeloid and lymphoid [4].
Despite advancements in treatment and management of these malignancies enhancing 5-year relative survival from 49.9% (1975–1977) to 72.04% (2014–2020), certain subtypes, like acute myeloid leukemia (AML), remain challenging, with a 5-year relative survival of 33.64% (2014–2020) and a particularly grim 2-year survival for those aged 75 + (12.5%), according to the SEER 8 database [5]. In addition to the poor prognosis of some hematologic malignancies, certain types, such as multiple myeloma, remain incurable [6, 7]. Beyond the disease and its symptoms, the treatment of malignancies and chemotherapy-induced side effects, prolonged hospitalization, social isolation, financial burden, fear of death, treatment failure and relapse can all contribute to the development of anxiety and depression in patients [8–13].
Depression and anxiety are among the most prevalent psychiatric disorders in patients diagnosed with cancer [14–17]. While the prevalence of these disorders varies based on geographical region, cancer type, and assessment scales, studies suggest a substantial burden of these conditions [18, 19]. In the meta-analysis conducted by Zhang et al. [14], the prevalence of anxiety and depression among patients with cancer was reported to be 31.3% and 32.5%, respectively. Studies specifically examining hematologic malignancies showed depression and anxiety prevalence ranging from 20 to 30% [8, 9, 15]. Individuals with hematologic malignancies have a 6.7-fold increased risk of developing depression compared to the general population [6].
Anxiety and depression significantly impact the prognosis and survival of patients with hematological malignancies [9, 18]. These psychological conditions can decrease treatment adherence, reduce the efficacy of therapies, and impair overall quality of life, potentially leading to increased mortality. Several studies have highlighted the association between depression and reduced survival in patients with AML [20]. Depressive symptoms have been linked to worse overall survival (OS) and event-free survival (EFS) in patients with AML, possibly due to decreased adherence to treatment regimens and the physiological effects of depression on the immune system [20–22]. Additionally, anxiety has also been associated with poorer OS, although its impact appears to be less consistent across studies [20, 22].
A systematic review and meta-analysis examining the impact of anxiety and depression on hematologic malignancy outcomes is crucial given the increasing evidence of their detrimental effects. While existing research indicates that these psychological factors adversely affect various types of cancer, findings related to hematologic malignancies remain inconsistent [18, 19, 23, 24]. This underscores the necessity for a comprehensive meta-analysis to consolidate the evidence and clarify the true effects of anxiety and depression on treatment response, disease progression, and overall survival.
This study tested the hypothesis that depression and anxiety are independently associated with reduced overall survival (OS) and event-free survival (EFS) in patients with hematologic malignancies. Specific aims included (1) quantifying the strength of this association across HM subtypes (e.g., AML, lymphoma), (2) exploring potential mechanisms (e.g., inflammatory vs. behavioral pathways), and (3) evaluating the impact of publication bias on pooled estimates.

Materials and methods

Protocol and registration
This systematic review was performed following the PRISMA guidelines [25]. We outlined our research question using the PICO framework (Population: individuals diagnosed with hematologic malignancies; Intervention: diagnosis with depression or anxiety; Comparator: Patients with HM but without a depression or anxiety diagnosis; Outcome: survival). The review protocol is registered with the International Prospective Register for Systematic Reviews (PROSPERO) under the identifier CRD42024568789 to ensure methodological transparency.

Literature search methodology
The PubMed, SCOPUS, and Web of Science databases were searched for broad coverage of medical and clinical studies up to July 2024. Search terms included “hematologic malignancy,” “leukemia,” “lymphoma,” “myeloma,” “depression,” “anxiety,” “survival,” and “prognosis”. Additionally, we searched the reference lists of relevant articles to identify any other publications that met our inclusion criteria. A comprehensive search strategy is provided in Supplementary Table 1.

