Association between the Geriatric Nutritional Risk Index and all-cause mortality in older adults with obstructive sleep apnea symptoms: A retrospective cohort study

Abstract

Nutritional status assessment is crucial for managing comorbidities and improving outcomes in older patients with obstructive sleep apnea (OSA) symptoms, but it is often neglected. The Geriatric Nutritional Risk Index (GNRI) is an objective tool to assess nutritional health. This study aimed to explore the association between GNRI and all-cause mortality in older patients with OSA symptoms. The analysis used data from 1642 older patients with OSA symptoms from the National Health and Nutrition Examination Survey during 2005 to 2008 and 2015 to 2018. We assessed GNRI’s correlation with all-cause mortality using weighted survival analysis and Cox regression models, evaluating GNRI both quantitatively and categorically. The stability of the results was evaluated using propensity score matching. To gauge robustness, we computed the E-value – the smallest confounder odds ratio capable of nullifying the GNRI–mortality link. Among the 1642 participants, 20.03% died after weighting was applied. The cohort had a median age of 71 years, with 53.07% being women. With a GNRI threshold <98, 13.2% of participants were in the low-GNRI group. Weighted Kaplan–Meier analysis indicated higher all-cause mortality among those with low GNRI. After full adjustment in the weighted Cox regression analysis, GNRI demonstrated a hazard ratio of 0.94 (95% confidence interval: 0.91–0.97, P < .001). Further subgroup evaluations revealed no substantial interaction effects involving GNRI and all-cause mortality. The E-values corresponding to the point estimate and upper confidence limit for all-cause mortality among older patients with OSA symptoms were 2.7 and 1.97, respectively. In older patients with OSA symptoms, lower GNRI is linked to higher mortality risk.

1. Introduction

Obstructive sleep apnea (OSA) syndrome is characterized by recurrent episodes of complete or partial upper airway obstruction while sleeping, causing airflow that is either diminished (hypopnea) or halted (apnea) for at least 10 seconds and is linked to either cortical arousal or reduced blood oxygen levels.^[1,2] As per a Lancet study,^[3] OSA is a global issue affecting approximately 936 million people, among whom 425 million have moderate to severe forms of the condition. Among individuals aged 60 years and older, the incidence of OSA ranges from 27% to 80%.^[4] OSA significantly heightens the likelihood of cardiovascular events and increases overall mortality rates, primarily due to oxidative stress from nocturnal hypoxia.^[5,6] Animal model studies provide experimental evidence showing that chronic hypoxia stemming from OSA can disrupt nutritional status in rats.^[7] Meanwhile, malnutrition adversely impacts disease outcomes, evident through heightened hospital readmission rates, reduced exercise tolerance, significant airflow obstruction, and elevated mortality.^[8,9] Despite a range of effective treatments being available, malnutrition is still not well-recognized and is poorly managed in clinical practice for patients with OSA. Similarly, axial spondylarthritis in chronic low back pain patients is often overlooked due to overlapping symptoms, leading to diagnostic delays of years and severe adverse outcomes.^[10] This highlights that easily neglected key factors – whether disease type or nutritional status – in chronic conditions may impact prognosis, emphasizing the importance of targeted screening.

The Geriatric Nutritional Risk Index (GNRI) has garnered significant attention for its straightforward yet objective and comprehensive approach to nutritional assessment, which relies on serum albumin levels, current body weight, and ideal body weight (IBW).^[11] The GNRI has shown its prognostic value in predicting mortality and adverse outcomes in a variety of conditions, such as chronic kidney disease necessitating dialysis,^[11] heart failure,^[12] and multiple types of cancers.^[13–15] Its straightforward nature facilitates easy calculation and comprehension, enhancing its utility in identifying individuals in need of nutritional support.^[15] Nevertheless, to our knowledge, the precise link between GNRI and all-cause mortality among older OSA patients has yet to be fully investigated.

Building on the aforementioned background, this research seeks to elucidate how GNRI relates to all-cause mortality in older OSA patients. The study endeavors to facilitate the incorporation of nutritional assessment into OSA management protocols, provide robust clinical guidance, and introduce innovative approaches to enhance clinical outcomes for this patient population. We hypothesize that diminished GNRI values are associated with a higher rate of all-cause mortality among older OSA patients.

