Impact of diabetes mellitus on mortality in pulmonary hypertension: A systematic review and meta-analysis

Abstract

BACKGROUND

Pulmonary hypertension (PH) is a progressive disease characterized by endothelial dysfunction and vascular remodeling and is a leading cause of mortality worldwide. Although it is independently associated with multiple comorbidities, the impact of diabetes mellitus (DM) on mortality in patients with PH remains uncertain. To address this issue, we conducted a systematic review and meta-analysis to investigate the effect of DM on survival in patients with pulmonary hypertension.

AIM

To investigate the impact of diabetes mellitus on mortality in pulmonary hypertension patients.

METHODS

We conducted a comprehensive search of four major electronic bibliographic databases like PubMed, Google Scholar, Scopus, and Embase, and identified 106 relevant studies, out of 1561 articles, published since the year 2000 for full-text review. Fourteen retrospective and prospective cohort studies that compared survival between patients with DM and those without DM in the context of PH were deemed eligible for inclusion in our meta-analysis. The study was registered on PROSPERO with the identifier CRD42023390232.

RESULTS

A total of 116455 patients with PH were included in the meta-analysis, of whom 41228 suffered from DM and 75227 did not. The results of our meta-analysis indicate an elevated mortality rate among PH patients with diabetes mellitus in comparison to those without DM [odds ratio (OR) = 1.40, 95%CI: 1.15–1.70, P = 0.0006]. The meta-regression analysis unveiled a statistically significant negative association between mean age and effect size (coefficient = -0.036, P value = 0.018). Conversely, a statistically significant positive association was detected between female proportion and effect size (coefficient = 0.000, P value < 0.001).

CONCLUSION

Our meta-analysis, which included approximately 116500 PH patients, revealed that the presence of diabetes mellitus was associated with increased odds of mortality when compared to non-diabetic patients. The meta-regression analysis indicates that studies with older participants and lower proportions of females tend to exhibit smaller effect sizes. Clinically, these findings underscore the importance of incorporating diabetes status into the risk stratification of patients with PH with more aggressive monitoring and early intervention to improve prognosis potentially.

Key Words: Pulmonary hypertension; Diabetes mellitus; Mortality; Systematic review; Meta-analysis

Core Tip: The high prevalence of comorbidities among pulmonary hypertension (PH) patients significantly complicates disease management and amplifies mortality risk. The presence of diabetes mellitus in PH patients is known to decrease survival rates, although the exact level of risk remains uncertain. Our study showed that in patients suffering from pulmonary hypertension, the presence of diabetes mellitus was significantly associated with decreased odds of survival when compared to nondiabetic patients from the same group. The sample's mean age and female proportion were also determined to moderate the effect size.

INTRODUCTION

Pulmonary hypertension (PH) is a complex interplay of cardiovascular and respiratory pathologies with a global impact, affecting approximately 1% of the world's population, especially among the elderly (> 65 years) [1,2]. PH is a rare multifactorial disease characterized by abnormal remodeling of the distal pulmonary vasculature, leading to significant health complications[3].

A pivotal proposal from the 6th World Symposium on Pulmonary Hypertension involved redefining the hemodynamic criteria for PH[4]. It is now defined by a mean pulmonary arterial pressure > 20 mmHg at rest, with specific criteria for PH, including a pulmonary vascular resistance > 2 Wood Units and a pulmonary arterial wedge pressure ≤ 15 mmHg[5,6]. The diagnostic algorithm for PH has been simplified, employing a three-step approach: Initial suspicion by frontline physicians, confirmation through echocardiography, and final validation through right heart catheterization in specialized PH centers.

The World Health Organization's Dana Point classification[7] categorizes PH into distinct groups based on underlying causes[8,9]: Group 1: Idiopathic Pulmonary Artery Hypertension. Group 2: Associated with left heart disease. Group 3: Associated with lung diseases and/or hypoxia. Group 4: Associated with pulmonary artery obstructions. Group 5: With unclear and/or multifactorial mechanisms.

Irrespective of its origins, PH consistently signifies worsening symptoms and an elevated risk of mortality[10]. Additionally, a high prevalence of comorbidities is noted among patients with PH, which includes sleep apnea, thyroid disease, obesity, hypertension, and diabetes mellitus, among others. The presence of these comorbidities not only complicates disease management for patients but also amplifies mortality risk within this population[11-13].

