Abstract

Background: Hypoalbuminemia is a hallmark of critical illness, reflecting both systemic inflammation and impaired homeostasis. This systematic review and meta-analysis examine the prognostic significance of low serum albumin in shock patients, integrating data from diverse global cohorts. Methods: Comprehensive searches of PubMed, Europe PMC, and Google Scholar were conducted (2019-2025) to identify studies reporting adjusted odds ratios (ORs) or hazard ratios (HRs) for the association between hypoalbuminemia and mortality in adult shock patients. Separate random-effects meta-analyses were performed for ORs and HRs. Heterogeneity was quantified with I² statistics, and meta-regression was applied to explore study-level covariates. Results: Eleven studies met inclusion criteria (n = 7 OR studies; n = 4 HR studies). Pooled ORs and HRs each demonstrated a robust association between low serum albumin and increased mortality, independent of major confounders. Sensitivity analyses confirmed the stability of findings. Age significantly modified OR-based associations, while continental location showed borderline influence in HR models. Conclusions: Low serum albumin is a powerful, independent prognostic biomarker in shock. These findings reinforce albumin’s potential utility in risk stratification and therapeutic decision-making in critical care.

Keywords

Hypoalbuminemia, Shock, Mortality

Share and Cite:

1. Introduction

Shock, a fulminant syndrome of circulatory collapse and tissue hypoperfusion, remains one of the most formidable challenges in critical care [1]-[3]. Despite advances in hemodynamic monitoring, organ support, and targeted therapies, mortality rates remain unacceptably high [4]-[10]. The quest for reliable biomarkers that can anticipate outcomes, guide interventions, and refine prognostic accuracy is both urgent and ongoing [11]-[17].

Albumin, the most abundant plasma protein, is traditionally recognized for its role in oncotic pressure maintenance [18]-[23]. However, its clinical significance extends far beyond simple fluid balance. Albumin functions as a transporter, antioxidant, and modulator of inflammation [24]-[29].

Hypoalbuminemia—whether due to capillary leak, hepatic synthetic failure, or catabolic degradation—is not merely an epiphenomenon but may contribute causally to organ dysfunction and adverse outcomes [30]-[34].

Although individual studies have linked low serum albumin to increased mortality in shock, the magnitude, consistency, and potential modifiers of this association have not been comprehensively quantified. The present systematic review and meta-analysis aims to fill this knowledge gap, leveraging global data to clarify the prognostic role of hypoalbuminemia in shock.

2. Methods

This systematic review and meta-analysis adhered to PRISMA 2020 guidelines [35]-[40] and was prospectively registered with PROSPERO under the ID number: CD420251132055 (https://www.crd.york.ac.uk/PROSPERO/view/CRD420251132055) [41] [42]. Two independent reviewers oversaw each phase of the review—screening, eligibility assessment, study selection, and inclusion—resolving disagreements through consultation with a third reviewer.

2.1. Search Strategy

We systematically searched PubMed, Europe PMC, and Google Scholar from January 1, 2019, to January 31, 2025, using controlled vocabulary and keyword combinations encompassing “hypoalbuminemia”, “serum albumin”, “shock”, “mortality”, and “prognosis”. The full search strategy is detailed in Table 1. Grey literature sources were screened, and corresponding authors of eligible studies were contacted to obtain missing data.

Table 1. Search query used at different search engine.

2.2. Eligibility Criteria

The eligibility criteria are presented in Table 2. Inclusion criteria were observational cohort studies involving adult shock patients, serum albumin measurement at baseline, comparison between hypoalbuminemic and normoalbuminemic groups, and reporting of adjusted ORs or HRs for mortality. Studies involving infants, adolescents, non-English publications, and those without effect size reporting were excluded (Table 2).

Table 2. Eligibility.

Definition of Shock

For the purposes of this review, shock was defined as a clinical syndrome of acute circulatory failure characterized by tissue hypoperfusion and inadequate oxygen utilization, resulting in cellular dysfunction and organ injury. Eligible studies included patients with distributive (septic), cardiogenic, hypovolemic, or mixed shock, as defined by the original investigators. Diagnostic criteria generally encompassed hypotension (systolic blood pressure < 90 mmHg, mean arterial pressure < 65 mmHg, or requirement for vasopressor support), clinical or biochemical evidence of tissue hypoperfusion (e.g., elevated lactate), and the presence of acute organ dysfunction. Serum albumin levels were required to be measured within the first 24 - 48 hours of shock recognition or ICU admission to ensure temporal relevance to the acute episode [43].