2. Methods

2.1. Study population and design

The National Health and Nutrition Examination Survey (NHANES) is a comprehensive series of studies designed to evaluate the health and nutritional status of adults and children in the United States. Conducted biennially by the National Center for Health Statistics, NHANES employs a meticulously selected, nationally representative sample of approximately 5000 individuals to reflect the broader US population. Effective analysis of large-scale data streams has been shown to improve feature extraction and analytical performance through optimized algorithmic design,^[16] and NHANES, as a classic large-scale population health data stream, provides a robust data foundation for exploring the prognostic factors of chronic diseases in the elderly population. Conducted under the auspices of the Centers for Disease Control and Prevention, the survey offers crucial information regarding the overall health and dietary well-being of the population. The survey consists of 2 main components: a household interview and a mobile medical examination. All participants provided written informed consent before data collection. Additionally, the study received ethical clearance from the Ethics Review Board of the National Center for Health Statistics. In this cohort study, older participants (aged ≥ 65 years) were recruited from NHANES cycles covering 2005 to 2008 and 2015 to 2018, as these cycles were the only source of relevant sleep questionnaires containing information on OSA. To ascertain possible OSA symptoms in participants, we adopted 3 key questions drawn from earlier work^[4] and classified individuals according to their answers: snoring 3 or more nights a week; having at least 3 episodes of snorting, gasping, or breath-holding weekly; and reporting extreme daytime sleepiness 16 to 30 times a month, despite sleeping at least 7 hours per night on workdays or work nights. Participants who answered affirmatively to any of these questions were classified as having OSA symptoms. The NHANES dataset remains a key resource for health and nutrition research. A total of 989 participants with incomplete mortality rates or covariate data were excluded from the study. The participant enrollment procedure is depicted in Figure 1.

Figure 1.

Flowchart of participant selection. BMI = body mass index, GNRI = Geriatric Nutritional Risk Index, NHANES = National Health and Nutrition Examination Survey, PIR = poverty income ratio, WBC = white blood cell.

2.2. Definitions of GNRI

GNRI was utilized in this study to evaluate the nutritional status of older patients with OSA. GNRI is calculated as follows:^[17] GNRI = 1.489 × serum albumin (g/L) + 41.7 × (present body weight [PBW, kg]/ideal body weight [IBW]), where IBW was defined as height² (m²) × 22. If PBW > IBW, the PBW/IBW ratio is set at 1. Consistent with clinical guidelines and previous research,^[18,19] a GNRI threshold of 98 has been established as a key marker for differentiating nutritional risk. Therefore, leveraging GNRI scores,^[20] we stratified OSA patients into 2 cohorts: those with low GNRI (≤98) and those with high GNRI (>98).

2.3. Outcomes

For this study, mortality data about OSA patients were merged with NHANES datasets. These datasets are publicly available on the NCHS Data Linkage portal (https://www.cdc.gov/nchs/linked-data/mortality-files/?CDC_AAref_Val). All-cause mortality was confirmed using the National Death Index through December 31, 2018. The duration of follow-up was determined from the date of the initial interview until the occurrence of death or the conclusion of the study period.

2.4. Potential covariates

Informed by prior NHANES research,^[21] we pinpointed several covariates associated with mortality as potential confounders. Initially, we extracted detailed demographic data from participants, including age, gender, and ethnicity. Specifically, participants were categorized into 4 ethnic groups: non-Hispanic Black, non-Hispanic White, Mexican American, and Other ethnicities. Subsequently, we extracted comprehensive data on participants’ body mass index (BMI), drinking habits, and smoking behavior. Additionally, the poverty income ratio (PIR) was divided into 3 categories: low (≤1.3), intermediate (>1.3, ≤3.5), and high (>3.5). For smoking status, participants were classified as: “never smokers” (those who smoked fewer than 100 cigarettes in their lifetime), “former smokers” (those who smoked at least 100 cigarettes but currently do not smoke), and “current smokers” (those who smoked at least 100 cigarettes and currently smoke, whether occasionally or daily). Alcohol consumption was divided into 5 groups: “never” (individuals who have never consumed alcohol), “former” (those who previously consumed alcohol but have abstained), “heavy” (defined as ≥3 drinks/day for women, ≥4 drinks/day for men, or binge drinking [≥4 drinks in 1 occasion for women, ≥5 drinks in 1 occasion for men] on ≥5 days/month), “moderate” (characterized by ≥2 drinks/day for women, ≥3 drinks/day for men, or binge drinking on ≥2 days/month, or a history of daily binge drinking), and “mild” (individuals not meeting the criteria for the other categories). Physical activity was stratified into 3 tiers: sedentary, moderate (marked by at least 10 minutes of movement in the past 30 days, inducing light perspiration or a noticeable yet manageable rise in breathing or heart rate), and vigorous (characterized by at least 10 minutes of vigorous activity within the past 30 days, leading to heavy sweating and a notable increase in breathing or heart rate). Hematological parameters were extracted from laboratory assessments. Supplementary lifestyle factors included marital status and educational attainment. Moreover, the analysis accounted for preexisting health conditions such as hypertension, heart failure, and diabetes, as well as medication history, including lipid-lowering agents, hypoglycemic agents, and antihypertensive medications.