Diabetes mellitus (DM), which is distinguished by its precipitous blood glucose levels, presents a substantial strain on healthcare systems worldwide[14]. The International Diabetes Federation estimates that 537 million individuals were afflicted with diabetes in 2021[15]. It ranked eighth in global death and disability in 2019, accounting for 1.5 million deaths, about 20% of which were of cardiovascular origin[16-18].

Comprehensive studies have elucidated diabetes's systemic effects, particularly on endothelial dysfunction, increasing the risk of myocardial infarction, stroke, and mortality[19,20]. It is currently theorized that the anaerobic pathway largely metabolizes glucose in blood vessels. DM, a chronic glucose metabolic disorder, can harm the vascular endothelium by increasing the mitochondrial division, causing sustained vascular contractions leading to PH[21]. In line with this, epidemiologically, diabetes is linked to an increased prevalence and mortality in pulmonary hypertension patients[22,23]. Understanding this link is critical for developing more specific guidelines for PH management and improved prognosis.

This systematic review and meta-analysis aims to address this gap in the literature regarding the association between diabetes mellitus and mortality in pulmonary hypertension patients.

MATERIALS AND METHODS

The present study was based on the principles outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement[24]. The study has been registered on PROSPERO with the ID CRD42023390232[25]. A part of this manuscript has been presented as an abstract at the CHEST 2023 conference[26].

Search strategy

To compile the results, we searched the following expansive databases, namely PubMed, EMBASE, Scopus, and Google Scholar, using relevant keywords like “pulmonary hypertension", "death", “outcome”, “survival”, "mortality", and “diabetes mellitus”. These were combined with appropriate Boolean operators to give us our desired search strategy.

We searched clinicaltrials.gov[27] for completed and ongoing randomized controlled trials (RCTs) to identify additional pertinent literature and eligible studies. Relevant articles published since the year 2000 onwards were considered, and only those with full text available were reviewed. Attempts were also made to obtain access to the complete articles by contacting the authors when necessary. The search strategy for different databases is detailed in Supplementary file 1.

Inclusion criteria

To establish a link between mortality in PH patients and diabetes as a comorbidity, this study employed specific inclusion criteria. Efforts were made to reach out to the study investigators to acquire any incomplete information in the publication. These criteria were as follows: (1) Full-text, peer-reviewed articles such as cohort studies, randomized controlled trials, and case-control studies; (2) Studies reporting survival outcomes or mortality events in PH patients with available comparative data for DM patients and non-DM patients; and (3) Articles published after the year 2000.

Exclusion criteria

To ensure the accuracy and reliability of the research, our systematic review and meta-analysis entailed strict exclusion criteria. We excluded studies that involved: Patients below the age of 18; Pregnant or lactating females. Publications in any language other than English. Animal studies, in vitro or molecular studies. Non-original data like editorials, letters, and protocols. Conference abstracts and case reports. All conflicts and discrepancies were addressed through mutual discussions.

Study selection

Three authors (SP, PM, and RK) downloaded all articles from electronic bibliographic databases and uploaded them on Rayyan AI^®, and duplicates were eliminated[28]. All authors independently evaluated titles and abstracts to identify key articles. Data screening in Rayyan AI^® was blinded to each author to reduce bias. Subsequently, all authors conducted a comprehensive review of the full text during the second screening phase. All reasons for exclusions were registered independently and individually. All discrepancies were addressed by arbitration and the lead author mediated the resolution of points of disagreement.

Data extraction

We included all articles that satisfied our search criteria. Our team of authors carefully selected appropriate articles, retrieved data, and organized it methodically using MS Excel^®. Several pieces of information were extracted, including the article title, first author's name, DOI, publication year, country, study design, study period, definition of PH, participants' mean age, gender, total number of PH patients with and without diabetes mellitus, mortality events among participants with and without diabetes mellitus, and continuous or categorical values. Any disagreements or conflicts were resolved unanimously through mutual discussion between the authors.