Inclusion and exclusion criteria for eligibility
Studies should assess depression and anxiety using clinical diagnoses based on established criteria such as the International Classification of Diseases (ICD) or the Diagnostic and Statistical Manual of Mental Disorders (DSM), through self-reported instruments like the Hospital Anxiety and Depression Scale (HADS) and Beck Depression Inventory (BDI), or any other recognized diagnostic criteria or validated scales depending on the methodology of each study.
Studies had to examine the association between either depression or anxiety and survival (e.g., OS, EFS, cancer-specific survival [CSS]) among those with hematologic malignancies (AML, ALL, CML, CLL, DLBCL, lymphoma (Hodgkin and non-Hodgkin), or myeloma, MDS). Neither the patients’ age nor the publication date was restricted, allowing for the inclusion of all relevant information studies. Studies including animal models, cell lines, or non-human patients were excluded. We also excluded non-original research, including review articles, book chapters, editorials, and studies published in languages other than English. Attempts were made to contact authors for studies with unavailable full texts, and studies without responses were excluded from further analysis.

Literature screening and data extraction
We performed the screening in two steps. In the first stage, two reviewers independently screened titles and abstracts of retrieved articles using the Rayyan web tool and predetermined eligibility criteria [26]. This preliminary assessment classified articles as “included” or “excluded.” In the second stage, a full-text review of eligible studies based on inclusion and exclusion criteria was conducted by two independent reviewers. Any discrepancies at any stage were resolved through discussion or consulting a third reviewer. Three reviewers extracted data, including bibliographic information (title, authors, year, study design, type of malignancy, age, follow-up period, scoring system, assessment time point (before/after)) and outcome-specific data (depression/anxiety, survival). A fourth reviewer resolved persistent disagreements.

Quality assessment
The Newcastle–Ottawa Scale (NOS) was used to evaluate the methodological quality of included observational studies [27]. The Newcastle–Ottawa Scale (NOS) awards a maximum score of nine, where scores of ≥ 7 represent high-quality studies and < 7 are classified as low quality. This scoring system ensured a rigorous quality assessment to enhance the validity of the systematic review process. We used RoB2 instructions to assess the risk of bias in the included randomized control trials (RCT) [28]. Two assessors independently assessed each study, and disagreements among assessors were resolved through discussion or referral to a third reviewer.
We retained all studies in the primary analysis. To evaluate the impact of study quality, we conducted three additional analyses: (1) we modeled the NOS score as a continuous moderator in a random-effects meta-regression, (2) we performed a subgroup analysis based on the categorical results of the quality assessment (poor vs. good quality), and (3) we also conducted a sensitivity analysis restricted to high-quality studies (NOS ≥ 7).

Data analyses
We performed a meta-analysis of studies exploring the relationship between depression/anxiety and survival in patients with hematologic malignancies. Hazard ratios (HRs) were used as the primary effect size measure to estimate the effect of these mental health conditions on OS, EFS, and CSS. HRs and 95% confidence intervals (CIs) were included in the analysis. For studies with only Kaplan–Meier survival curves, HRs and CIs were calculated from the curves using the Automeris digitizer and Guyot reconstruction method [29, 30]. A random-effects model was used to allow for anticipated heterogeneity in design, patient characteristics, and methodologies between studies. Heterogeneity was evaluated by the I2 statistic, with I2 values ranging from 0 to 25% indicating low heterogeneity, 26–50% indicating moderate heterogeneity, and values above 50% indicating high heterogeneity.
We predefined subgroup analyses based on study-level factors that might influence the association: exposure type (depression, anxiety, combined depression/anxiety), geographic region (Asia, Europe, North America), assessment timing (baseline/pre-treatment vs after treatment), assessment method (clinical diagnosis via ICD/DSM/structured interview vs questionnaire-based instruments), HR source (reported adjusted multivariate analysis, reported unadjusted univariate analysis, or reconstructed from Kaplan–Meier curves), sample size (≤ 200, 200–1,000, 1,000–10,000, > 10,000), and BMT/HSCT cohort (yes vs no). Subgroup meta-analyses employed random-effects models; differences between subgroups were assessed using a Q-test for subgroup differences. Additionally, since AML cohorts were identified separately in the included studies, we conducted a separate analysis to assess the effect size in AML patients.
Sensitivity analysis was performed employing a leave-one-out method, in which each study was omitted one at a time to determine its impact on overall results. Funnel plots and Egger’s test were used to assess publication bias. Statistical analyses were conducted in R software (v4.3.2) and “meta”, “metafor”, “ggplot”, and “tidyverse” packages.

