Prevalence of anxiety and depression in idiopathic pulmonary fibrosis: a systematic review and meta-analysis Authors: Lili Xu, Danyang Zang, Ruonan Yan, Zhihui Liang, Shuguang Yang, Ningxia Yu, Xueqing Yu Published: 2026 ABSTRACT AbstractBackgroundIdiopathic pulmonary fibrosis (IPF) is a chronic, progressive, fibrosing interstitial lung disease. Anxiety and depression are prevalent psychiatric comorbidities among IPF patients; however, reported prevalence rates vary substantially, and a consolidated estimate is lacking. This systematic review and meta-analysis aimed to determine the global pooled prevalence of anxiety and depression in IPF patients and to explore potential sources of heterogeneity, thus providing a robust evidence base for clinical management and future research.MethodsThis study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. A systematic literature search was conducted in PubMed, Web of Science, Cochrane Library, and Embase from database inception to October 23, 2025, to identify studies reporting the prevalence of anxiety or depression in patients with IPF. Two reviewers independently screened the literature, extracted data, and assessed the methodological quality of the included studies using the Joanna Briggs Institute (JBI) critical appraisal checklist. Two separate single-arm meta-analyses were performed, one for studies using International Classification of Diseases, Tenth Revision (ICD-10) criteria and one for those using screening tools (e.g., Hospital Anxiety and Depression Scale, Patient Health Questionnaire-9). Pooled prevalence estimates were calculated using a generalized linear mixed model (GLMM), with corresponding 95% confidence intervals (CIs) and 95% prediction intervals (PIs). Heterogeneity was assessed using Cochran’s Q test, the I² statistic, and τ². Sensitivity analyses were conducted using the leave-one-out method and the Freeman-Tukey double arcsine transformation. Subgroup analyses were performed based on study design, geographic region, assessment tool, methodological quality, and sample size. Publication bias was evaluated using Doi plots and the LFK index.ResultsA total of 25 studies involving 78,060 IPF patients were included. Among these, 15 studies reported the prevalence of anxiety and 25 reported the prevalence of depression. Studies that did not specify the assessment tool were excluded from the meta-analysis, resulting in the inclusion of 14 studies for the anxiety meta-analysis and 18 studies for the depression meta-analysis. Quantitative analyses were performed separately for the ICD-10 group and the screening tool group.Anxiety in IPF: The pooled prevalence was 14.52% (95% CI: 8.90%–22.79%) in the ICD-10 group (2 studies), with substantial heterogeneity (τ² = 0.1599, I² = 98.9%, P < 0.001); the 95% PI was not calculated due to the small number of studies. In the screening tool group (12 studies), the pooled prevalence was 42.04% (95% CI: 25.28%–60.86%), with a 95% PI of 3.36%–93.80% and substantial heterogeneity (τ² = 1.7068, I² = 88.3%, P < 0.001).Depression in IPF: The pooled prevalence was 6.87% (95% CI: 3.23%–14.04%) in the ICD-10 group (3 studies), with a 95% PI of 0.00%–99.95% and substantial heterogeneity (τ² = 0.4853, I² = 99.8%, P < 0.001). In the screening tool group (15 studies), the pooled prevalence was 38.26% (95% CI: 24.65%–54.01%), with a 95% PI of 3.89%–90.46% and substantial heterogeneity (τ² = 1.4877, I² = 82.3%, P < 0.001).Subgroup analyses based on the screening tool group showed no statistically significant differences in pooled prevalence across study design, assessment tool, geographic region, methodological quality, or sample size (all P > 0.05). Sensitivity analyses confirmed the robustness of the pooled estimates for both anxiety and depression. Publication bias assessment revealed significant bias in the screening tool groups for both outcomes.ConclusionThe prevalence of anxiety and depression in IPF patients is substantial and is significantly influenced by the assessment tool used: estimates based on ICD-10 clinical diagnoses are substantially lower than those derived from screening tools. Therefore, future large-scale, high-quality studies employing standardized assessment tools are needed to accurately evaluate the psychiatric comorbidity burden in IPF patients, thereby providing robust evidence for the development of targeted clinical intervention strategies.Trial registrationPROSPERO; No.: CRD420251174885.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12890-026-04400-1.Keywords: Idiopathic pulmonary fibrosis, Anxiety, Depression, Systematic review, Meta-analysis Introduction Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive, fibrosing interstitial lung disease of unknown etiology, characterized by progressive fibrosis of the lung parenchyma and destruction of normal alveolar architecture, ultimately leading to impaired gas exchange [1]. IPF predominantly affects older men, with progressive dyspnea and irreversible decline in lung function being the hallmark clinical features. The median survival after diagnosis is only 2–3 years, reflecting its dismal prognosis and substantial public health burden [1]. The burden of IPF, however, extends far beyond the respiratory system. Persistent severe dyspnea and refractory dry cough, combined with the risk of rapid disease progression, limited treatment options, and significant socioeconomic and psychological distress, collectively elevate the risk of psychiatric comorbidities in this patient population [2–4]. Among these, anxiety and depression represent the most prevalent mental health disorders. These conditions can exacerbate respiratory symptoms such as dyspnea and cough, significantly impair health-related quality of life, compromise treatment adherence, thereby contributing to poorer clinical outcomes [5, 6]. Clinically, anxiety in IPF patients often manifests as excessive worry or fear regarding dyspnea episodes or disease prognosis, whereas depression is typically characterized by persistent low mood, anhedonia, and fatigue [7, 8]. In recent years, a growing body of research has investigated the prevalence of anxiety and depression in IPF patients, providing important insights into their mental health burden. However, reported prevalence rates vary substantially across studies, ranging from relatively low percentages to rates exceeding 50% [9–13]. This heterogeneity likely stems from methodological variations in assessment tools, study design, population characteristics and sample size, making it difficult to derive reliable epidemiological estimates from individual studies. Consequently, this inconsistency has hindered the development of evidence-based clinical guidelines and systematic intervention strategies. To address these evidence gaps, this systematic review and meta-analysis aims to synthesize the existing literature, provide a more precise pooled estimate of the prevalence of anxiety and depression in IPF patients globally, and systematically explore potential sources of heterogeneity. The findings of this synthesis are expected to provide high-level evidence for accurately assessing the burden of psychiatric comorbidities in IPF, thereby informing the development of integrated, patient-centered management strategies aimed at improving overall health-related quality of life and clinical outcomes. Methods The study protocol was registered in advance on the International Prospective Register of Systematic Reviews (PROSPERO) under registration number CRD420251174885. Search strategy This systematic review and meta-analysis was conducted in accordance with the PRISMA statement [14]. The search strategy was developed following the PICOS framework: Population: patients with IPF; Outcome: prevalence of anxiety or depression; Study design: observational studies (no restriction on specific design subtypes). A comprehensive literature search was performed using a combination of subject headings and free-text terms to identify relevant studies published from database inception to October 23, 2025, in the following electronic databases: PubMed, Web of Science, Cochrane Library, and Embase. Search terms included “idiopathic pulmonary fibrosis”, “anxiety”, “anxious”, “depression”, “depressive”, among other related terms. The complete search strategy for each database is provided in Supplementary Material 1. Furthermore, we manually screened the reference lists of included articles and relevant reviews to identify additional eligible studies. Selection