Data synthesis and statistical analysis

The primary outcome of our meta-analysis was all-cause mortality in diabetic patients with pulmonary hypertension. To investigate this, we ran a meta-analysis of fourteen cohort studies that compared mortality rates between diabetic and non-diabetic PH patients. For each study, we also calculated a crude odds ratio (OR) using raw data for events and non-events and used a random-effects model to account for variability between and within studies. The results were examined using the Mantel-Haenszel method for binary data to calculate combined odds ratios using Review Manager (RevMan) Version 5.4^® from the Cochrane Collaboration^® in 2020[29]. The P value below 0.05 for OR was considered significant. We anticipated heterogeneity and therefore used a random-effects model to calculate summary statistics, with heterogeneity assessed using the Cochrane Q value and I² statistics. We defined I² values of 25%–50%, 50%–75%, and > 75% as low, moderate, and high heterogeneity, respectively. A sensitivity analysis was performed to assess the impact of each study's effect size on the overall summary. The presence of publication bias was investigated using the funnel plot and Egger test.

Exploring heterogeneity & moderation

To explore potential moderators of the effect size and heterogeneity across studies, we conducted subgroup analyses and meta-regression.

Statistical analysis

The included studies were categorized into five predetermined subgroups: The mean age of the overall population mentioned in the studies (< 60 years vs ≥ 60 years), study location by country (United States vs non-United States), duration of follow-up (< 5 years vs ≥ 5 years), sample size (< 200 vs ≥ 200), and observed risk of bias for individual studies (high, moderate and unclear risk of bias). To evaluate the robustness of our findings and investigate potential moderators of the effect size, we conducted sensitivity analyses employing a sequential removal approach within each subgroup. We tracked changes in heterogeneity (measured by the I² statistic) and effect size estimates within each subgroup, as well as across the entire analysis, following the removal of each study. This systematic procedure helped identify potentially influential studies and assess the generalizability of the subgroup analysis findings.

Multivariate meta-regression

To further explore potential moderators of the effect size across studies, we conducted two separate meta-regression analyses. The first model included the continuous moderator variables (mean age, female proportion, hypertension proportion, and duration of follow-up period) in each study. The second model included the categorical subgroup variables country (United States vs non-United States), follow-up period (< 5 years vs ≥ 5 years), sample size (< 200 vs ≥ 200), and observed risk of bias for individual studies (high, moderate, and unclear risk of bias). Both models were fitted using a restricted maximum likelihood approach to account for potential between-study heterogeneity. The overall significance of each model was assessed using the omnibus P value. Individual covariate effects within each model were examined through their respective coefficients and P values. All statistical tests were two-tailed with a P value ≤ 0.05.

This two-pronged approach of subgroup analysis and meta-regression allowed us to comprehensively explore potential factors influencing the effect size and heterogeneity in our meta-analysis.

Risk of bias assessment. The National Institutes of Health (NIH) Quality Assessment Tool^® for Observational Cohort and Cross-Sectional Studies was used to measure the risk of bias in cohort studies[30]. Three reviewers (SP, RC, and VG) assessed bias likelihood individually. Two other reviewers (SP and PM) independently performed the final assessment for the NIH instrument after assessing the risk of bias associated with each negative response. Studies were graded according to the quality of the evidence as very low (= high risk of bias), low (= moderate risk of bias), fair (= unclear risk of bias), good (= low risk of bias but has some potential flaws), and high (= very low risk of bias). A third knowledgeable reviewer (RK/VB) helped to resolve disagreements through consensus.

RESULTS

After an initial and thorough atheneum search conducted as mentioned above, 1561 potentially relevant articles were identified from these sources. Out of these, 269 duplicate research papers were omitted, leaving 1292 articles for further review. Out of these, 106 articles fulfilled the requirements for full-text evaluation. The remaining manuscripts were scrutinized further, and 92 were excluded for various reasons: 16 articles had no full text available despite efforts to contact the respective authors; 28 papers did not establish a direct correlation between DM and mortality in PH patients; and 48 studies did not provide explicit data for calculating the odds ratio. Ultimately, 14 studies were included for qualitative synthesis and meta-analysis[31-44] (Figure 1). The PRISMA checklist is provided in Supplementary file 1.

Figure 1 PRISMA flow diagram for included studies. DM: Diabetes mellitus; PH: Pulmonary hypertension.