Results

Study identification
A comprehensive search of relevant databases initially identified 8,242 studies, of which 4,931 were excluded due to duplication. After screening titles and abstracts, 3194 studies were excluded due to irrelevant focus or other eligibility criteria. Six studies were inaccessible and therefore excluded. The remaining 111 studies underwent full-text screening, and 29 studies met our criteria [20, 21, 31–57]. These studies assessed the association between depression or anxiety and survival outcomes in patients with hematologic malignancies. The reason for exclusion at each stage is clearly outlined in Fig. 1.

Characteristics of the included studies
The included studies varied in terms of study design, patient populations, and assessment methods for depression and anxiety. In total, 29 studies containing 31 cohorts were included, comprising 419,054 patients [20, 21, 31–57]. The studies covered various hematologic malignancies, including AML, ALL, CML, CLL, MM, DLBCL, NHL, HL, and MDS. The average age of patients in the studies ranged from 34 to 80.4 years. Methods of depression and anxiety assessment differed, with five studies using clinical interviews, including ICD-10 [39, 51, 56], DSM-III [40], and DSM-IV [54], and 24 studies using questionnaires. Follow-up periods ranged from 1 year to 9.5 years, and outcomes such as overall survival (OS), event-free survival (EFS), and cancer-specific survival (CSS) were the primary focus. The main characteristics of the studies are listed in Table 1.

Quality assessment of included studies
The observational studies had a mean score of 7.46 on NOS. Most studies demonstrated strong representativeness of the exposed cohort and adequate ascertainment of exposure. All studies clearly demonstrated that the outcome of interest was not present at the start of the study. However, comparability between cohorts was frequently limited, with many studies not adjusting for the most important confounding factors. Follow-up was generally sufficient, and outcome assessment was consistently adequate across studies. We also converted Newcastle–Ottawa scales to AHRQ standards using thresholds outlined in the Agency for Healthcare Research and Quality (AHRQ) guidelines and assessed the quality of studies as good, fair, or poor. Thirteen studies were considered to be of good quality, and 15 were of poor quality. There were some concerns regarding the quality of the one RCT study. The scores are presented in Table 1, and Supplementary Table 2 demonstrates the results in more detail.

Association between depression/anxiety and survival outcomes
Wherever possible, we used multivariable-adjusted effect sizes reported by the primary studies rather than unadjusted estimates, in order to minimize confounding. However, the covariates included in these models varied considerably across studies (Supplementary Table 3). Most studies adjusted for age, sex, disease stage or risk category, performance status, comorbidities, and treatment type. Fewer studies accounted for socioeconomic indicators (e.g., education, income), lifestyle behaviors (e.g., smoking, alcohol use), or relevant biomarkers.

Overall survival
Patients with HM and depression or anxiety had lower survival rates compared to patients with HM and without depression or anxiety (HR = 1.20; 95% CI: 1.12–1.29; p < 0.0001; I2 = 85%). The forest plot for OS is illustrated in Fig. 2. The subgroup analysis for depression and OS included 28 cohorts of patients from 27 studies and demonstrated a pooled HR of 1.17 (95% CI: 1.09–1.27; p < 0.001). The analysis suggests that depression is significantly associated with a modest increase in the risk of poor overall survival. Anxiety, assessed in 9 studies, did not show a significant association with OS with a pooled HR of 1.20 (95% CI: 1.00–1.44; p = 0.044). Furthermore, some studies reported depression and anxiety together (shown as depression/anxiety in the figures). The pooled HR of these studies was 1.27 (95% CI: 1.14–1.40; p < 0.001). The heterogeneity for OS in each subgroup was high, with I2 values of 86%, 80%, and 59% for depression, anxiety, and depression/anxiety subgroups, respectively.