2.5. Statistical analysis

When analyzing the NHANES data, this study adhered to the NHANES guidelines. To achieve nationally representative estimates, all analyses incorporated the Mobile Examination Center examination weight variables. The primary sampling unit variable and the stratification variable were employed to account for standard errors (SEs) arising from complex survey designs. Descriptive and regression analyses were conducted using weighted samples. For the descriptive analysis, continuous variables were expressed as means with SEs, and categorical variables were represented as weighted proportions (%). Differences among normally distributed continuous variables were evaluated via the t test. Additionally, categorical variables were examined using the χ² test. We utilized multivariable Cox proportional hazards regression models to assess the relationship between GNRI and all-cause mortality in older adults with OSA symptoms, estimating hazard ratios (HRs) and their corresponding 95% confidence intervals (CIs). We developed both unadjusted and 3 progressively adjusted models. More clinically significant covariates were included, as guided by previous reports.^[21,22] A multicollinearity assessment confirmed that variance inflation factors for all variables were below 5, indicating no multicollinearity in the regression model. In model 1, adjustments were made for sociodemographic factors, including age, gender, education, and race. In model 2, additional adjustments were included for marital status, BMI, PIR, alcohol consumption, smoking status, diabetes, hypertension, and heart failure. In model 3, we included serum uric acid as an additional covariate. Moreover, we conducted subgroup analyses via Cox regression models across various factors, including age, gender, race, hypertension, diabetes, smoking status, heart failure, and serum uric acid. Interactions between these subgroups and the GNRI were assessed using the likelihood ratio test. To elucidate the influence of the GNRI on all-cause mortality while mitigating potential biases from confounding factors, we utilized a Cox regression model in conjunction with propensity score matching (PSM). Specifically, a 1:1 nearest neighbor matching algorithm was implemented with a caliper width set at 0.2. The covariates from model 3 were employed to generate the propensity scores. The balance achieved through PSM was evaluated using the standardized mean difference, with a threshold of <0.1 deemed acceptable. Additionally, to assess the robustness of our logistic regression model, we computed the E-value. This metric signifies the minimum odds ratio that unmeasured confounders would need to achieve, given the measured covariates, to nullify the observed association between GNRI and all-cause mortality.

Since the sample size was dictated exclusively by the available data, no a priori statistical power analysis was performed. Statistical analyses were carried out using R software (https://www.r-project.org/, The R Foundation, Vienna, Austria) and Free Statistics software version 2.2 (Beijing FreeClinical Medical Technology Co., Ltd, Beijing, China).^[23,24] Statistical significance was determined by a two-sided P-value cutoff of <.05 across all analyses.

3. Results

3.1. Patient characteristics

The study utilized a dataset of 1642 geriatric patients with OSA symptoms, which is considered reflective of a population-based sample comprising 14,349,462 individuals. Table 1 delineates the fundamental attributes of the 1642 participants. The average age was 72.06 years, with a SE of 0.2 years. Among the entire cohort, 868 participants (46.93%) were male, with a median GNRI level of 103.21 mg/L. Participants were categorized into 2 groups according to their baseline GNRI, with 98 serving as the established cutoff value. Table 1 presents a comprehensive overview of the demographic, clinical, and laboratory profiles for both groups. The low-GNRI cohort (<98) was older, predominantly female, more likely to live alone, smoke, and have heart failure, and exhibited lower hemoglobin concentrations (all P < .05). PIR, education, BMI, alcohol intake, physical activity, hypertension, diabetes, and remaining laboratory indices showed no intergroup differences (all P > .05, Table 1).