Patient and study characteristics

The study population consisted of 116455 PH patients internationally. We analyzed the data by gender, and it was discerned that the majority of patients were males (n = 108920, 93.53%), while females accounted for 6.47% of the total population (n = 7535). Out of these, 41228 (35.40%) patients concurrently suffered from diabetes mellitus. Another most frequently reported comorbidity was hypertension, which was recorded in 13 out of the 14 papers studied. These studies included a total of 5960 patients with PH, of which 69.67% (n = 4152/5960) were classified as having hypertension. Other notable comorbidities included chronic pulmonary diseases, coronary artery disease, hyperlipidemia, chronic kidney disease, obesity, and smoking, among others. The average duration through which the patients were followed was 100.57 months. The vast majority of pertinent research that fit our inclusion criteria was carried out in the United States (n = 6)[34,35,36,39,43,44], with remaining studies conducted in the Czech Republic (n = 1)[41], Taiwan (n = 1)[42], Korea (n = 1)[40], France (n = 1)[37], Poland (n = 1)[38], Israel (n = 1)[32], China (n = 1)[45], and Hong Kong (n = 1)[31]. Supplementary Table 1 offers a detailed summary of the different attributes of the studies analyzed. In conducting the risk of bias assessment, we found that 3 studies[37,38,40] had very low-quality evidence, suggesting a high risk of bias, while 10 studies[32,34,35,36,39,41,42,43,44,45] had low-quality evidence, which was in turn suggestive of a moderate risk of bias. Only one study had a fair quality of evidence[31], indicating an unclear risk of bias and the need for more evidence to derive any conclusion (Supplementary Table 2).

Meta-analysis: The results of our meta-analysis indicate an elevated mortality rate among PH patients with diabetes mellitus in comparison to those without DM (OR = 1.40, 95%CI: 1.15–1.70, P = 0.006). Notably, there was considerable heterogeneity observed, with an I² value of 60%. Importantly, sensitivity analyses validated that none of the individual studies exerted a substantial influence on the overall odds ratio. The detailed forest plots for mortality depicting subgroup analyses as well as their sensitivity analyses excluding the most dominant study are presented in Supplementary Figure 1-14. The overall forest plots for mortality and the sensitivity analysis excluding Trammell et al[34], are visualized in Figure 2.

Figure 2 Forest plot. A: Overall forest plot for all-cause mortality for diabetics vs non-diabetics pulmonary hypertension patients; B: Forest plot depicting sensitivity analysis after excluding the study by Trammell et al[34].

Meta-regression

We conducted two separate meta-regression analyses to explore potential moderators of the effect size across studies.

Continuous variables

The overall model exhibited statistical significance (Omnibus P value = 0.034), indicating an association between at least one of the continuous variables (mean age, female proportion, hypertension proportion, follow-up period) and the effect size. Meta-regression analysis revealed a significant negative correlation between mean age and effect size (coefficient = -0.036, P value = 0.018). This suggests that studies with older participants tended to report smaller effect sizes than those with younger participants. Conversely, a significant positive correlation was found between female proportion and effect size (coefficient = 0.000, P value < 0.001), indicating that studies with a higher proportion of female participants tended to yield larger effect sizes compared to those with a lower proportion of females. However, there was no statistically significant association between the proportions of participants with hypertension (coefficient = -0.003, P value = 0.534) or the duration of follow-up (coefficient = -0.028, P value = 0.205) and the effect size. These findings suggest that these variables do not significantly moderate the effect size according to the analysis conducted. The results indicate that these variables do not have a significant impact on the effect size based on the analysis performed. The results of multivariate regression analysis for continuous variables can be found in Table 1.

Table 1 Multivariate regression analysis for continuous and categorical variables.

Covariate	Coefficients	Lower bound	Upper bound	Std. error	P value	Omnibus, P value
Model 1 for continuous variables
Intercept	2.774	0.958	4.591	0.927	0.003	0.034
Mean age	-0.036	-0.066	-0.006	0.015	0.018
Female proportion	0.000	-0.000	0.001	< 0.001	0.075
Hypertension proportion	-0.003	-0.013	0.007	0.005	0.534
Follow-up period in years	-0.028	-0.071	0.015	0.022	0.205
Model 2 for subgroup variables (categorical)
Intercept	0.429	-0.820	1.677	0.637	0.501	0.002
United States vs non-United States	0.088	-0.188	0.364	0.141	0.534
Sample size < 200 vs sample size ≥ 200	0.407	-0.097	0.911	0.257	0.113
Follow up < 5 years vs ≥ 5 years	-0.541	-1.251	0.169	0.362	0.135
Mean age < 60 years vs ≥ 60 years	-0.301	-0.637	0.036	0.172	0.080
Study quality score based on NIH risk of bias	0.250	-0.101	0.601	0.179	0.162

NIH: National Institutes of Health.