criteria Studies were included if they met all of the following criteria: (1) population: patients diagnosed with IPF; (2) study type: cross-sectional studies or baseline prevalence data from cohort studies; (3) outcome: reported prevalence of anxiety or depression among IPF patients. Studies were excluded if they were: (1) duplicate publications; (2) conference abstracts, meta-analyses, case series, or other ineligible study types; (3) studies for which the full text was unavailable; (4) studies from which relevant data could not be extracted. Study selection and data extraction Two reviewers independently conducted literature screening and data extraction, with cross-verification performed at each stage. Any discrepancies were resolved through discussion with a third reviewer. The retrieved records were first imported into EndNote 21.5 for duplicate removal. Titles and abstracts were then screened to exclude studies clearly not meeting the eligibility criteria. For records where relevance could not be determined from the abstract, the full text was reviewed. The following data were extracted from the included studies: first author, publication year, country, study design, study setting, funding source, sample size, participant gender and age, number of anxiety or depression cases, prevalence rates, and the assessment tool used. Quality assessment Two researchers independently appraised the methodological quality of the included studies, with cross-verification of their assessments. Any disagreements were resolved through discussion with a third reviewer. The Joanna Briggs Institute (JBI) critical appraisal checklist for prevalence studies was used for quality evaluation. The checklist consists of nine items, each rated as “Yes,” “No,” “Unclear,” or “Not applicable.” Studies meeting most criteria (≥ 7 “Yes” responses) were rated as having low risk of bias; those meeting some criteria (4–6 “Yes” responses) were rated as moderate risk; and those with ≤ 3 “Yes” responses were rated as high risk of bias. A higher number of “Yes” responses indicated higher study quality [15]. Statistical analysis To investigate the influence of different diagnostic criteria on prevalence estimates, two separate meta-analyses were performed based on ICD-10 criteria and screening tools, e.g., the Hospital Anxiety and Depression Scale (HADS) and the Patient Health Questionnaire-9 (PHQ-9). Studies that did not report the assessment tool were only described narratively and were not included in the meta-analysis. The same statistical methods were applied to both analyses, and the results were reported separately. All analyses were conducted using R software (version 4.3.3) with the meta package. Forest plots were generated using the RevMan layout. The significance level was set at α = 0.05 (two-tailed). A generalized linear mixed model (GLMM) was used for the single-arm meta-analysis. The GLMM directly models the original binomial data without requiring continuity correction for zero-event studies, thereby providing more accurate estimates of within- and between-study variance [16]. Parameters were estimated using maximum likelihood (ML), and the outcome measure was the logit‑transformed prevalence. The 95% confidence intervals (CIs) for the pooled prevalence were calculated on the logit scale based on the normal approximation (conditional approach) and then back-transformed to the original probability scale using the logistic transformation. When at least three studies were available, the 95% prediction interval (PIs) was calculated on the logit scale as: , where is the logit of the pooled effect, is the between-study variance, is the standard error of the pooled logit, and is the 0.025 quantile of the t-distribution with k-2 degrees of freedom. The resulting limits were then back-transformed to the prevalence scale [17]. Heterogeneity was quantified by estimating τ² using the restricted maximum likelihood (REML) method, and the following measures were reported: Cochran’s Q test (with α = 0.10 and P < 0.10 indicating significant heterogeneity), and the I² statistic (with 0%–25% considered low, 25%–75% moderate, and ≥ 75% substawith α = 0.10, P < 0.10 indicating significant heterogeneityntial heterogeneity) [18, 19]. To assess the robustness of the primary findings, the following two sensitivity analyses were performed [20]: (1) Leave-one-out analysis: each study was sequentially removed, and the GLMM analysis was repeated on the remaining studies to evaluate the influence of individual studies; the results were presented as a leave-one-out forest plot. (2) Alternative effect measure transformation: the Freeman-Tukey double arcsine transformation (PFT) with a random-effects model was used to re-pool the data, and the 95% PI was calculated. Heterogeneity was also assessed, reporting τ², Cochran’s Q test, and the I² statistic. To explore potential sources of heterogeneity, subgroup analyses were performed based on study design, assessment tool, geographic region, methodological quality (score 4–6 vs. ≥ 7), and sample size (dichotomised into large vs. small using the median as the cut-off). All subgroup analyses were conducted using a random-effects model. Publication bias was evaluated using Doi plots and the LFK index, which are more reliable than funnel plots or Egger’s test [21]. The LFK index was interpreted as follows: |LFK| < 1 indicates no significant asymmetry, 1 ≤ |LFK| ≤ 2 indicates minor asymmetry, and |LFK| > 2 indicates significant asymmetry [22]. This assessment was performed only for subgroups with at least three studies. If significant publication bias is detected, the trim-and-fill method is employed as an exploratory sensitivity analysis. Results Study selection The systematic search yielded 1,425 records, comprising 171 from PubMed, 219 from Web of Science, 948 from Embase, and 87 from Cochrane Library. An additional 13 studies were identified through manual searches of reference lists, resulting in 1,438 records in total. After removing 355 duplicates, 1,083 records remained. Screening of titles and abstracts led to the exclusion of 993 records. Subsequently, 55 articles were excluded as their full texts were unavailable. After full-text assessment, 10 studies were excluded for insufficient data, resulting in the final inclusion of 25 studies. The detailed study selection process is presented in Fig. 1. Flow diagram of the study selection process Chateristics of included studies Among the 25 studies included in this meta-analysis [9–13, 23–42], the sources of prevalence data can be classified into three categories: Five studies [27, 32, 34, 35, 40] were based on administrative claims databases, 19 studies [9–13, 23–26, 28–31, 33, 36, 38, 39, 41, 42] were based on hospital-based clinical cohorts, and one study [37] was based a community-based cohort. 15 studies reported the prevalence of anxiety in IPF patients [9, 11–13, 24–29, 31, 35, 39, 41, 42], while all 25 studies reported the prevalence of depression. Detailed characteristics of the included studies are presented in Table 1. Table 1The characteristics of the people the included studiesStudy (year)regionStudy typeStudy CenterStudy periodFundNo.of patientsGender(male/female)Mean(SE) or Median(IQR) of ageMental Health Outcomes AssessedPrevalencen(%)Assessment toolsAnxietyDepressionAnxietyDepressionAguilar-Duran H 2025 [9]South Americacross-sectional studymulticenter2021–2022No7561/1474 ± 7Both21(28%)26(35%)HADSHADSAkhtar AA 2013 [10]Europecross-sectional studysingle-center2010–2011No11860/5868.1 ± 11.35Depression/58(49.2%)/WDILee YJ 2017 [11]AsiaProspective cohortmulticenter2012–2015Yes11275/3772Both29(25.9%)24(21.4%)HADSHADSMatsuda T 2017 [12]Asiaretrospective cohort studysingle-center2009–2013No12199/2266.8 ± 7.6Both26(21.49%)27(22.31%)HADSHADSX. He 2024 [13]AsiaProspective cohort studysingle-center2021–2022Yes5645/1167(58–72)Both42(75%)35(62.5%)GADPHQ-9A. Tzouvelekis 2020 [23]EuropeProspective cohort studymulticenter2016–2017/9880/2070.8 ± 8.1Depression/42(42.9%)/BDI-IIYalnız E 2019 [24]Europecross-sectional studysingle-center2016–2017/5038/1267.4 ± 7.1Both49(98%)50(100%)HADSHADSHolland AE 2014 [25]Oceaniacross-sectional studysingle-center2010–2013/48//Both13(27.08%)14(29.17%)HADSHADSBloem AEM 2020 [26]Europecross-sectional studysingle-center2018–2019No5945/1473(70–78)Both5(9.1%)5(8.9%)HADSHADSLee JH 2023 [27]Asiaretrospective cohort studymulticenter2011–2019Yes21,11115,029/6,082/Both4,238(20.07%)3,142(14.88%)ICD-10ICD-10Edwards GD 2022 [28]EuropeProspective