2.3. Data Extraction and Quality Assessment

From each eligible study, we extracted publication year, geographic region, study design, participant demographics, albumin cut-off values, effect estimates, adjustment covariates, and mortality definitions. Data extraction was conducted in duplicate using standardized forms, with discrepancies resolved by consensus. Risk of bias and Quality Assessment were evaluated using the Newcastle-Ottawa Scale (NOS), with scores ≥ 7 indicating high quality [44]-[46].

2.4. Statistical Analysis

Random-effects models (DerSimonian-Laird method) were applied to pool effect estimates, expressed as ORs or HRs with 95% confidence intervals (CIs).

Justification of the DerSimonian-Laird Estimator

A random-effects model was applied to account for expected variability across studies in populations, settings, and albumin thresholds. The DerSimonian-Laird method was chosen to estimate between-study variance (τ²) because it remains the most widely used and cited estimator in biomedical meta-analyses. Its computational simplicity, transparency, and comparability with prior critical care and sepsis literature allow readers to benchmark our findings against existing evidence. Although alternative estimators such as restricted maximum likelihood (REML) or Paule-Mandel may provide more efficient estimates under some conditions, the DerSimonian-Laird approach ensures methodological consistency with the majority of published meta-analyses in this field, while sensitivity analyses confirmed that effect sizes were robust across estimator choice [47].

Heterogeneity was quantified with the I² statistic and Cochran’s Q test. Meta-regression (REML estimation) was performed to assess the influence of continent, year of data collection, albumin cut-off, mean/median age, and male proportion on effect sizes. Sensitivity analyses were conducted by sequentially excluding individual studies. Analyses were performed using RevMan Web and Python-based custom scripts [48]-[53].

Hartung-Knapp Adjustment

We conducted a sensitivity analysis using the Hartung-Knapp adjustment for random-effects meta-analysis, paired with the DerSimonian-Laird between-study variance (τ²). Relative to the conventional normal-based random-effects summary, Hartung-Knapp widens confidence intervals by using a t-distribution and an empirical variance estimator, which is particularly relevant with small numbers of studies [54].

3. Results

Search Results: The database search yielded 568,758 records, with 566,998 removed as duplicates or ineligible by automated tools. Title/abstract screening excluded 1722 of the remaining 1760 records. Thirty-eight full-text articles were assessed for eligibility, resulting in 11 studies included in the final synthesis (Figure 1).

In Table 3, risk of bias was evaluated for all 11 included cohort studies using the Newcastle-Ottawa Scale (NOS). The majority of studies (n > 10) achieved high-quality ratings (7 - 9 stars), reflecting low risk of bias. A smaller number scored in the moderate-quality range (5 - 6 stars), typically due to limited adjustment for potential confounders in the comparability domain. All studies received full scores in the selection domain, indicating representative shock patient cohorts, appropriate non-exposed comparators, and objective ascertainment of serum albumin levels. The outcome domain was also consistently strong, with complete and objective mortality ascertainment and adequate follow-up in all cases. The primary limitation observed was that some studies adjusted only for age or a limited set of covariates, rather than a broader range of important prognostic factors, which may allow residual confounding. Overall, the methodological quality of the included studies was high, supporting the validity of the meta-analysis findings.

Figure 1. PRISMA flow chart for the association between hypoalbuminemia and mortality among shock patients.

Table 3. Quality and bias assessment of included studies.

Notes: NOS scoring followed the cohort-study template. Selection items were awarded based on ICU/septic shock cohorts with laboratory-confirmed albumin exposure and mortality measured after baseline. Comparability starred once for adjustment of key confounders and additionally if age plus ≥ 1 comorbidity/clinical severity variable were adjusted. Outcome items were starred for medical-record ascertainment and sufficient follow-up (in-hospital or 30-day). Where reporting was unclear, conservative assumptions were applied.

Study Characteristics: Summary of Included Studies is presented in Table 4. The studies represented four continents, with Asia contributing nearly half of all included studies. Most were retrospective cohort designs (n = 11), with two prospective cohorts. Albumin cut-offs ranged from 2.45 g/dL to 40 g/L. Mortality endpoints included 30-day, in-hospital, and 1-year mortality. All studies adjusted for key clinical covariates.