Subgroup variables

We further conducted a multivariate meta-regression analysis on predefined subgroups, revealing a statistically significant overall model (omnibus P value = 0.002). This suggests that at least one of the subgroup covariates [country (United States vs non-United States), sample size (< 200 vs ≥ 200), follow-up period (< 5 years vs ≥ 5 years), mean age (< 60 years vs ≥ 60 years) or study quality] may be associated with the odds ratio. However, upon closer examination of individual covariate effects, certain limitations emerged. While covariates such as sample size (less than 200 vs 200 or more) and follow-up period (less than 5 years vs 5 years or more) displayed trends indicating a potential association with the effect size, these associations did not achieve statistical significance at our chosen alpha level (P value ≤ 0.05. Similarly, the association between study quality and effect size was not statistically significant. Furthermore, the analysis suggested a potential moderating effect of participant age (less than 60 years vs 60 years or older). Studies involving younger participants appeared to exhibit different effect sizes compared to those involving participants aged 60 years or older. However, it is crucial to note that this association did not attain statistical significance (P value = 0.080). This may be due to constraints such as the small sample size (14 studies) or the incorporation of multiple covariates in the model. Therefore, the finding regarding age should be interpreted cautiously and warrants further investigation in future research with larger sample sizes. The results of subgroup analysis and meta-regression can be found in Table 2.

Table 2 Subgroup analysis of studies based on country, sample size, follow-up period, mean age, or study quality.

Group	Included studies	Odds ratio (95%CI)	P value	P-within subgroups heterogeneity	I2%	P-meta-regression
Overall risk of mortality	14	1.40 (1.15-1.70)	< 0.001	0.002	60	NA
Subgroup analysis
Country United States vs non-United States
United States	6	1.17 [0.89-1.53]	0.26	0.27	22	< 0.001
Non-United States	8	1.55 [1.25-1.92]	< 0.001	0.14	36
Sample size < 200 vs ≥ 200
Sample size < 200	7	1.22 [0.71-2.12]	0.47	0.06	50	0.340
Sample size ≥ 200	7	1.46 [1.19-1.80]	< 0.001	0.003	70
Follow up < 5 years vs ≥ 5 years
Follow up < 5 years	5	1.52 [0.78-2.98]	0.22	0.06	56	0.640
Follow up ≥ 5 years	9	1.37 [1.13-1.67]	0.001	0.006	63
Overall mean age < 60 years vs ≥ 60 years
Mean age < 60 years	7	1.61 [1.27-2.05]	< 0.001	0.40	3	0.138
Mean age ≥ 60 years	7	1.26 [0.99-1.61]	0.06	0.004	68
Study quality score based on NIH risk of bias
High	3	2.52 [1.47-4.32]	0.0008	0.90	0	0.807
Moderate	10	1.23 [1.05-1.43]	0.010	0.20	26
Unclear	1	1.78 [1.40-2.25]	Not applicable	Not applicable	Not applicable

NIH: National Institutes of Health.

Publication bias

The funnel plot is illustrated in Figure 3. According to the figure, the intercept estimate was 0.67 (SE = 0.29), with a statistically significant P value of 0.041 (t-test = 2.29). This suggests potential publication bias, as the positive intercept indicates a tendency for studies with larger effect sizes to be included in the analysis. While the slope estimates of 0.05 (SE = 0.01) provided limited additional information, the positive intercept, and significant P value combined suggest a potential publication bias favoring studies with larger or statistically significant effects. The visual inspection of the funnel plot asymmetry is indicative that three studies with imputed data fell on the left side (negative effect sizes), three studies appeared above the top of the funnel plot, and two studies fell on the adjusted CES line, further supporting this finding. These inferences suggest that studies with negative or non-significant results might be missing from the analysis, potentially influencing the overall effect size estimate.