Event-free survival
The forest plot for EFS, shown in Fig. 3A, included three studies assessing both depression and anxiety. Depression was not significantly associated with EFS, showing a pooled HR of 1.11 (95% CI: 0.67–1.84; p = 0.74). However, anxiety was weakly associated with poorer EFS, with a pooled HR of 1.39 (95% CI: 1.05–1.83; p = 0.025). Although the heterogeneity for anxiety was low, the HRs in depression showed high heterogeneity, with I2 values of 9% and 63%, respectively.

Cancer-specific survival
The pooled HR for depression and CSS, evaluated in two studies, was 1.24 (95% CI: 0.75–2.05; p = 0.39), suggesting no significant association between depression and CSS. Anxiety, assessed in two studies, showed a pooled HR of 1.38 (95% CI: 0.90–2.12; p = 0.15), with no significant effect on CSS. The heterogeneity for CSS was high in both groups (I2 = 90% and 60%, respectively), indicating variability in the studies included (Fig. 3B).

Publication bias
Publication bias was assessed using funnel plots, and visual inspection of the funnel plot for OS (Supplementary Fig. 1) showed some asymmetry, suggesting potential publication bias. Egger’s test for publication bias indicated a significant result (p = 2.8e-07), supporting the possibility of small-study effects influencing the overall findings.
The trim and fill method was applied to adjust for publication bias. After adjusting for missing studies, the results for OS were majorly affected, confirming that the findings were sensitive to potential publication bias. For depression, the overall HR for OS after adding 12 studies with the trim and fill adjustment was 1.01 (95% CI: 0.87–1.17). Furthermore, applying this method for anxiety and OS results showed an HR of 1.0091 (95% CI: 0.81–1.26) after adding four studies, suggesting publication bias significantly impacts the results.

Subgroup analysis and sensitivity analysis
To evaluate the robustness of the association between depression or anxiety and survival outcomes in different groups of patients, we performed several subgroup analyses. We categorized studies based on the continent, HR source, sample size, psychological score, score type, assessment time, hematologic malignancy, and BM transplant. The results are shown in Table 2.
Also, in a subgroup analysis based on the categorical results of the quality assessment (poor vs. good quality), two groups showed no significant difference in OS (P-value = 0.69). Excluding low-quality studies (NOS < 7) yielded k = 23 with HR = 1.19 (95% CI: 1.10–1.30, I2 = 87.8) for OS; conclusions remained unchanged compared to the primary analysis (Supplementary Fig. 2).
To evaluate the robustness of the pooled results, we performed a leave-one-out sensitivity analysis (Supplementary Fig. 3). The results indicated that excluding individual studies did not significantly alter the overall HR for OS, confirming the stability of the findings. The pooled HR for OS remained consistent, and heterogeneity was not notably affected.
We repeated the leave-one-out sensitivity analysis, only on high-quality studies (NOS ≥ 7), which showed that removing individual studies does not significantly alter the overall pooled HR for OS (Supplementary Fig. 4). The point estimates for HR stay consistent across the different exclusions, indicating that the analysis results are robust and suggesting that using only high-quality studies for the analysis provides a reliable estimate of the link between depression/anxiety and OS, with no single study disproportionately affecting the pooled estimate.

Meta-regression
Meta-regression analyses were performed to explore the impact of various factors on the observed associations. Specifically, we examined sample size, country, NOS score, and gender as potential moderators. Our results showed a significant relationship between sample size and effect size for OS (P-value = 0.0006), indicating that sample size substantially affects the overall results. However, we found no significant effect of sex, publication year, and quality assessment of papers (in NOS format) on the association between depression/anxiety and OS (P-value = 0.877, 0.148, and 0.113, respectively) (Supplementary Fig. 5–8).