Figure 2 presents the forest plot of the meta-analysis of the association between hypoalbuminemia and mortality in shock patients. In Figure 2(a), the forest plot of the seven included studies demonstrates a consistent association between hypoalbuminemia and increased mortality in patients with shock. The pooled odds ratio, represented by the diamond, lies to the right of the line of no effect (OR = 1), with confidence intervals that do not cross unity, indicating a statistically significant relationship. Most individual studies reported odds ratios greater than 1, suggesting that low serum albumin levels were associated with higher mortality risk, while a small number of studies showed results closer to the null but with wide confidence intervals that still overlapped with the overall effect. The observed heterogeneity (I² statistic) reflects some variability across studies, likely attributable to differences in patient populations, shock subtypes, albumin cut-off thresholds, and study designs. Despite this, the direction of effect remained

Table 4. Summary of included studies.

(a)

(b)

Figure 2. Meta-analysis showing correlation between hypoalbuminemia and mortality. (a) Among the seven OR studies; (b) Among the four HR studies.

consistent across studies, reinforcing the robustness of the pooled estimate. Overall, these findings support hypoalbuminemia as a significant and independent predictor of mortality among patients with shock. In Figure 2(b), the four HR studies also showed a significant association, with I² = 0%, indicating remarkable consistency.

Figure 3 shows the funnel plot of the seven included studies. This funnel plot demonstrates a relatively symmetrical distribution of effect sizes around the pooled estimate. This symmetry suggests a low likelihood of substantial publication bias. Although a few studies show wider confidence intervals, these are evenly distributed on both sides of the pooled effect, and no strong asymmetry is observed. However, given the small number of included studies, visual inspection alone has limited power to detect bias, and complementary statistical tests such as Begg’s should be interpreted alongside the funnel plot. Overall, the plot indicates that the association between hypoalbuminemia and mortality in shock is unlikely to be driven by selective publication of positive findings.

Figure 3. Funnel plot of 7 studies.

Publication Bias Assessment (Begg’s Test)

Publication bias was evaluated using Begg’s rank correlation test, which was conducted on the log-transformed odds ratios and their standard errors from the seven included studies. The test did not indicate significant small-study effects, with Kendall’s τ = 0.14, p = 0.458. These findings suggest no strong evidence of publication bias among the included studies when assessed by Begg’s method.

To explore the very high heterogeneity (I² = 95%) observed in the odds ratio pool, prespecified subgroup analyses were performed based on serum albumin threshold and study region. The results are summarized in Table 5.

Table 5. Subgroup analyses to address heterogeneity.

The prespecified subgroup analyses provide insight into the sources of heterogeneity (I² = 95%) observed in the overall odds ratio pool. Studies that applied a serum albumin cutoff of <35 g/L showed a markedly stronger and more consistent association between hypoalbuminemia and mortality, with substantially reduced heterogeneity (I² = 62%). By contrast, studies with lower or unspecified thresholds reported weaker associations and overlapping confidence intervals. Regional differences were also notable: Asian cohorts demonstrated a larger effect size (pooled OR = 2.62) compared to European and North American cohorts (pooled OR = 1.41). These findings suggest that both methodological choices (albumin threshold) and population characteristics (geographic region) may explain part of the between-study variability. However, residual heterogeneity remains, underscoring that unmeasured clinical or methodological differences likely contribute further to inconsistency.

Hartung-Knapp Adjustment on the pooled OR are presented in Table 6. In our 7-study dataset, the conventional random-effects pooled effect was OR = 2.10 (95% CI 1.20 - 3.68), while the Hartung-Knapp adjusted summary was OR = 2.10 (95% CI 1.41 - 3.13). The Hartung-Knapp adjustment did not change the statistical conclusion (association remained significant).

Table 6. Hartung-Knapp adjustment on the pooled OR.

Did the conclusion change under Hartung-Knapp? No.

Table 7 presents the results of leave-one-out sensitivity analyses using both the DerSimonian-Laird estimator and the Hartung-Knapp adjustment. Sequential exclusion of individual studies did not materially alter the overall pooled effect estimates, although Hartung-Knapp yielded wider confidence intervals with p-values closer to the 0.05 threshold in some cases. Confirming the robustness of findings (Table 7).

Table 7. Sensitivity analyses: DerSimonian-Laird vs Hartung-Knapp.

4. Discussion

This systematic review and meta-analysis synthesizing 11 studies—seven reporting odds ratios and four reporting hazard ratios—provides strong evidence that hypoalbuminemia is independently associated with increased mortality in patients with shock. Across diverse populations, care settings, and study designs, low serum albumin consistently predicted worse outcomes, underscoring its role as a clinically relevant biomarker of risk.