Figure 3 The Funnel Plot with Eggers test depicting the publication bias. SE: Standard error; CILL: Confidence interval lower limit; CIUL: Confidence interval upper limit; Adjusted CES: Adjusted combined effect size.

DISCUSSION

This systematic review and meta-analysis, representing a pioneering endeavor, investigates the prognostic implications of diabetes mellitus in individuals with PH. Our comprehensive review of 14 cohort studies involving 116455 participants reveals that those with concurrent DM and PH exhibit significantly higher all-cause mortality rates compared to their counterparts with PH alone. The meta-regression analysis indicates that mean age and female proportion may moderate the effect size. Studies with older participants and lower proportions of females tend to exhibit smaller effect sizes. Despite advancements in understanding this association, the precise mechanistic interplay between PH and adverse outcomes in individuals with DM remains elusive[46]. While some researchers theorize that insulin resistance, accompanied by elevated levels of inflammatory cytokines, may directly contribute to endothelial cell damage, others have highlighted the potential exacerbating effects of deficiencies in apoE and PPARγ on pulmonary hypertension[47,48]. Additionally, adipokines such as adiponectin and resistin, which undergo dysregulation in the context of insulin resistance, have emerged as key players in the pathophysiology of PAH, exerting deleterious effects on the pulmonary vasculature and remodeling the myocardial tissue[49,50]. Notably, elevated HbA1c levels and insulin resistance have also been linked to the incidence of systolic heart failure, hinting at a potential association with right ventricular dysfunction in the context of PAH[51-53]. Lastly, accumulating evidence implicates the triad of glucose intolerance, metabolic syndrome, and insulin resistance in the progression of PAH, though further elucidation is required regarding the temporal relationship between glucose abnormalities and PAH manifestation[54-56].

Given the significant morbidity and mortality linked to pulmonary hypertension, comprehending the impact of diabetes mellitus becomes paramount. Our research findings echo those of previous studies conducted in various countries, including Korea[57], Kenya[58], Spain[59], Poland[22], Australia[60,61], and the United States[62], indicating a consistent association across different ethnicities and geographic regions. Likewise, an analysis of the REVEAL registry by Poms et al[63] underscored how diabetes and COPD independently augmented mortality risk in PH patients compared to other comorbidities. Poms and her colleagues also went a step further to stratify diabetes and obesity, to remove the confounding effects of obesity on survival. In the same vein as our results, a COMPERA analysis conducted by Rosenkranz et al[64] revealed that when multiple comorbidities were incorporated into the Cox proportional hazard model, only coronary heart disease and diabetes exhibited associations with heightened mortality risk. Similarly, Belly et al[65] concluded that long-term outcomes in PH patients correlated with HbA1c levels in a dose-dependent manner. The multivariate analysis calculated a 2.2-fold increase in the risk of all-cause mortality in PH patients per 1-unit rise in HbA1c levels. Aligning with our results, these observations in pre-diabetic patients may indicate that impaired glucose metabolism plays a crucial role in mortality in this patient cohort. The largest contribution to our study stemmed from the United States Veteran Health Affairs System analysis of a substantial cohort of 110495 veterans with PH[34]. It also revealed that DM elevated the risk of death by 28% to 33%, while overweight and obese veterans were shielded compared to underweight or normal-weight individuals. A comparable pattern was observed in several additional studies, wherein the adjusted hazard ratios for patients with DM ranged from 1.80 to 11.5[40,41,66]. Moreover, in, a 2020 analysis of nearly 3800 PH patients also revealed that older age was significantly associated with decreased survival[67].

Nevertheless, conflicting findings emerged from a study conducted by Strange et al[68] on PH patients in Australia and New Zealand. This prospective cohort study revealed that only male gender and 6-minute walk distance remained independent predictors of survival on multivariate analysis. Although comorbidities like hypertension, coronary artery disease, and diabetes were more prevalent, none exerted enough influence to remain independent predictors. From 2001 to 2012, Anand et al[69] observed trends and outcomes of PH-related hospitalizations using the largest inpatient national database in the United States. Despite observing an increased burden of comorbidities in these inpatient PH cases, the presence of diabetes was surprisingly associated with a decreased mortality rate. However, the study was limited by the fact that patients who died with a different discharge diagnosis were not included in this database. Similarly, Datta et al[70] also reported no correlation of DM with survival in PH patients, although the study was limited by a small sample size.