Discussion
In this meta-analysis, we evaluated the impact of anxiety and depression on survival outcomes in patients with hematologic malignancies. The results indicated that depression was associated with worse OS, while anxiety did not show a significant association, highlighting their varying effects. Subgroup analyses further suggested that these associations were influenced by geographic region, assessment tools, and sample size, underscoring heterogeneity across studies. Anxiety was also linked to poor EFS; however, we found no evidence for the association of EFS with depression. No significant association between depression or anxiety and cancer-specific survival (CSS) was found. Our results should be interpreted carefully, as there was high heterogeneity across studies and significant publication bias.
A recent meta-analysis including 8 studies showed that depression in patients diagnosed with cancer and undergoing HSCT worsened the OS significantly with an HR of 1.07 (95% CI 1.03–1.11), which is in line with our findings [58]. Another 2019 pan-cancer meta-analysis of 51 cohort studies by Wang et al. showed that depression and anxiety increased the cancer-specific mortality (HR = 1.66; 95% CI 1.43–1.93) in hematopoietic cancers but did not affect their incidence (HR = 1.15; 95% CI 1.00–1.33) [18]. Previous meta-analyses have similarly demonstrated the detrimental effects of depression and anxiety on survival in various other cancers, including breast, glioma, prostate, and gastrointestinal tumors [18, 59–63].
In line with our findings, previous studies have shown a more significant association between depression and cancer survival compared to anxiety [61–64]. One study analyzed data from 19,966 patients with common cancers and revealed that depression is strongly associated with worse survival across all cancer types, while anxiety's impact varies and can even correlate with better survival in females when adjusted for depression [63]. A recent meta-analysis on the effects of anxiety and depression on colorectal cancer outcomes found that, unlike anxiety, depression was associated with all-cause mortality rates of 1.89 (95% CI 1.68–2.13) [61]. These findings show that anxiety and depression may also adversely affect survival outcomes through distinct mechanisms. Anxiety is often associated with heightened arousal, worry, and physiological stress responses, while depression involves low mood, reduced energy, and anhedonia. Depression’s more pronounced impact on survival could also stem from its stronger association with committing suicide, while anxiety may even lead to healthier behaviors, more medical care seeking, and better adherence to medical treatments [63, 65].
The inconsistent findings for anxiety, showing no significant association with OS but a weak association with EFS, may stem from the nature and sensitivity of these endpoints. EFS reflects earlier disease events, such as relapse, progression, or treatment failure, which may be more immediately affected by anxiety through mechanisms such as heightened physiological stress responses, less effective coping strategies, or delays in initiating or completing treatment. In contrast, OS captures all causes of death over a longer period, during which the initial impact of anxiety may lessen due to effective symptom management, psychological adaptation, or competing mortality risks. Also, the smaller number of studies examining the relationship between anxiety and OS may reduce statistical power, contributing to the absence of a significant association. The absence of a significant association between either depression or anxiety and cancer-specific survival may result from the small number of included studies in the CSS analyses and the high between-study heterogeneity, reducing statistical power. Moreover, CSS does not capture mortality through non-cancer pathways, including cardiovascular disease, infections, and suicide, which may be contributors to mortality in patients with depression and anxiety [66].
Our subgroup analysis showed a more prominent HR for Asian populations, consistent with the findings of the pan-cancer meta-analysis by Wang et al. that showed a greater effect size for the impact of depression and anxiety on all-cause mortality among cancer patients in the Asian subgroup [18]. This may be attributed to a combination of delayed diagnosis and treatment of mental illnesses due to cultural stigma, limiting access to care due to socioeconomic disparities, and genetic factors [67]. Furthermore, the association of depression and anxiety and OS was not significant in the subgroup of studies with HRs extracted from univariate analysis. This pattern suggests the presence of negative confounding, where certain variables attenuate the crude association but, when controlled for, reveal a stronger link between psychological symptoms and survival. Potential negative confounders could include treatment type or intensity, age, and performance status, as these factors may be associated with both psychological distress and better survival outcomes, thereby masking the effect in unadjusted analyses.