Table 1.

Baseline characteristics of the study population.

Characteristics	Total (n = 1642)	Low-GNRI (n = 216)	High-GNRI (n = 1426)	P-value
Weighted participants (%)	14,349,462	1,705,325 (11.88)	12,644,136 (88.12)
Age, mean (SE)	72.06 (0.20)	73.67 (0.59)	71.85 (0.22)	.014
Gender, N (%)				.005
Male	868 (46.93)	99 (35.13)	769 (48.52)
Female	774 (53.07)	117 (64.87)	657 (51.48)
Race, N (%)	901 (81.41)	108 (75.17)	793 (82.25)	.003
Non-Hispanic White	901 (81.41)	108 (75.17)	793 (82.25)
Non-Hispanic Black	262 (6.09)	59 (11.19)	203 (5.40)
Mexican American	189 (3.78)	23 (5.03)	166 (3.61)
Other	290 (8.73)	26 (8.61)	264 (8.74)
Marital, N (%)				.015
Married or living with partners	1060 (69.24)	121 (59.74)	939 (70.52)
Living alone	582 (30.76)	95 (40.26)	487 (29.48)
PIR, N (%)				.054
Low income	391 (13.33)	58 (19.77)	333 (12.46)
Medium income	784 (44.16)	102 (44.31)	682 (44.14)
High income	467 (42.52)	56 (35.92)	411 (43.41)
Education, N (%)				.089
<9th grade	262 (7.60)	26 (7.17)	236 (7.66)
9–12th grade	632 (36.91)	88 (43.49)	544 (36.02)
>12th grade	748 (55.50)	102 (49.34)	646 (56.33)
BMI, mean (SE)	29.77 (0.23)	29.37 (0.57)	29.83 (0.26)	.23
Alcohol drinking, N (%)				.053
Never	313 (13.97)	36 (16.37)	277 (13.65)
Former	337 (19.17)	74 (29.62)	263 (17.77)
Mild	716 (49.43)	74 (35.96)	642 (51.24)
Moderate	172 (11.97)	17 (11.36)	155 (12.06)
Heavy	104 (5.45)	15 (6.69)	89 (5.29)
Smoke status, N (%)				.007
Never	763 (46.65)	83 (43.37)	680 (47.09)
Former	724 (45.08)	93 (40.23)	631 (45.73)
Current	155 (8.28)	40 (16.40)	115 (7.18)
Physical, N (%)				.10
Sedentary	959 (51.14)	127 (62.18)	832 (49.65)
Moderate	441 (32.48)	58 (24.52)	383 (33.56)
Vigorous	242 (16.38)	31 (13.30)	211 (16.80)
Hypertension, N (%)	907 (53.48)	130 (54.67)	777 (53.32)	.82
Heart failure, N (%)	161 (7.72)	42 (13.07)	119 (7.00)	.010
Diabetes, N (%)	443 (22.99)	74 (25.57)	369 (22.65)	.53
GNRI, mean (SE)	103.21 (0.22)	103.44 (0.26)	102.29 (0.35)	.018
Serum uric acid (mg/mL), mean (SE)	5.71 (0.05)	5.89 (0.15)	5.69 (0.06)	.43
WBC, mean (SE)	7.10 (0.09)	7.29 (0.20)	7.08 (0.09)	.24
Lymphocyte, mean (SE)	1.99 (0.04)	1.92 (0.09)	2.00 (0.04)	.75
Neutrophil, mean (SE)	0.61 (0.01)	0.62 (0.02)	0.61 (0.01)	.62
Hemoglobin, mean (SE)	14.12 (0.06)	13.46 (0.15)	14.22 (0.06)	<.001
Platelet, mean (SE)	236.11 (2.18)	247.99 (7.41)	234.51 (2.47)	.10
Antihypertensive drugs, N (%)	513 (31.71)	46 (21.91)	467 (33.03)	.049
Hypoglycemic drugs, N (%)	427 (22.26)	69 (24.12)	358 (22.01)	.62
Lipid-lowering drugs, N (%)	815 (50.94)	109 (47.82)	706 (51.36)	.48
All-cause mortality, N (%)	398 (20.03)	73 (31.09)	325 (18.54)	.001

Continuous variables are presented as weighted means (weighted standard errors) and categorical variables as raw frequencies (weighted percentages).