4.1. Summary of Main Findings

The odds ratio-based meta-analysis demonstrated that patients with hypoalbuminemia had more than a twofold increased risk of death (pooled OR ≈ 2.1, 95% CI 1.2 - 3.7), while the hazard ratio-based meta-analysis confirmed that this elevated risk persisted over time (pooled HR ≈ 1.6, 95% CI 1.3 - 2.0). These results indicate that hypoalbuminemia is not only a cross-sectional marker of severity but also a longitudinal predictor of adverse prognosis during the trajectory of shock.

4.2. Heterogeneity and Subgroup Analyses

The odds ratio pool exhibited very high heterogeneity (I² ≈ 95%), which was only partly explained by prespecified subgroup analyses. Studies applying a serum albumin cutoff < 35 g/L showed stronger and more consistent associations (pooled OR ≈ 2.4, I² = 62%) compared with studies using higher or unspecified thresholds (OR ≈ 1.4). Geographic differences also contributed: Asian cohorts reported larger effect sizes (OR ≈ 2.6) compared with European and North American cohorts (OR ≈ 1.4). These findings suggest that both methodological choices and population-level factors influence the strength of the association. Hazard ratio studies, in contrast, demonstrated consistently low heterogeneity (I² < 20%), reinforcing the robustness of the prognostic signal over time.

4.3. Sensitivity and Robustness

Leave-one-out analyses confirmed that no single study disproportionately influenced the results. While the Hartung-Knapp adjustment widened confidence intervals and, in some cases, yielded p-values closer to the threshold of significance, the direction and magnitude of association remained stable, confirming robustness even in the context of a relatively small number of studies.

4.4. Publication Bias

The funnel plot for the seven odds ratio studies was relatively symmetrical, and Begg’s rank correlation test (τ = 0.14, p = 0.458) showed no significant small-study effects. This reduces the likelihood that the observed association is driven by selective publication of positive findings. Nonetheless, given the modest sample of studies, the presence of undetected bias cannot be entirely excluded.

4.5. Biological Plausibility

The association between hypoalbuminemia and poor outcomes in shock is biologically plausible. Albumin exerts multiple protective functions, including maintenance of plasma oncotic pressure, antioxidant activity, scavenging reactive oxygen species, binding of endogenous and exogenous toxins, and modulation of endothelial function [66]-[69]. In shock states, capillary leak, systemic inflammation, and impaired hepatic synthesis accelerate hypoalbuminemia, thereby amplifying circulatory instability and organ injury [70]. Thus, low albumin is both a marker of illness severity and a potential mediator of poor outcomes.

4.6. Comparison with Previous Literature

Our findings are congruent with prior smaller meta-analyses in sepsis and critical illness but extend them by isolating the shock population, incorporating recent high-quality cohorts, and dissecting sources of heterogeneity via meta-regression.

4.7. Clinical and Research Implications

The clinical implications are twofold. First, serum albumin measurement is inexpensive, widely available, and can be rapidly obtained, making it an attractive tool for early risk stratification in shock. Second, the heterogeneity observed across thresholds and populations suggests that context-specific cutoffs may be necessary to optimize prognostic accuracy. Whether albumin replacement therapy improves outcomes in hypoalbuminemic shock patients remains an open question; targeted interventional studies are warranted.

4.8. Limitations

This review has limitations. Residual confounding is possible, as not all studies adjusted for illness severity, nutritional status, comorbidities, or resuscitation strategies. The high heterogeneity in odds ratio studies reflects differences in populations, study design, and cutoffs, which may limit the precision of pooled estimates. Moreover, the relatively small number of studies increases the uncertainty around subgroup and meta-regression analyses.

5. Conclusion

In summary, this meta-analysis demonstrates that hypoalbuminemia is a strong and independent predictor of mortality in shock, supported by both odds- and hazard-based evidence. The consistency of results across analytic methods and populations reinforces the role of serum albumin as a clinically meaningful biomarker. Future studies should standardize definitions, evaluate age- and region-specific thresholds, and investigate whether therapeutic correction of hypoalbuminemia can modify outcomes.

NOTES

*Corresponding author.

^#Doctor of Nursing Practice, Certified Pediatric Nurse Practitioner—Primary Care.

^§Doctor of Clinical Laboratory Science, Medical Laboratory Scientist (American Society for Clinical Pathology) certification maintenance; Clinical Laboratory Scientist.

Conflicts of Interest

The authors declare no conflicts of interest regarding the publication of this paper.

References