cohort studysingle-center2012–2017/235163/7273 ± 9Both81(34.47%)86(36.6%)HADSHADSKette S 2025 [29]Europeretrospective cohort studysingle-center2011–2024No10162/3973(11)Both0(0%)1(1%)//Lee CT 2024 [30]North Americaretrospective cohort studysingle-center/Yes389313/7670.6 ± 7.9Depression/50(12.9%)//Glaspole IN 2017 [31]Oceaniaretrospective cohort studymulticenter2012–2014Yes516347/16971.3 ± 8.6Both156(38.2%)146(35%)HADSHADSXu H 2024 [32]North Americaretrospective cohort studymulticenter2014–2022Yes10,2896,077/4212/Depression/2,088(20.3%)//Ortiz ADR 2024 [33]EuropeProspective cohort studymulticenter2015–2021/232186/4669.9 ± 7.9Depression/18(7.8%)//Dempsey TM 2021 [34]North Americaretrospective cohort studymulticenter2014–2019No10,9966,073/492373.4 ± 9.6Depression/1,564(14.2%)//Liao KM 2024 [35]Asiaretrospective cohort studymulticenter2017–2019Yes1,6801,278/40274.1 ± 9.9Both171(10.2%)51(3%)ICD-10ICD-10Hyldgaard C 2014 [36]Europeretrospective cohort studysingle-center2003–2009/12193/2867.4 ± 8.4Depression/8(7%)//Fernández Pérez ER 2009 [37]North Americaretrospective cohort studymulticenter1997–2005Yes4728/1973.5 ± 7.8Depression/5(11%)//De Vries J 2001 [38]EuropeProspective cohort studymulticenter//4115/2663.5(38–80)Depression/8(23.5%)/BDIKatie Pink 2014 [39]Europeretrospective cohort studysingle-center2010–2012/3724/13/Both15(40.5%)9(24.3%)HADSHADSSho Takuma 2025 [40]Asiaretrospective cohort studymulticenter2013–2018Yes31,38622,891/8,495/Depression/2,100(6.7%)/ICDYenibertiz D 2020 [41]Europeretrospective cohort studysingle-center//2725/165.6 ± 7.6Both18(66.7%)17(63%)HADSHADSCoskun F 2021 [42]Europecross-sectional study studymulticenter2020–2020No11582/3368.43 ± 7.44Both44(43.56%)31(30.69%)HADSHADS The characteristics of the people the included studies Of these, the 15 studies that assessed the prevalence of anxiety were conducted between 2014 and 2025. Five studies were cross-sectional in design [9, 24–26, 42], while 10 studies were cohort studies [11–13, 27–29, 31, 35, 39, 41]. Six studies were multicenter [9, 11, 27, 31, 35, 42], and nine studies were single-center [12, 13, 24–26, 28, 29, 39, 41]. Sample sizes ranged from 27 to 21,111. Geographically, the studies were conducted across diverse populations: seven from Europe [24, 26, 28, 29, 39, 41, 42], five from Asia [11–13, 27, 35], one from South America [9], and two from Oceania [25, 31]. Regarding funding, five studies [11, 13, 27, 31, 35] reported receiving funding, five [9, 12, 26, 29, 42] reported no funding, and five [24, 25, 28, 39, 41] did not mention funding status. Anxiety was primarily assessed using the HADS in 11 studies [9, 11, 12, 24–26, 28, 31, 39, 41, 42], two studies [27, 35] employed the ICD-10 codes, one study [13] used the Generalized Anxiety Disorder (GAD) scale, and one study [29] did not specify the assessment tool used. The 25 studies assessing the prevalence of depression in IPF were published between 2001 and 2025. Six studies [9, 10, 24–26, 42] were cross-sectional, and 19 studies were cohort studies [11–13, 23, 27–41]. Thirteen studies were multicenter [9, 11, 23, 27, 31–35, 37, 38, 40, 42], while 12 studies were single-center [10, 12, 13, 24–26, 28–30, 36, 39, 41]. Sample sizes ranged from 27 to 31,386. Geographically, the studies included 12 from Europe [10, 23, 24, 26, 28, 29, 33, 36, 38, 39, 41, 42], six from Asia [11–13, 27, 35, 40], four from North America [30, 32, 34], one from South America [9], and two from Oceania [25, 31]. Regarding funding, nine studies reported receiving support [11, 13, 27, 30–32, 35, 37, 40], seven reported no funding [9, 10, 12, 26, 29, 34, 42], and nine did not report their funding status [23–25, 28, 33, 36, 38, 39, 41]. Depression assessment tools varied: the HADS was used in 11 studies [9, 11, 12, 24–26, 28, 31, 39, 41, 42], ICD-10 codes were used in three studies [27, 35, 40], and one study each used the Well-being Index (WDI) [10], Patient Health Questionnaire-9 (PHQ-9) [13], Beck Depression Inventory (BDI) [38], and Beck Depression Inventory-II (BDI-II) [23]. Seven studies did not specify the depression assessment tool [29, 30, 32–34, 36, 37]. Study quality appraisal Based on the quality assessment of the 25 included studies, 10 studies were rated as having a low risk of bias [12, 25, 27, 28, 31, 32, 34, 35, 40, 42], while 15 studies were rated as having a moderate risk of bias [9–11, 13, 23, 24, 26, 29, 30, 33, 36–39, 41]. 17 studies [9–13, 23–26, 28–30, 33, 36, 38, 39, 41] did not adopt a population-based sampling framework, 12 studies [9–11, 13, 23, 24, 26, 30, 33, 36, 38, 41] did not clearly report the sampling method, 14 studies [9–13, 23–26, 29, 36, 37, 39, 41] did not provide the basis for sample size calculation, and 10 studies [27, 29–37] did not explicitly state the validation status of the measurement tools used. This distribution indicates generally favorable methodological quality across the included body of literature. Detailed quality assessment results are summarized in Table 2. Table 2Quality assessment of prevalence studies using JBI critical appraisal checklistStudy (year)Q1Q2Q3Q4Q5Q6Q7Q8Q9Total score regarding study risk of biasRisk of biasAguilar-Duran H 2025 [9]NUUYYYYYY6MAkhtar AA 2013 [10]NUUYYYYYY6MLee YJ 2017 [11]NUUYYYYYY6MMatsuda T 2017 [12]NYUYYYYYY7LX. He 2024 [13]NUUYYYYYY6MA. Tzouvelekis 2020 [23]NUUYYYYYY6MYalnız E 2019 [24]NUUYYYYYY6MHolland AE 2014 [25]NYUYYYYYY7LBloem AEM 2020 [26]NUUYYYYYY6MLee JH 2023 [27]YYYYYUYYY8LEdwards GD 2022 [28]NYYYYYYYY8LKette S 2025 [29]NYUYYUUYY5MLee CT 2024 [30]NUYYYUYYY6MGlaspole IN 2017 [31]YYYYYUYYY8LXu H 2024 [32]YYYYYUYYY8LOrtiz ADR 2024 [33]NUYYYUUYY5MDempsey TM 2021 [34]YYYYYUYYY8LLiao KM 2024 [35]YYYYYUYYY8LHyldgaard C 2014 [36]NUUYYUYYY5MFernández Pérez ER 2009 [37]YYUYYUUYY6MDe Vries J 2001 [38]NUYYNYYYN5MKatie Pink 2014 [39]NYUYNYYYN5MSho Takuma 2025 [40]YYYYYYYYY9LYenibertiz D 2020 [41]NUNYUYYYU4MCoskun F 2021 [42]YYYYYYYYY9LList of the nine questions (Q1–Q9) applied to the studies: Q1: Was the sample frame appropriate to address the target population? Q2: Were study participants sampled in an appropriate way? Q3: Was the sample size adequate? Q4: Were the participants and the setting described in detail? Q5: Was the data analysis conducted with sufficient coverage of the identified sample? Q6: Were valid methods used for the identification of the condition? Q7: Was the condition measured in a standard, reliable way for all participants? Q8: Was there appropriate statistical analysis? Q9: Was the response rate adequate, and if no, was the low response rate managed appropriately?To calculate the overall score for the risk of bias in the studies, the scores from individual studies were summed. The JBI checklist consists of 9 items, with responses to each item categorized as “Yes,” “No,” “Unclear,” or “Not Applicable.” A response of “Y” corresponds to a score of “1,” while responses of “N,” “U,” or “NA” correspond to a score of “0.” Studies that met most of the criteria (7 or more “Y” responses) were judged as having a low risk of bias; those that met some criteria (4–6 “Y” responses) were judged as having a moderate risk; and studies with fewer than 3 “Y” responses were judged as having a high risk of bias Quality assessment of prevalence studies using JBI critical appraisal checklist List of the nine questions (Q1–Q9) applied to the studies: Q1: Was the sample frame appropriate to address the target population? Q2: Were study participants sampled in an appropriate way? Q3: Was the sample size adequate? Q4: Were the participants and the setting described in detail? Q5: Was the data analysis conducted with sufficient coverage of the identified sample? Q6: Were valid methods used for the identification of the condition? Q7: Was the condition measured in a standard, reliable way for all participants? Q8: Was there appropriate statistical analysis? Q9: Was the response rate adequate, and if no, was the low response rate managed appropriately? To calculate the overall score for the risk of bias in the studies, the scores from individual studies were summed. The JBI checklist consists of 9 items, with responses to each item categorized as “Yes,” “No,” “Unclear,” or “Not Applicable.” A response of “Y” corresponds to a score of “1,” while responses of “N,” “U,” or “NA” correspond to a score of “0.” Studies that met most of the criteria (7 or more “Y” responses) were judged as having a low risk of bias; those that met some criteria (4–6 “Y” responses) were judged as having a moderate risk; and studies with