BMI = body mass index, GNRI = Geriatric Nutritional Risk Index, PIR = poverty income ratio, SE = standard error, WBC = white blood cell.

3.2. Survival analysis assessing the impact of GNRI on all-cause mortality

Weighted Kaplan–Meier analysis revealed that the lower GNRI group exhibited a higher rate of cumulative all-cause mortality events (Fig. 2, P < .001).

Figure 2.

Kaplan–Meier survival estimates for long-term all-cause mortality (weight). GNRI = Geriatric Nutritional Risk Index.

3.3. Relationships between GNRI and mortality outcomes in older individuals with OSA symptoms

By December 31, 2018, 398 participants (20.03%) had died. The mean follow-up duration was 5.59 years, with a SE of 0.24. Three distinct models were constructed to evaluate the influence of GNRI on all-cause mortality in older patients with OSA symptoms, utilizing adjusted weighted Cox proportional hazards models (Table 2). A negative association was found between GNRI and all-cause mortality among older patients with OSA symptoms when GNRI was modeled as a continuous variable in the unadjusted model. In particular, each 1-unit increment in GNRI corresponded to an 8% decrease in the risk of all-cause mortality (HR: 0.92, 95% CI: 0.89–0.96, P < .001). This significant relationship persisted in the fully adjusted model, with a HR of 0.94 (95% CI: 0.91–0.97, P < .001, model 3). Upon categorizing GNRI into groups, a lower GNRI level was associated with an elevated long-term risk of all-cause mortality compared with the higher GNRI group. In the unadjusted model, a lower GNRI was associated with a HR of 0.39 (95% CI: 0.27 to 0.57). When potential confounders were taken into account, a lower GNRI maintained a significant association with higher all-cause mortality risk, with a HR of 0.48 (95% CI: 0.34–0.67). The consistency of results across models highlights the strength of the statistical analyses. The restricted cubic spline model did not reveal a significant nonlinear relationship between GNRI and all-cause mortality (P = .081), as shown in Figure S1, Supplemental Digital Content, https://links.lww.com/MD/R630.

Table 2.

All-cause mortality hazard ratios for participants according to Geriatric Nutritional Risk Index (weight).

Status	Number of deaths/total	Crude model		Model 1		Model 2		Model 3
		HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value
All-cause
GNRI	398/1642	0.92 (0.89–0.96)	<.001	0.93 (0.90–0.97)	<.001	0.94 (0.91–0.97)	<.001	0.94 (0.91–0.97)	<.001
GNRI (category)
Low-GNRI	73/216	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
High-GNRI	325/1426	0.39 (0.27–0.57)	<.001	0.42 (0.30–0.57)	<.001	0.48 (0.34–0.67)	<.001	0.48 (0.34–0.67)	<.001

Crude model did not adjust for covariates.

Model 1 was adjusted for age, gender, race, and education.

Model 2 = model 1 + marital status + PIR + BMI + alcohol drinking + smoke status + hypertension + diabetes + heart failure.

Model 3 = model 2 + serum uric acid.

BMI = body mass index, CI = confidence interval, GNRI = Geriatric Nutritional Risk Index, HR = hazard ratio, PIR = poverty income ratio, Ref = reference.

3.4. Subgroup analysis

Figure 3 presents the findings of the weighted subgroup analysis, which shows a stable link between GNRI and all-cause mortality. Following stratification by key clinical and demographic factors, including gender, race, hypertension, diabetes, smoking status, heart failure, and serum uric acid levels, no significant interactions were detected across the subgroups (all P > .05).

Figure 3.

Subgroup analyses between malnutrition and all-cause mortality in older patients with obstructive sleep apnea symptoms. GNRI was treated as a continuous variable in the analysis. Serum uric acid levels were divided into 3 groups based on tertiles. CI = confidence interval, GNRI = Geriatric Nutritional Risk Index, HR = hazard ratio.

