The CardioCHUVI-AF registry (formerly known as Registry of Atrial Fibrillation of Vigo, Spain) is a retrospective community study of patients with AF.12 The medical records of all patients were reviewed, and AF diagnosis was confirmed by the presence of a compatible electrocardiogram. Clinical, laboratory and therapeutic data were collected in an encoded database. The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the local ethics committee (Autonomous Committee of Research Ethics of Vigo, code HAC-ACO-2018-01, registry 2018/258). Due to the retrospective nature of the study, informed consent was not required. The CardioCHUVI-AF registry is registered in ClinicalTrials.gov with the identifier NCT04364516

All consecutive patients with a confirmed diagnosis of AF between January 2014 and January 2018 in the health area of Vigo (Galicia, Spain) were included (n=16 203). We excluded patients with missing baseline and follow-up data (n=146), and without data on weight/height (n=429) or nutritional status (n=779). The final cohort was composed of 14 849 AF patients. Patients were classified according to BMI in 4 groups: underweight (< 18.5kg/m2), normal weight (18.5-24.9kg/m2), overweight (25-29.9kg/m2), and obesity (≥ 30kg/m2). Subsequently, patients were classified according to nutritional status. For this purpose, the Controlling Nutritional Status (CONUT) score was employed.13 The score is derived from the values of serum albumin, total cholesterol, and lymphocyte counts. Albumin represents protein reserves, total cholesterol represents calorie depletion, and lymphocyte count represents immune defense. The CONUT score was calculated according to the original study (figure 1). This score has been used due to its simplicity and its prognostic validation in inpatients and outpatients with other cardiovascular diseases.14,15 A score of 0 to 1 was considered normal; scores of 2 to 4, 5 to 8, and 9 to 12 reflected mild, moderate, and severe malnutrition, respectively. Details of the creation of the cohort are shown in the supplementary data (figure 1 of the supplementary data).

Figure 1.

Scoring system for the CONUT score.

(0.41MB).

The primary outcome was all-cause mortality during follow-up, while the secondary outcome was embolic and hemorrhagic events. Embolic events included ischemic stroke and SEE, defined according to cardiovascular and stroke endpoint definitions for clinical trials performed by the Standardized Data Collection for Cardiovascular Trials Initiative (SCTI) and the US Food and Drug Administration.16 Hemorrhagic events included major bleeding, defined according to the statement of the International Society on Thrombosis and Haemostasis.17 Two data managers independently reviewed the clinical history of each patient to identify events. All suspected events were independently reviewed by 2 specialists in the clinically relevant areas of cardiology and internal medicine. Both data managers and specialists underwent a study-specific training course on event definitions and the adjudication process. Finally, an independent clinical events committee revised and adjudicated all events.

Continuous data are presented as the mean ±standard deviation and were compared using unpaired t tests. Categorical data are presented as counts (proportions) and were compared using chi-square tests. The incidences of all-cause mortality, embolic and hemorrhagic events were estimated using Kaplan-Meier curves. We used Cox proportional hazard regression models to estimate the association with all-cause mortality. For stroke/SEE and major bleeding, death served as a competing risk. Therefore, the incidence of each outcome was estimated using weighted cumulative incidence curves after the use of the Fine-Gray proportional subdistributions hazards model. The proportionality assumption was verified by testing for an interaction between the exposure variable and time, and no relevant violations were found. Three multivariate models were performed: a) adjusted for age and sex; b) adjusted for age, sex, statin therapy, anticoagulation therapy, Charlson index,18 HAS-BLED score,19 and CHA2DS2-VASc20 score; c) the same as model 2 but adding the CONUT score or BMI, as appropriate. Effect estimates from Cox models were reported as hazard ratios (HRs) while those from Fine-Gray models were reported as subdistribution HRs along with 95% confidence intervals (95%CIs).

Statistical analyses were conducted using STATA software, version 15 (Stata Corp, College Station, United States). A 2-sided P<.05 was considered statistically significant.

A total of 14 849 patients (75.6 ±10.3 years, 50.9% women) were followed up for 4.4 ±1.8 years. The mean BMI value was 30.3±4.8, with overweight and obesity observed in 42.6% and 46.0%, respectively (figure 2A). The mean CONUT score was 1.4±1.7, with mild and moderate/severe malnutrition observed in 29.0% and 5.3%, respectively (figure 2B). The baseline characteristics of the study population and differences according to nutritional status and BMI are summarized in the supplementary data (tables 1 and 2 of the supplementary data).

Figure 2.

Distribution of the study population according to body mass index (BMI) and nutritional status. Prevalence of overweight and obesity (A), and prevalence of malnutrition (B). The relationship between the CONUT score and BMI (C). The rate of malnutrition according to BMI groups (D).