The subgroup analysis of the studies that used clinical assessment tools such as ICD-10 or DSM criteria showed a significantly greater association between depression diagnosis and reduced OS compared to those using self-report questionnaires such as the HADS score. This finding is in line with previous systematic reviews on other cancers and may result from self-report tools capturing milder symptoms that may not considerably affect OS [18, 68]. Additionally, this finding may show a dose–response relationship, as clinical interviews generally apply standardized diagnostic thresholds and capture more severe, functionally impairing cases, whereas self-report tools may detect milder or transient symptoms, increasing the risk of non-differential misclassification and underestimating true associations. Variation in assessment methods likely contributed to heterogeneity in our pooled estimates. Standardizing case definitions in future research, ideally using structured clinical interviews conducted by trained professionals, would improve comparability between studies and enhance the precision of effect estimates.
We did not find a significant difference in depression’s association with OS between studies assessing depression before versus after cancer diagnosis, which aligns with previous meta-analyses in breast cancer and various other cancer types [18, 19]. However, a colorectal cancer meta-analysis showed lower survival in patients with depression diagnosed after cancer diagnosis (HR = 1.35 (95% CI 0.99–1.84) for before vs. 1.79 (95% CI 1.30–2.46) for after) [61]. They attributed their findings to an earlier diagnosis of depression being associated with better management and control, leading to better depression outcomes and cancer survival. The inconsistent results could be attributed to varying biological mechanisms that influence the interaction between depression and survival, as well as varying amounts of psychological distress caused by different cancer treatments. We found a lower HR for the impact of depression on OS in post-bone marrow transplant (BMT) patients, likely due to the intensive monitoring and psychological support they receive. Survivor bias may also play a role, as BMT recipients tend to be healthier with better prognoses, which reduces the observed effect of depression on survival.
Our meta-regression results showed that sex and age did not significantly affect the association between depression and anxiety and OS, which is in line with previous studies not showing a difference between men and women [18, 19] or different ages [18]. However, some previous meta-analyses found a larger effect size in either the younger [61, 68] or older [19] subgroups. Some studies show a negative correlation between age and depression severity, which could support the studies that show a poorer prognosis for young patients [69, 70]. However, the effects may vary by type of cancer and their different mechanisms of mortality.
Depression and anxiety can worsen survival outcomes in patients with HM through several interconnected mechanisms. Unhealthy behaviors, such as sedentary lifestyle, smoking, alcohol consumption, poor nutrition, and obesity, are more common among people who suffer from depression and anxiety [71]. Additionally, low adherence to treatment and follow-up appointments is common in patients with depression and can lead to a poor prognosis in patients with cancer [72]. One study showed that patients who experienced depression or anxiety before pancreatic cancer diagnosis had a reduced likelihood of receiving chemotherapy (OR = 0.58, p = 0.04) and a decrease in overall survival (HR = 1.32, p = 0.04) [60]. Prolonged psychological distress can also cause systemic inflammation, marked by high levels of pro-inflammatory cytokines like TNF-α and IL-6, which induce angiogenesis, tumor growth, and metastasis [73]. The hypothalamic–pituitary–adrenal (HPA) axis is frequently dysregulated in people with depression, leading to aberrant cortisol release [73, 74]. This exacerbates inflammation, inhibits immunological surveillance, and encourages the survival and growth of cancer cells [73]. Also, sympathetic nervous system activation brought on by stress increases the expression of enzymes such as matrix metalloproteinases, which break down the extracellular matrix and promote the development of tumors [73]. Moreover, patients with depression have decreased activity of DNA repair enzymes and natural killer cells, which are crucial for cancer defense [74].