3.5. Sensitivity analysis

In all 4 PSM analyses, patients with higher GNRI values consistently exhibited a lower long-term all-cause mortality risk than those with lower GNRI values, as shown in Figure S2, Supplemental Digital Content, https://links.lww.com/MD/R630. To further substantiate the reliability of our initial results, E-values were utilized to assess the potential impact of unmeasured confounding factors. For geriatric patients with OSA symptoms, the E-values (relative risk) corresponding to the point estimate and upper confidence limit for all-cause mortality were 2.7 and 1.97, respectively, as depicted in Figure 4.

Figure 4.

E-values for other effect measures. RR = risk ratio.

4. Discussion

This study explored how GNRI is linked to all-cause mortality among older patients with OSA symptoms. Lower GNRI values were associated with an elevated risk of all-cause mortality in this cohort. This correlation remained stable across various subgroups defined by gender, race, hypertension, diabetes, smoking status, heart failure, and serum uric acid levels, with no significant interactions detected. Overall, our study underscores that a lower GNRI (≤98) signals an elevated mortality risk in this population.

Significantly, the all-cause mortality risk observed in older patients with OSA symptoms is consistent with prior research on GNRI-associated mortality across various patient groups, such as those in cardiac intensive care,^[25] with multiple organ dysfunction,^[26] pancreatic cancer,^[15] chronic kidney disease,^[20] and chronic obstructive pulmonary disease. Contrasting with prior research^[27] on OSA-related mortality in the NHANES database, our study utilized 8 years of comprehensive OSA data, thereby enhancing the robustness of our findings. In light of prior research identifying serum uric acid levels as a significant and independent predictor of mortality in OSA patients,^[21] we integrated serum uric acid levels into Cox regression model 3 and incorporated them into our stratified analysis based on trisection. The association between all-cause mortality and GNRI remained consistent.

While the HR (0.94) per unit increase in GNRI may seem modest, its clinical significance is substantial given its broad clinical range (0–141.7) derived from the original formula.^[17] Determined by serum albumin (0–55 g/L) and the ratio of actual to ideal weight (0–1, capped at 1 when exceeding ideal weight), this range enables cumulative unit effects: a 10-unit increase (e.g., 88 elevated to 98, crossing the nutritional risk cutoff) correlates with a 46.3% relative reduction in all-cause mortality (HR = 0.94¹⁰ ≈ 0.537). Categorical analysis with 98 as the cutoff further confirms this value, with the high-GNRI group (>98) showing a 52% lower mortality risk than the low-GNRI group (≤98) (HR = 0.48, 95% CI: 0.34–0.67). Consistent with literature across diverse populations, such as geriatric burn^[28] and acute myocardial infarction patients,^[29] the HR for mortality per unit GNRI increase ranges from 0.93 to 0.99 (Table S2, Supplemental Digital Content, https://links.lww.com/MD/R631), validating its robust cross-population protective effect.

Although the exact mechanism linking low-GNRI levels to increased mortality in older patients with OSA symptoms is not fully understood, several potential explanations can be explored. Among those with OSA, recurrent upper airway obstructions during sleep result in chronic intermittent hypoxia. This persistent secondary hypoxia disrupts adenosine triphosphate production, impairs mitochondrial function, and interrupts the canonical insulin signaling pathway, ultimately compromising protein synthesis.^[7] Concurrently, intermittent hypoxia impairs mitochondrial adenosine triphosphate and protein synthesis, while blunting intestinal mucosal responses to anorexia; combined with reduced intake and nutrient loss, this fosters malnutrition, frailty, and immune decline, linking low GNRI to excess mortality.