fewer than 3 “Y” responses were judged as having a high risk of bias The prevalence of anxiety and depression in IPF patients To investigate the influence of different diagnostic criteria on prevalence estimates, two separate meta-analyses of anxiety or depression in IPF patients were performed based on International ICD-10 criteria and screening tools (e.g., HADS, PHQ-9). A total of 15 studies reported the prevalence of anxiety in IPF. Among them, two studies based on ICD-10 criteria [27, 35] yielded a pooled anxiety prevalence of 14.52% (95% CI: 8.90%–22.79%), with substantial heterogeneity (τ² = 0.1599, I² = 98.9%, P < 0.001) (Fig. 2). The 95% PI was not calculated because fewer than three studies were available in this subgroup. The remaining 12 studies based on screening tools [9, 11–13, 24–26, 28, 31, 39, 41, 42] showed a pooled anxiety prevalence of 42.04% (95% CI: 25.28%–60.86%), with a 95% PI of 3.36%–93.80% and substantial heterogeneity (τ² = 1.7068, I² = 88.3%, P < 0.001) (Fig. 3). One additional study that did not report the assessment tool could not be included in the meta-analysis; it reported a prevalence of 0% for anxiety among 101 IPF patients. Forest plot showing the pooled prevalence of anxiety in IPF patients using ICD-10 diagnostic criteria. CI, confidence interval Forest plot showing the pooled prevalence of anxiety in IPF patients using screening tools. CI, confidence interval A total of 25 studies reported the prevalence of depression in IPF. Among them, three studies based on ICD-10 criteria yielded a pooled depression prevalence of 6.87% (95% CI: 3.23%–14.04%), with a 95% PI of 0.00%–99.95% and substantial heterogeneity (τ² = 0.4853, I² = 99.8%, P < 0.001) (Fig. 4). The 15 studies based on screening tools showed a pooled depression prevalence of 38.26% (95% CI: 24.65%–54.01%), with a 95% PI of 3.89%–90.46% and substantial heterogeneity (τ² = 1.4877, I² = 82.3%, P < 0.001) (Fig. 5). In studies on the prevalence of depression in IPF, a total of seven studies did not report the depression assessment tool and could not be included in the meta-analysis, thus only qualitative description was provided. These seven studies had sample sizes ranging from 47 to 10,996, with a cumulative sample size of 23,175 cases, and reported depression prevalence ranging from 1% to 20.3% (median 11%). Among them, four studies were based on hospital cohorts (prevalence ranging from 1% to 13%), two were based on administrative claims databases (prevalence of 14.2% and 20.3%, respectively), and one on a community-based cohort (11%). Forest plot showing the pooled prevalence of depression in IPF patients using ICD-10 diagnostic criteria. CI, confidence interval Forest plot showing the pooled prevalence of depression in IPF patients using screening tools. CI, confidence interval Subgroup analysis The prevalence of anxiety in IPF patients To explore potential sources of heterogeneity in the pooled prevalence of anxiety in IPF, subgroup analyses were performed based on the 12 studies in the screening-tool group. Stratification factors included study design, assessment tool, geographic region, methodological quality, and sample size. The detailed results are as follows. Study design: Five cross-sectional studies yielded a pooled prevalence of 43.41% (95% CI: 11.17%–82.39%), with substantial within‑subgroup heterogeneity (τ² = 4.0315, I² = 89.0%, P < 0.001). Seven cohort studies gave a pooled prevalence of 41.72% (95% CI: 28.34%–56.44%), also with substantial heterogeneity (τ² = 0.5732, I² = 89.2%, P < 0.001). The test for subgroup differences showed no statistically significant difference between study designs (P = 0.9431) (Fig. 6). Subgroup analysis of anxiety prevalence in IPF by study design using screening tools. CI, confidence interval Geographic region: Three studies from Asia gave a pooled prevalence of 39.24% (95% CI: 16.04%–68.59%), with substantial heterogeneity (τ² = 1.0961, I² = 95.4%, P < 0.001). Six studies from Europe yielded a pooled prevalence of 51.01% (95% CI: 19.46%–81.77%), also with substantial heterogeneity (τ² = 3.1365, I² = 88.8%, P < 0.001). Three studies from other regions produced a pooled prevalence of 33.77% (95% CI: 26.10%–42.40%), with moderate heterogeneity (τ² = 0.0230, I² = 56.3%, P = 0.101). The test for subgroup differences was not statistically significant (P = 0.6207) (Fig. 7). Subgroup analysis of anxiety prevalence in IPF by geographical region using screening tools. CI, confidence interval Assessment tool: Eleven studies using the HADS produced a pooled prevalence of 38.82% (95% CI: 22.52%–58.07%), with substantial heterogeneity (τ² = 1.6349, I² = 84.2%, P < 0.001) (Fig. 8). Only one study used the GAD, reporting a prevalence of 75.00%; it could not be pooled because of the insufficient number of studies. Subgroup comparison was therefore not performed. Subgroup analysis of anxiety prevalence in IPF by HADS tool. CI, confidence interval Methodological quality: Seven studies with moderate risk of bias (score 4–6) reported a pooled prevalence of 51.66% (95% CI: 21.98%–80.21%), with substantial heterogeneity (τ² = 3.0507, I² = 92.5%, P < 0.001). Five studies with low risk of bias (score ≥ 7) gave a pooled prevalence of 32.59% (95% CI: 26.88%–38.87%), with moderate heterogeneity (τ² = 0.0581, I² = 69.4%, P = 0.0108). No statistically significant difference was observed between the two quality subgroups (P = 0.2533) (Fig. 9). Subgroup analysis of anxiety prevalence in IPF by methodological quality using screening tools. CI, confidence interval Sample size: Six large‑sample studies (n ≥ 75) gave a pooled prevalence of 31.48% (95% CI: 26.40%–37.05%), with substantial heterogeneity (τ² = 0.0585, I² = 70.2%, P = 0.0049). Six small-sample studies (n < 75) yielded a pooled prevalence of 56.59% (95% CI: 21.92%–85.82%), also with substantial heterogeneity (τ² = 3.4565, I² = 92.5%, P < 0.001). The subgroup difference was not statistically significant (P = 0.1888) (Fig. 10). Subgroup analysis of anxiety prevalence in IPF by sample size using screening tools. CI, confidence interval The prevalence of depression in IPF patients To explore potential sources of heterogeneity in the pooled prevalence of depression in IPF, subgroup analyses were performed based on the 15 studies in the screening-tool group. Stratification factors included study design, assessment tool, geographic region, methodological quality, and sample size. The detailed results are as follows. Study design: Six cross-sectional studies yielded a pooled prevalence of 47.81% (95% CI: 12.98%–84.90%), with substantial within-subgroup heterogeneity (τ² = 4.7608, I² = 82.2%, P < 0.001). Nine cohort studies gave a pooled prevalence of 34.89% (95% CI: 25.72%–45.33%), also with substantial heterogeneity (τ² = 0.3770, I² = 84.3%, P < 0.001). The test for subgroup differences showed no statistically significant difference between study designs (P = 0.5733) (Fig. 11). Subgroup analysis of depression prevalence in IPF by study design using screening tools. CI, confidence interval Geographic region: Three studies from Asia gave a pooled prevalence of 33.20% (95% CI: 16.11%–56.26%), with substantial heterogeneity (τ² = 0.6477, I² = 93.6%, P < 0.001). Nine studies from Europe yielded a pooled prevalence of 43.36% (95% CI: 19.76%–70.41%), also with substantial heterogeneity (τ² = 2.8162, I² = 81.8%, P < 0.001). Three studies from other regions produced a pooled prevalence of 34.44% (95% CI: 30.55%–38.56%), with no heterogeneity (τ² = 0, I² = 0.0%, P = 0.722). The test for subgroup differences was not statistically significant (P = 0.8066) (Fig. 12). Subgroup analysis of depression prevalence in IPF by geographical region using screening tools. CI, confidence interval Assessment tool: Eleven studies using the HADS to assess depression produced a pooled prevalence of 37.26% (95% CI: 19.75%–58.90%), with substantial heterogeneity (τ² = 2.0801, I² = 74.9%, P < 0.001) (Fig. 13). Only one study each used the BDI, BDI-II, PHQ-9, and WDI, with prevalence estimates of 19.51%, 42.86%, 62.50%, and 49.15%, respectively; these could not be pooled because of the insufficient number of studies. Subgroup comparison was therefore not performed. Subgroup analysis of depression prevalence in IPF by assessment tool CI, confidence interval Methodological quality: Ten studies with moderate risk of bias (score 4–6) reported a pooled prevalence of 44.18% (95% CI: 22.02%–68.93%), with substantial heterogeneity (τ² = 2.5850, I² = 86.0%, P < 0.001). Five studies