Our findings underscore the importance of routine screening and an integrated and multidisciplinary management of depression and anxiety in patients with HM. Various interventions have been suggested to help improve anxiety and depression in patients with cancer such as psychological, educational, and psychosocial interventions [75]. A systematic review of 29 RCTs showed that physical, psychological, complementary, and spiritual therapies can improve the quality of life and anxiety and depression symptoms in people with blood cancers [76]. Similar benefits were observed in various other cancer types [77, 78]. One study showed that mental health treatment programs were associated with lower all-cause and cancer-specific mortality in veterans with a diagnosis of non-small cell lung cancer with pre-existing mental health disorders [79]. Regarding pharmacologic management and using antidepressants, the findings are less consistent, with some studies even showing increased mortality in patients with breast cancer using SSRIs [68]. However, pharmacotherapy can still be helpful in patients who prefer it or don’t have access to first-line treatment such as cognitive behavior therapy, behavioral activation, mindfulness-based stress reduction, physical activity, or psychosocial interventions [75]. Although survival benefits are unclear, improving depression and anxiety symptoms significantly enhances quality of life and has important public health implications [80, 81].
In this study, we conducted a comprehensive literature review and applied advanced meta-analysis methods such as meta-regression and detailed sensitivity analyses to confirm the robustness of our findings. However, our study has several limitations. First, we found significant heterogeneity across studies, which may have resulted from the variability in assessment tools for depression and anxiety, follow-up durations, treatment protocols, and patient populations. Publication bias was also present and had majorly affected our results, as revealed by the trim and fill method. The high heterogeneity across studies and the presence of significant publication bias substantially limits the interpretability and clinical applicability of our findings. Although our use of random-effects models, subgroup analyses, and sensitivity analyses aimed to account for these issues, the trim-and-fill adjustment for publication bias reduced the pooled HR for depression and OS from 1.17 to 1.01, effectively nullifying the association. While it should be taken into consideration that Egger’s test only shows small-study effects, and the trim and fill method has decreased reliability when between-study heterogeneity is large, the apparent link between depression and OS should be interpreted cautiously, and our results should be viewed as hypothesis-generating rather than conclusive. Clinically, this means that while screening and management of depression remain important for quality of life and potential survival benefit, our meta-analysis alone cannot establish a causal or uniformly applicable effect on survival without confirmation from large, standardized, and multicenter studies. Furthermore, inconsistent adjustments among the included studies make identifying a standardized association based on adjustments for the same variables impossible. Failure to consistently adjust for potential confounders such as disease stage, performance status, treatment type, comorbidity indices, lifestyle behaviors, and inflammatory biomarkers may bias results and overestimate or underestimate the association between psychological symptoms and survival. Because we relied on published HRs, we could only use the adjustments reported by each study. Multivariable HRs were sometimes missing because they were not measured, not reported, or excluded from the final model.
Our results advocate for three immediate clinical actions: (1) routine integration of validated tools like the HADS during initial oncology assessments, (2) referral of high-risk patients (e.g., pre-transplant or relapsed cases) to mental health specialists, and (3) trialing multidisciplinary interventions, such as cognitive-behavioral therapy or anti-inflammatory regimens, to mitigate depression-driven mortality. Future studies should test whether such interventions improve survival, particularly in high-heterogeneity subgroups like older AML patients.

Conclusion
Depression is significantly associated with reduced overall survival in patients with hematologic malignancies, while anxiety showed no statistically significant association with overall survival but was linked to poorer event-free survival. These effects may arise through distinct mechanisms: depression may influence outcomes through systemic inflammation, hypothalamic–pituitary–adrenal axis dysregulation, and reduced treatment adherence, while anxiety may act through heightened stress responses and behavioral changes. Given the high prevalence of depression and anxiety among cancer patients, integrating mental health care into the management of patients with HM is crucial to improving both quality of life and clinical outcomes. Further studies are needed to clarify the mechanisms and causal pathways linking depression and anxiety with cancer outcomes. Prospective cohort studies with standardized psychological assessments, as well as interventional trials in psycho-oncology, are warranted. Research should also explore the impact of tailored mental health interventions on hematologic malignancy outcomes and the role of socioeconomic and financial factors.

Supplementary Information