Notably, the 2 core components of GNRI serum albumin and body weight changes are closely associated with OSA pathophysiology. Undiagnosed myasthenia gravis in the postpartum period is also masked by the physiological manifestations of pregnancy, ultimately leading to severe respiratory failure and collapse crisis after delivery.^[30] This is similar to nutritional insufficiency in older OSA patients – its related symptoms are easily obscured by fatigue and somnolence caused by the disease itself, becoming a potential risk factor often neglected in clinical practice. Serum albumin serves as a marker of oxidative stress and systemic inflammation and can reflect the chronic intermittent hypoxic injury induced by OSA.^[7] Body weight fluctuations are common in OSA patients who often have reduced physical activity due to daytime sleepiness or suffer from nocturnal eating disorders. The mini nutritional assessment relies heavily on subjective self-reported items such as appetite and feelings of frailty. This shortcoming makes it prone to reporting bias in older OSA patients with daytime sleepiness and potential cognitive impairment, thus limiting its reliability in this population. Serum albumin alone neglects dynamic weight changes and BMI fails to distinguish muscle mass from fat mass. Neither of them can comprehensively reflect the nutritional disturbances related to OSA. Additionally, GNRI has well-validated prognostic value for mortality in geriatric chronic disease cohorts that share chronic hypoxia and systemic inflammation with OSA patients.^[11–15] Selecting GNRI enables horizontal comparisons with existing studies and fills the knowledge gap in older OSA patients, which further supports the clinical relevance of our findings.

This investigation offers several notable strengths. As far as we are aware, it stands as one of the few studies conducted within the US population. The use of weighted samples in our analyses enhances the applicability of our findings to the broader US adult population, an analytical strategy similar to the research conclusion that advanced algorithms enable efficient information mining from large-scale complex datasets.^[31] Additionally, our multivariate regression analysis accounted for laboratory variables more comprehensively than comparable studies. Finally, our subgroup analysis revealed that the observed association remained consistent across various subgroups.

Nevertheless, interpreting the findings must consider the study’s limitations. First, the absence of polysomnography in NHANES may introduce reporting bias in OSA symptom identification. However, this questionnaire-based approach has been widely adopted in prior NHANES OSA studies,^[4,27,32] ensuring the comparability of our results with existing literature and mitigating such bias. Second, exclusion of participants with missing serum albumin, body weight, or height data for GNRI calculation may induce potential selection bias. Yet this data-processing method is widely used in previous NHANES-based epidemiological studies,^[5,32] which ensures result comparability and renders the potential bias generally acceptable in large-scale cohort studies. Third, despite adjusting for numerous potential confounders and employing various alternative analytic methods that produced consistent results, residual bias from unmeasured confounding factors may still exist. Therefore, we employed the PSM method to validate the findings of the adjusted Cox regression model, and the results were consistent. Meanwhile, our E-value (2.7) exceeds the effect sizes of key confounders in similar populations: frailty (HR = 1.32^[27]) and clinical factors (diabetes: 1.15^[33]; diet score: 0.61–0.71^[4]; uric acid: 1.51^[21]; neutrophil-to-lymphocyte ratio: 1.16^[34]), as shown in Table S1, Supplemental Digital Content, https://links.lww.com/MD/R631. The potential impact of unmeasured confounders on the results is relatively limited, further supporting the robustness of the association between GNRI and all-cause mortality. Fourth, the NHANES data included in this study were from 2 cycles (2005–2008 and 2015–2018), and the temporal gap between the cycles may lead to heterogeneity in healthcare practices and mortality risk factors, thereby potentially affecting the study results. Finally, due to the cross-sectional design of the study, we are unable to establish definitive causality. The findings should be interpreted with caution, considering the study’s limitations, and viewed as a foundation for further investigation.

5. Conclusion

In older individuals with OSA symptoms, a lower GNRI is associated with a higher risk of all-cause mortality. Tailoring therapeutic strategies to individual nutritional needs in a clinical context is essential for reducing mortality risk.

Acknowledgments

We gratefully thank Dr Jie Liu of the Department of Vascular and Endovascular Surgery, Chinese PLA General Hospital & Physician-Scientist Center of China for his contribution to statistical support, study design consultations, and comments on the manuscript.

Author contributions

Conceptualization: Yan-Hong Lu, Chao Wu.

Funding acquisition: Chao Wu.

Data curation: Yan-Hong Lu, Zhao-Yao Wang.

Methodology: Yan-Hong Lu, Chao Wu.

Investigation: Chao Wu, Ji-Peng Liu.

Project administration: Chao Wu, Ji-Peng Liu.

Supervision: Chao Wu, Ji-Peng Liu.

Software: Zhao-Yao Wang.

Validation: Yan-Hong Lu.

Writing – original draft: Yan-Hong Lu, Chao Wu, Ji-Peng Liu.

Writing – review & editing: Yan-Hong Lu, Chao Wu, Ji-Peng Liu, Zhao-Yao Wang.