with low risk of bias (score ≥ 7) gave a pooled prevalence of 30.84% (95% CI: 25.86%–36.32%), with moderate heterogeneity (τ² = 0.0380, I² = 61.6%, P = 0.034). No statistically significant difference was observed between the two quality subgroups (P = 0.2886) (Fig. 14). Subgroup analysis of depression prevalence in IPF by methodological study quality using screening tools. CI, confidence interval Sample size: Eight large-sample studies (n ≥ 75) gave a pooled prevalence of 33.20% (95% CI: 27.26%–39.73%), with substantial heterogeneity (τ² = 0.1294, I² = 78.9%, P < 0.001). Seven small-sample studies (n < 75) yielded a pooled prevalence of 48.39% (95% CI: 15.85%–82.35%), also with substantial heterogeneity (τ² = 4.3457, I² = 86.9%, P < 0.001). The subgroup difference was not statistically significant (P = 0.4452) (Fig. 15). Subgroup analysis of depression prevalence in IPF by sample size using screening tools. CI, confidence interval Sensitivity analysis To assess the robustness of the findings, the following sensitivity analyses were performed: (1) leave-one-out analysis: each study was sequentially removed and the GLMM analysis was repeated to evaluate the influence of individual studies; (2) alternative effect measure transformation: the Freeman-Tukey double arcsine transformation with a random-effects model was used to re-pool the data. The prevalence of anxiety in IPF patients (1) Leave-one-out analysis: The ICD-10 group was not subjected to leave-one-out sensitivity analysis because fewer than three studies were available. For the screening-tool group (12 studies), the results showed that after removing any single study, the pooled prevalence ranged from 34.79% to 46.29%, all of which fell within the 95% CI of the primary analysis (25.28%–60.86%), indicating robustness (Fig. 16). (1) Leave-one-out analysis: The ICD-10 group was not subjected to leave-one-out sensitivity analysis because fewer than three studies were available. For the screening-tool group (12 studies), the results showed that after removing any single study, the pooled prevalence ranged from 34.79% to 46.29%, all of which fell within the 95% CI of the primary analysis (25.28%–60.86%), indicating robustness (Fig. 16). (2) Freeman-Tukey double arcsine transformation with REML: In the ICD-10 group, the pooled anxiety prevalence was 14.83% (95% CI: 6.54%–25.73%), which was similar to the primary analysis result (14.52%, 95% CI: 8.90%–22.79%). Heterogeneity measures were I² = 99.2%, τ² = 0.0096, and Q test P < 0.001. Because only two studies were included, the 95% PI could not be calculated (Fig. 17). In the screening-tool group, the pooled anxiety prevalence was 41.95% (95% CI: 26.47%–58.28%), with a 95% PI of 0.21%–95.75%, which was highly consistent with the primary analysis result (42.04%, 95% CI: 25.28%–60.86%). Heterogeneity was I² = 94.8%, τ² = 0.0803, and Q test P < 0.001. (Fig. 18). These results indicate that the pooled estimates were robust across different effect measures. (2) Freeman-Tukey double arcsine transformation with REML: In the ICD-10 group, the pooled anxiety prevalence was 14.83% (95% CI: 6.54%–25.73%), which was similar to the primary analysis result (14.52%, 95% CI: 8.90%–22.79%). Heterogeneity measures were I² = 99.2%, τ² = 0.0096, and Q test P < 0.001. Because only two studies were included, the 95% PI could not be calculated (Fig. 17). In the screening-tool group, the pooled anxiety prevalence was 41.95% (95% CI: 26.47%–58.28%), with a 95% PI of 0.21%–95.75%, which was highly consistent with the primary analysis result (42.04%, 95% CI: 25.28%–60.86%). Heterogeneity was I² = 94.8%, τ² = 0.0803, and Q test P < 0.001. (Fig. 18). These results indicate that the pooled estimates were robust across different effect measures. Sensitivity analysis using the leave-one-out method for anxiety in IPF patients using screening tools Sensitivity analysis using the Freeman-Tukey transformation for anxiety in IPF patients using ICD-10 diagnostic criteria Sensitivity analysis using the Freeman-Tukey transformation for depression in IPF patients using screening tools The prevalence of depression in IPF patients (1) Leave-one-out analysis: In the ICD-10 group, after removing any single study, the pooled prevalence ranged from 4.61% to 10.07%, all within the primary 95% CI (3.23%–14.04%), confirming the robustness of the primary analysis. In the screening-tool group, the leave-one-out results showed that the pooled prevalence after removing any single study ranged from 32.45% to 41.23%, all within the primary 95% CI (24.65%–54.01%). These findings suggest that the pooled prevalence estimates for both groups were robust (Figs. 19 and 20). (1) Leave-one-out analysis: In the ICD-10 group, after removing any single study, the pooled prevalence ranged from 4.61% to 10.07%, all within the primary 95% CI (3.23%–14.04%), confirming the robustness of the primary analysis. In the screening-tool group, the leave-one-out results showed that the pooled prevalence after removing any single study ranged from 32.45% to 41.23%, all within the primary 95% CI (24.65%–54.01%). These findings suggest that the pooled prevalence estimates for both groups were robust (Figs. 19 and 20). (2) Freeman-Tukey double arcsine transformation with REML: In the ICD-10 group (3 studies), the pooled depression prevalence was 7.55% (95% CI: 2.32%–15.43%), which was comparable to the primary analysis result (6.87%, 95% CI: 3.23%–14.04%). Heterogeneity was τ² = 0.0122, I² = 99.8%, and Q test P < 0.001. (Fig. 21). Because only three studies were available, the 95% PI was extremely wide (95.00%–89.44%) and of no practical value. In the screening-tool group (15 studies), the pooled depression prevalence was 38.91% (95% CI: 25.85%–52.83%), with a 95% PI of 0.49%–91.63%, which was highly consistent with the primary analysis result (38.26%, 95% CI: 24.65%–54.01%). Heterogeneity was τ² = 0.0730, I² = 94.3%, and Q test P < 0.001. (Fig. 22). These results demonstrate that changing the effect measure and estimation method did not substantially alter the pooled estimates, confirming the excellent robustness of the meta-analysis findings. (2) Freeman-Tukey double arcsine transformation with REML: In the ICD-10 group (3 studies), the pooled depression prevalence was 7.55% (95% CI: 2.32%–15.43%), which was comparable to the primary analysis result (6.87%, 95% CI: 3.23%–14.04%). Heterogeneity was τ² = 0.0122, I² = 99.8%, and Q test P < 0.001. (Fig. 21). Because only three studies were available, the 95% PI was extremely wide (95.00%–89.44%) and of no practical value. In the screening-tool group (15 studies), the pooled depression prevalence was 38.91% (95% CI: 25.85%–52.83%), with a 95% PI of 0.49%–91.63%, which was highly consistent with the primary analysis result (38.26%, 95% CI: 24.65%–54.01%). Heterogeneity was τ² = 0.0730, I² = 94.3%, and Q test P < 0.001. (Fig. 22). These results demonstrate that changing the effect measure and estimation method did not substantially alter the pooled estimates, confirming the excellent robustness of the meta-analysis findings. Sensitivity analysis using the leave-one-out method for depression in IPF patients using ICD-10 diagnostic criteria Sensitivity analysis using the leave-one-out method for depression in IPF patients using screening tools Sensitivity analysis using the Freeman-Tukey transformation for depression in IPF patients using ICD-10 diagnostic criteria Sensitivity analysis using the Freeman-Tukey transformation for depression in IPF patients using screening tools Publication bias Publication bias was assessed using Doi plots and the LFK index. The prevalence of anxiety in IPF patients The ICD-10 group included only two studies, so publication bias was not assessed. For the screening-tool group (12 studies), the LFK index was 2.337 ( > 2), indicating significant publication bias (Fig. 23). Further exploratory correction using the trim-and-fill method showed that no hypothetical studies were imputed (number of added studies = 0). The adjusted pooled prevalence was 40.53% (95% CI: 25.45%–57.63%), which was highly consistent with the original pooled estimate (42.04%, 95% CI: 25.28%–60.86%). Heterogeneity remained high (I² = 88.3%) (Fig. 24). Given that the trim-and-fill method did not impute any studies and the adjusted results were essentially consistent with the primary analysis, the impact of publication bias on the pooled estimate was considered limited. Therefore, we retained the primary analysis results as the main findings