(0.39MB).

As BMI increased, the CONUT score tended to decrease, but the relationship was not linear (figure 2C). Malnutrition rates were lower as BMI increased: from 48.1% in patients with underweight, to 36.8%, 35.1% and 33.0% in patients with normal weight, overweight, and obesity, respectively (P<.001 for trend; figure 2D).

During the follow-up, 3335 patients died (5.1 per 100 patients/y). The association between BMI and CONUT score with outcomes is detailed in table 1 and table 2. BMI was inversely associated with mortality (HR per 5kg/m2 0.76; 95%CI, 0.68-0.86; P<.001) in the univariate analysis; however, after various multivariate analyses including the CONUT score, BMI was not an independent predictor of mortality (table 1). The mortality rate was lower in patients with BMI ≥ 25kg/m2 (5.0 deaths per 100 patients/y) compared with those with BMI<25kg/m2 (6.2 deaths per 100 patients/y). However, when we adjusted for the presence of malnutrition (figure 3), those differences disappeared (HR for mortality comparing BMI ≥ vs <25kg/m2 [HR 0.97, 95%CI, 0.92-1.02; P=.176]. Neither obesity (HR, 0.91; 95%CI, 0.81-1.01; P=.088) nor overweight (HR, 1.04; 95%CI, 0.94-1.16; P=.438) were predictors of mortality in the multivariate analysis adjusted for nutritional status. The prognostic impact of overweight and obesity in patients with normal nutrition, mild malnutrition, and moderate-severe malnutrition is shown in figure 4a and table 3 of the supplementary data. Regarding nutritional status, both mild and moderate-severe malnutrition were associated with higher mortality in all BMI groups (normal weight, overweight, and obesity) (figure 4B and table 3 of the supplementary data). In patients with moderate-severe malnutrition, BMI was inversely associated with mortality, but not in those with mild or absent malnutrition (table 3 of the supplementary data). However, we did not find statistical significance for the association between BMI and mortality after adjusting for age, sex, comorbidities and moderate-severe malnutrition (table 4 of the supplementary data). The association between BMI and the CONUT score with mortality and composite endpoint (mortality, stroke/SEE and/or major bleeding) was maintained in various clinical subgroups (table 5 and table 6 of the supplementary data).

Figure 3.

Kaplan-Meier curve for patients with body mass index ≥ and <25kg/m2, stratified by the presence or absence of malnutrition. BMI, body mass index.

(0.21MB).

Figure 4.

Prognostic impact of overweight/obesity and malnutrition nutritional status. A: the risk of mortality and clinical outcomes for overweight/obesity according to nutritional status. B: the risk of mortality and clinical outcomes for malnutrition according to body mass index (BMI) groups. In view of small sample size (n=27), patients with underweight (body mass index <18.5kg/m2) were excluded from these analyses. 95%CI, 95% confidence interval; SEE, systemic embolic event.

(0.56MB).

A total of 984 patients had a stroke/SEE during follow-up, and 1317 had a major bleeding event. After we stratified by groups according to nutritional status (good nutrition, mild malnutrition, and moderate-severe malnutrition), neither overweight nor obesity were associated with embolic and hemorrhagic outcomes (figure 5). After various multivariate analyses including age, sex, comorbidities, and nutritional status, BMI was not associated with composite endpoint (mortality, stroke/SEE and/or major bleeding), in contrast with the CONUT score (table 1 and table 2).

Figure 5.

Incidence of the composite endpoint (mortality, ischemic stroke, systemic embolism, and/or major bleeding) according to nutritional status and body mass index (BMI) groups. Patients with underweight (BMI <18.5kg/m2, n=27) were excluded from the analysis.

(0.48MB).

This study has the inherent limitations of a retrospective study. Nutritional status was assessed by a simple screening tool, the CONUT score. We did not use more complex comprehensive nutritional assessments. Although our analysis indicates that a diagnosis of malnutrition attenuates the survival advantage of obesity, our diagnosis of malnutrition may be confounded by severity-of-illness markers that may not be related to nourishment and factors that may be more closely linked to nourishment but may not be sensitive to the response to nourishment. Furthermore, since total cholesterol is a parameter of the nutritional scale, the CONUT score could be affected by statin treatment. For this reason, multivariate adjustment included both the Charlson comorbidity index and concomitant statin therapy. Given that nutritional status was evaluated only at a single time point, we did not investigate changes in nutritional status over time and their relationship with clinical outcomes. Moreover, we did not evaluate the association of nutritional status with inflammatory markers or with body composition. In our study, data are limited to white patients, without information on educational attainment, marital status, or socioeconomic characteristics that might help us understand the contributing causes of malnutrition. Confirmation of our findings by other investigators with different health care and social systems would be welcome.