Comparing the findings of our meta-analysis with those of other comorbidities, a similar association was observed. Tang et al[71] demonstrated increased all-cause mortality in chronic kidney disease (CKD) patients (RR = 1.44, 95%CI: 1.17–1.76) and those with end-stage renal disease (ESRD) on maintenance dialysis (RR = 2.32, 95%CI: 1.91-2.83). Interestingly, the MA on body mass index (BMI) demonstrated a non-linear dose-response relationship with mortality in PH patients, presenting a summary RR per 5-unit BMI increment of 0.83 (95%CI: 0.77–0.89). This reveals a distinct pattern, potentially attributed to the "obesity paradox" described by Jiang et al[72], where the lowest risk was observed at BMI 32–38 kg/m².

This analysis has several notable strengths: Firstly, the substantial sample size, along with the robustness and consistency of outcomes observed in both primary and subgroup analyses, enhances the reliability of the findings. Furthermore, additional meta-regression was conducted to account for confounding variables. Secondly, the study's outcomes' relevance holds significance in critical care and pulmonary hypertension research. Our findings underscore the importance of including diabetic status in the risk stratification of PH. Recognizing this elevated risk could guide clinical management decisions, leading to more aggressive monitoring and early intervention to improve patient outcomes. Lastly, we believe this is a pioneering SR/MA published on this topic, further contributing to its overall strength.

As with any other research, our study has inherent limitations. Firstly, the observational nature of the included studies prevents us from establishing causality. While our study identifies an association with increased mortality, it does not clarify the underlying cause. Secondly, despite sensitivity analysis, moderate to high heterogeneity persisted, indicating clinical diversity among the included studies. Other factors that could potentially affect heterogeneity would include variations in the sample size, age ranges as well as gender proportions, and differences in the definition of PH in each study. Third, some studies adjusted for factors like obesity and hypertension, which may act as intermediates or confounders, influencing the identified correlations. Fourth, it is important to acknowledge that the meta-regression analysis was conducted with a relatively small number of studies (n = 14). This raises concerns about the power of the analysis to detect true associations, particularly for the non-significant findings. Additionally, including four continuous variables in the model with a small sample size increases the risk of overfitting. Therefore, these findings should be interpreted with caution. Lastly, our search was confined to major databases, omitting others like ISI Web of Knowledge. Our search was limited to English literature, as research published in languages apart from English would require extensive resources for their review, data interpretation, and translation, which we lack at the moment. Along with this, we have also excluded grey literature, such as conference articles and patents, as these sources provide insufficient data for robust statistical analysis, adding to the potential publication bias and limiting the comprehensiveness of our findings. Furthermore, the lack of data on diabetes severity (HbA1C ranges) in our studies prevented the stratification of findings and assessing its impact on mortality in pulmonary hypertension. In addition, having all pulmonary hypertension groups, rather than group-specific adds to the limitation of this study.

Further exploration of the intricate relationship between DM and pulmonary hypertension is crucial for refining interventions and enhancing clinical outcomes in this complex patient population. We need research studies to delve into the pathophysiological mechanisms of diabetes and its impact on mortality in PH patients. Conducting randomized or clinical trials is pivotal in establishing a causal relationship, particularly in the context of how controlling diabetes mellitus affects pulmonary hypertension outcomes, aiming to reduce fatality. Risk stratification studies are needed to understand the influence of incremental increases in HbA1c levels, uncontrolled vs controlled DM, or the duration of DM on outcomes in pulmonary hypertension patients.

CONCLUSION

Pulmonary hypertension is a major cause of death worldwide, and numerous comorbidities have been identified as exacerbating the risk. Our systematic review and meta-analysis indicate a significantly reduced survival rate among patients with diabetes mellitus compared to those without DM. The meta-regression analysis suggests that mean age and female proportion might be potential moderators of the effect size. However, dedicated animal studies and large-scale RCTs are required to gain a comprehensive understanding of the effect of DM on survival in pulmonary hypertension and to investigate other potential modifiers. These endeavors are imperative for delineating the precise role of DM in influencing outcomes in PH patients.