without adjustment. Publication bias assessment using Doi plot for anxiety prevalence in IPF patients using screening tools Trim-and-fill analysis for publication bias adjustment of anxiety prevalence in IPF patients using screening tools The prevalence of depression in IPF patients The ICD-10 group included only three studies, so publication bias was not assessed. For the screening-tool group (15 studies), the LFK index was 3.148 ( > 2), indicating significant publication bias (Fig. 25). Exploratory correction using the trim-and-fill method revealed that no hypothetical studies were imputed (number of added studies = 0). The adjusted pooled prevalence was 34.76% (95% CI: 26.01%–44.66%), which was largely consistent with the original pooled estimate (38.26%, 95% CI: 24.65%–54.01%). Heterogeneity remained high (I² = 84.9%) (Fig. 26). As the trim-and-fill method did not alter the original conclusion, we retained the primary analysis results as the main findings without adjustment. Publication bias assessment using Doi plot for depression prevalence in IPF patients using screening tools Trim-and-fill analysis for publication bias adjustment of depression prevalence in IPF patients using screening tools Discussion This systematic review and meta-analysis included 25 studies comprising 78,060 patients with IPF and provided the first pooled estimate of the prevalence of comorbid anxiety and depression in this population. The pooled prevalence of anxiety based on ICD-10 clinical diagnostic criteria was 14.52% (95% CI: 8.90%–22.79%), and that of depression was 6.87% (95% CI: 3.23%–14.04%). In contrast, using screening tools (e.g., HADS, PHQ-9), the pooled prevalence of anxiety was 42.04% (95% CI: 25.28%–60.86%) and that of depression was 38.26% (95% CI: 24.65%–54.01%). The estimated prevalence of anxiety and depression in IPF patients varied significantly according to the assessment method, with estimates derived from ICD-10 criteria being substantially lower than those from screening tools for both conditions. Notably, although the pooled point estimate from screening tools was approximately 40%, the 95% PIs were extremely wide (anxiety: 3.36%–93.80%; depression: 3.89%–90.46%), indicating that the true prevalence may vary considerably across different IPF populations. Therefore, this point estimate should be interpreted as an average trend across existing studies rather than a fixed constant applicable to all populations. Clinical decision-making should integrate individual patient characteristics and give greater weight to the range of the PI rather than relying solely on the point estimate. The core reason for the discrepancy between estimates based on ICD-10 criteria and those based on screening tools lies in their different thresholds and purposes. ICD-10 prioritises clinical confirmation and high specificity, requiring strict criteria for symptom duration, severity, and functional impairment, and relies on clinical interviews and coding records, which can lead to under-detection. In contrast, screening tools prioritise symptom detection and high sensitivity, actively identifying subclinical or untreated psychological symptoms but potentially including mild or transient mood disturbances, which may result in overestimation. These findings suggest that although the proportion of IPF patients meeting formal diagnostic criteria for a psychiatric disorder is relatively low, a considerable proportion experience clinically significant anxiety and depressive symptoms that warrant clinical attention and further evaluation. This phenomenon is not unique to IPF. In patients undergoing hip arthroscopy, the assessment method similarly influences the estimated prevalence of psychiatric conditions: a systematic review and meta-analysis of 38 studies reported that the prevalence derived from validated screening tools was significantly higher than that based on ICD codes [43]. Among critically ill patients receiving extracorporeal membrane oxygenation, the estimated prevalence of delirium also depends heavily on the assessment tool. A systematic review and meta-analysis of 13 studies and 8,679 patients found that the pooled prevalence of delirium was as high as 57% when assessed with the Nursing Delirium Screening Scale but only 7% when based on ICD-10 codes [44]. From a clinical perspective, relying solely on ICD-10 criteria may lead to underdiagnosis of clinically significant anxiety and depression. Many IPF patients who do not meet the formal diagnostic thresholds nevertheless experience substantial psychiatric symptoms; if left unaddressed, these may further reduce treatment adherence, impair quality of life, and even accelerate disease progression. Conversely, using screening tools alone may result in overdiagnosis and unnecessary referrals, increasing healthcare costs, while false‑positive results may also exacerbate patients’ psychological distress. Based on the above discrepancies, we recommend a stepwise assessment strategy: first, administer validated screening tools (e.g., HADS) to all IPF patients to identify at‑risk individuals; then, confirm screen‑positive cases using semi‑structured diagnostic interviews; and finally, initiate standardized mental health interventions. Collectively, this evidence supports that the choice of assessment tool is a fundamental methodological factor determining reported prevalence rates. Therefore, to improve the accuracy and comparability of future research, studies should clearly report the core characteristics, measurement objectives, and psychometric properties of the assessment tools used, and give priority to validated semi-structured diagnostic interviews [45]. Such efforts will generate more reliable epidemiological evidence, accurately define the mental health burden in IPF patients, and inform targeted clinical intervention strategies. In this study, the pooled I² statistic exceeded 80% for both the ICD-10 group and the screening-tool group, indicating substantial between-study heterogeneity. To explore potential sources of this heterogeneity, subgroup analyses were performed using study design, assessment tool, geographic region, methodological quality, and sample size as stratification factors (the ICD-10 group could not be analysed because of the small number of studies). For the screening‑tool group, the subgroup analyses revealed no statistically significant differences in the pooled prevalence of anxiety or depression across study design (cross-sectional vs. cohort), geographic region (Asia, Europe, other), methodological quality (moderate risk vs. low risk), or sample size (large vs. small) (all P > 0.05). However, these negative findings should be interpreted with great caution, and one should not directly infer that the above factors are unrelated to the prevalence estimates. A plausible explanation is insufficient statistical power. The power of subgroup analyses is influenced by the number of studies within each subgroup, the degree of between‑study heterogeneity, and the magnitude of the between‑group effect size difference. When the number of studies within a subgroup is small, the power of subgroup analyses is generally low, making it difficult to detect genuine effect modification. Furthermore, high residual heterogeneity (I² > 80%) persists within the subgroups, which further inflates the standard errors of the effect estimates and reduces the sensitivity of subgroup comparisons. In the screening-tool group of the present study, most subgroups contained fewer than ten studies, which may have compromised statistical power and thus masked true differences that might otherwise have been detected. In the subgroup analyses of this study, although the differences between geographic regions did not reach statistical significance, the point estimates of the prevalence of anxiety and depression in IPF patients showed a trend of lower estimates in Asia than in Europe: the prevalence of anxiety was 39.24% (95% CI: 16.04%–68.59%) in Asia and 51.01% (95% CI: 19.46%–81.77%) in Europe; the prevalence of depression was 33.20% (95% CI: 16.11%–56.26%) in Asia and 43.36% (95% CI: 19.76%–70.41%) in Europe. However, given the small number of studies within the subgroups and the considerable overlap of the 95% confidence intervals, the statistical robustness of this trend is limited, and the findings should be interpreted with caution. If future studies with larger sample sizes validate this observed difference, it may be attributable to variations in healthcare systems and sociocultural factors. At the level of healthcare systems, some European countries have integrated mental health screening into routine respiratory care, with greater accessibility to psychological services, making it more likely that patients are asked about mental health issues during clinical encounters, thereby increasing the detection and reporting of prevalence estimates [46]. In contrast, in some Asian regions, access to mental health services remains limited, and only a small proportion of individuals with mental disorders seek help from primary care physicians, which may lead to under-recognition of psychological distress among IPF patients and consequently lower prevalence estimates in screening studies, that is, systematic under-reporting [47, 48]. At the sociocultural level, in certain Asian cultural contexts, the social stigma associated with mental illness remains pervasive; patients may deliberately conceal or downplay their emotional symptoms for fear of discrimination, tending to underreport negative emotional experiences even during standardised screening [49]. Such response bias may lower the point estimates of prevalence derived from self‑report scales. Furthermore, the substantial overlap of confidence intervals between the two groups suggests that the observed differences cannot rule out the influence of sampling error and heterogeneity in measurement tools. In summary, the regional differences observed in this study may reflect a complex interplay between the true epidemiological profile of psychiatric comorbidities in IPF patients and systematic variations across regions in the identification, reporting, and management of mental health problems. All the above explanations warrant further investigation in future studies. Within the screening-tool group, point estimates of the prevalence of anxiety and depression showed certain differences between the sample size subgroups and the methodological quality subgroups. Regarding sample size, the pooled prevalence of anxiety in small-sample studies was 56.59% (95% CI: 21.92%–85.82%), higher than that in large-sample studies (31.48%, 95% CI: 26.40%–37.05%); a similar trend was observed for depression. This difference may be attributable to the greater susceptibility of small-sample studies to methodological limitations, particularly selection bias induced by non-probability sampling. Such studies often adopted single-center convenience sampling, which is prone to Berkson’s bias: patients with more severe disease and comorbid psychiatric symptoms have a higher propensity to seek medical care and participate in research, leading to overestimation of prevalence. Furthermore, the large sampling variability inherent in small-sample studies results in unstable effect estimates, potentially reducing the robustness of the pooled meta-analytic results [50, 51]. Regarding methodological quality, the pooled prevalence of anxiety in low-quality studies was 51.66% (95% CI: 21.98%–80.21%), which was higher than that in high-quality studies (32.59%, 95% CI: 26.88%–38.87%); a similar trend was observed for depression (moderate risk 44.18% vs. low risk 30.84%). These findings suggest that studies with higher methodological quality tend to report lower prevalence estimates. Further analysis of the quality assessment details of the included studies revealed that studies with low risk of bias strictly adhered to rigorous design standards, including clear sampling methods, adequate sample sizes, and standardized assessment tools, thereby effectively controlling for selection bias, information bias, and measurement bias, and making their prevalence estimates closer to the true population values. Within the screening-tool group, the distribution of assessment tools was unbalanced. For both anxiety and depression, apart from HADS, the other tools (GAD, BDI, BDI-II, PHQ-9, and WDI) each appeared in only a single study, precluding subgroup meta-analysis. Specifically, for the prevalence of anxiety in IPF, the pooled prevalence from 11 studies using HADS was 38.82% (95% CI: 22.52%–58.07%), whereas the single GAD study reported a prevalence of 75.00%. For the prevalence of depression, the pooled prevalence from 11 HADS-based studies was 37.26% (95% CI: 19.75%–58.90%), with substantial within-subgroup heterogeneity (τ² = 2.0801, I² = 74.9%, P < 0.001); the remaining tools (BDI, BDI-II, PHQ-9, and WDI) each appeared in only a single study, with prevalence estimates ranging from 19.51% to 62.50%. Moreover, significant heterogeneity was observed within the HADS subgroups for both anxiety and depression, which may be related to between-study differences in the implementation of HADS. Variations in the diagnostic cut-off values used, language versions of the scale, and modes of administration (self-administered vs. interviewer-assisted) can all influence the response patterns of IPF patients, leading to discrepancies in prevalence estimates. In summary, the unbalanced distribution of assessment tool types and the operational differences within the same tool jointly represent important sources of heterogeneity in the screening-tool group. Beyond the methodological factors examined in this study, clinical variables that were commonly underreported in the original studies also represent important confounders affecting the pooled prevalence estimates of anxiety and depression in IPF patients. Core indicators of disease severity, such as DLCO and Gender-Age-Physiology stage, may directly influence patients’ mental status and are closely associated with anxiety and depression scores, indicating that patients with more severe IPF have a higher probability of comorbid anxiety and mood disorders [5, 23]. Antifibrotic therapy, as the cornerstone of IPF management, can effectively slow the decline in FVC and stabilise the disease, thereby improving patients’ psychological state [52]. Moreover, several studies have confirmed that pulmonary hypertension, obstructive sleep apnoea, and gastro-oesophageal reflux disease are independent risk factors for anxiety and depression [53–55]. In addition, dyspnoea and various comorbidities significantly increase the prevalence of mood disorders in patients with interstitial lung disease [56]. Thus, the presence of pulmonary hypertension, obstructive sleep apnoea, gastro-oesophageal reflux disease, and other comorbidities in IPF patients may further increase the burden of anxiety and depression and should be considered as important clinical confounders. Consequently, differences across study populations in disease severity, antifibrotic regimens, and comorbidity profiles are potential sources of substantial between-study heterogeneity and bias in the pooled prevalence estimates. However, because the original studies did not provide prevalence data of anxiety or depression stratified by disease severity, antifibrotic use, or comorbidities, we could not perform the corresponding subgroup analyses to disentangle the confounding effects and accurately estimate the independent impact of these clinical characteristics on the pooled prevalence. Therefore, the prevalence estimates reported in this study reflect the overall average under the influence of heterogeneous clinical characteristics across included studies. Future studies should systematically collect and fully report these key clinical covariates to provide data support for precise confounding adjustment and to enhance the credibility of results in subsequent meta-analyses. In this study, publication bias was assessed using Doi plots and the LFK index. In the screening-tool group, we detected significant publication bias, with LFK indices of 2.337 for anxiety and 3.148 for depression (both > 2). Exploratory correction using the trim-and-fill method imputed no hypothetical studies; the adjusted pooled prevalence (anxiety: 40.53%; depression: 34.76%) was highly consistent with the original pooled estimates (anxiety: 42.04%; depression: 38.26%). These findings indicate that although publication bias was pronounced, its impact on the pooled effect estimates was limited, and the original results can be considered relatively robust. Nevertheless, the marked asymmetry indicated by the LFK indices still suggests a potential risk of overestimation of the pooled prevalence, and the possible influence of unpublished negative or null results on the overall estimates cannot be completely excluded. To the best of our knowledge, this is the first systematic review and meta-analysis to report the pooled prevalence of anxiety and depression in IPF patients separately based on ICD-10 diagnostic criteria and screening tools, directly revealing that differences in diagnostic criteria are a core source of heterogeneity. This study strictly adhered to the PRISMA guidelines, included 25 studies with 78,060 patients, and thus had good sample representativeness. Methodologically, we employed a GLMM, conducted dual sensitivity analyses (leave-one-out and Freeman-Tukey transformation), used Doi plots with the LFK index to assess publication bias, and reported 95% PI, thereby overcoming the limitations of traditional methods in small sample settings. Multivariable subgroup analyses systematically explored sources of heterogeneity within the screening-tool group. Our findings provide empirical support for integrating mental health assessment into the clinical management of IPF patients and offer valuable references for standardising the design of future studies. When interpreting the results of this study, several limitations should be acknowledged. (1) The assessment tools used in the included studies exhibited considerable heterogeneity. Although we stratified the analyses by diagnostic criteria (ICD-10 vs. screening tools) and explored the influence of different screening tools through subgroup analyses, inherent differences in symptom items, diagnostic thresholds, and psychometric properties across tools may have biased the pooled prevalence estimates and reduced cross-study comparability. Moreover, the ICD-10 group included very few studies with limited sample sizes and extremely high heterogeneity (I² > 98% for both outcomes). Consequently, the pooled prevalence estimates for this subgroup were unstable, with extremely wide 95% PIs, making it difficult to accurately reflect the true prevalence of clinically confirmed anxiety and depression in IPF patients. (2) Although we used random-effects models and sensitivity analyses to address heterogeneity, substantial heterogeneity persisted, and subgroup analyses did not identify clear sources of heterogeneity, suggesting the presence of unmeasured confounders that could not fully account for the observed heterogeneity. (3) Significant publication bias was detected in the screening-tool groups. Although sensitivity analyses indicated that this bias had a limited impact on the conclusions, it may still have led to overestimation of the prevalence of anxiety and depression based on screening tools. (4) The literature search was limited to English-language publications. Although English dominates international academic communication, this language restriction may have led to the omission of relevant studies published in other languages, thereby introducing potential publication bias. (5) Owing to the limited number of included studies, we were unable to perform meta-regression analyses to quantify the associations between various influencing factors and the prevalence of anxiety and depression. (6) Some original studies did not clearly report the assessment tools used and were therefore excluded from the meta-analysis and only described narratively, which may have introduced selective reporting bias. (7) The geographic distribution of the included studies was markedly imbalanced, with the vast majority coming from Asia and Europe and very few or no studies from Africa, Southeast Asia, and other regions. Thus, the generalisability of our findings to other cultural settings remains to be validated. Future research should adopt standardised diagnostic tools, systematically collect key covariates such as disease stage and treatment details, encourage the inclusion of multi-language publications, and conduct individual participant data meta-analyses to reduce heterogeneity and obtain more precise prevalence estimates. To reduce the impact of publication bias on pooled estimates, pre-registration of study protocols should be encouraged, along with active reporting of null or negative results. Additionally, prospective, multicentre designs with unified standardised assessment protocols are recommended to enhance diagnostic accuracy and cross-study comparability. When more high-quality studies become available, meta-regression analyses incorporating covariates such as age, sex, disease severity, use of antifibrotic medications, and geographic region should be considered to identify and quantify the sources of heterogeneity in prevalence estimates, thereby clarifying which factors independently influence the prevalence of anxiety and depression in patients with IPF. At the same time, attention should be paid to differences across regions (e.g., Latin America, Africa, Southeast Asia) and the impact of cross-cultural variations in symptom expression on the validity of assessment tools. On this basis, large-scale, multicentre, prospective cohort studies should be carried out to explore causal relationships between anxiety/depression and clinical outcomes in IPF, thereby providing higher-level evidence for the development of targeted comprehensive intervention strategies. Furthermore, given the high risk of psychiatric comorbidity observed in IPF patients, we recommend incorporating standardised psychological assessment tools into routine clinical care pathways. Clinicians should be aware of the elevated risk of anxiety and depression in this population and integrate regular mental health screening throughout patient management to enable early identification and stratified intervention for high-risk individuals. Ultimately, establishing evidence-based biopsychosocial management strategies for IPF is crucial for improving patients’ quality of life and long-term prognosis. Conclusion This systematic review and meta-analysis confirms that anxiety and depression are common among patients with IPF and are significantly influenced by diagnostic criteria: prevalence estimates based on ICD‑10 clinical diagnoses are substantially lower than those derived from screening tools. Although considerable heterogeneity was observed across studies and pre‑specified subgroup analyses did not identify a single major source of heterogeneity, sensitivity analyses confirmed the robustness of the pooled results. Our findings highlight the urgent need for uniform diagnostic criteria in research on psychiatric comorbidities in IPF. Future large‑scale, well‑designed prospective studies that adopt standardized assessment tools and consistent diagnostic criteria are required to more accurately estimate the burden of psychiatric comorbidity in IPF patients and to provide high‑quality evidence for the development of targeted comprehensive intervention strategies. Supplementary Information Supplementary Material 1. Acknowledgements Not applicable. Abbreviations Beck Depression Inventory Beck Depression Inventory-II confidence intervals Generalized Anxiety Disorder Hospital Anxiety and Depression Scale International Classification of Diseases, Tenth Revision Idiopathic pulmonary fibrosis Joanna Briggs Institute Patient Health Questionnaire-9 prediction intervals Preferred Reporting Items for Systematic Reviews and Meta-Analyses Well-being Index Authors’ contributions LLX and XQY conceived and designed the study. LLX and DYZ conducted the comprehensive literature search. LLX and RNY were responsible for data extraction and statistical analysis. ZHL and SGY assessed the quality of the included studies. LLX drafted the manuscript. DYZ, NXY and XQY revised the manuscript. XQY supervised the final review of the paper. All authors read and approved the final manuscript for submission. Funding This study was supported by the Noncommunicable Chronic Diseases-National Science and Technology Major Project of China (Nos. 2024ZD0522900 and 2024ZD0522901); the National Natural Science Foundation of China’s Young Scientists Fund (No. 82505809); the Key Project of the Joint Fund (Dominant Discipline Cultivation Category) under the Henan Provincial Science and Technology R&D Program (No. 242301420020); and the Clinical Collaboration Project of Integrated Traditional Chinese and Western Medicine for Severe and Refractory Diseases of China (No. ZDYN-2024-A-021). Data availability All data generated or analyzed during this study are included in the published article and its supplementary material. Declarations Ethics approval and consent to participate Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement All data generated or analyzed during this study are included in the published article and its supplementary material.