Association between preoperative serum albumin levels with risk of death and postoperative complications after bariatric surgery: a retrospective cohort study

Abstract

Background:

Hypoalbuminemia is common among individuals with obesity who qualify for bariatric surgery, but its relevance to clinical outcomes after bariatric surgery remains to be established.

Objectives:

To examine the association of preoperative serum albumin with 30-day postoperative outcomes.

Setting:

Data from the 2015–2019 Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program Participant Use Files were used.

Methods:

Preoperative serum albumin level was categorized as hypoalbuminemia (<3.5 g/dL), and normoalbuminemia (3.5–5.5 g/dL) among patients who underwent bariatric surgery. Multivariate logistic regression models were used to determine the association of preoperative hypoalbuminemia with 30-day postoperative mortality and other co-morbid outcomes.

Results:

Among 633,011 adult patients, 85.1% were women and the mean (standard deviation) age was 44.8 (12.0) years. The prevalence of hypoalbuminemia was 6.13% (n = 38,792). After adjustment for procedure type and demographic, lifestyle, and co-morbidity covariates, the odds ratio (OR) (95% confidence interval [CI]) for mortality was 1.42 (1.10, 1.82) for hypoalbuminemia. For all other outcomes, the ORs (95% CIs) for hypoalbuminemia ranged from 1.03 (.67–1.60) for cardiac arrest requiring CPR to 2.32 (1.66–3.25) for failure to be discharged by day 30. The ORs for several associations were higher for severe hypoalbuminemia than marginal hypoalbuminemia.

Conclusion:

Preoperative hypoalbuminemia was associated with several negative 30-day postoperative bariatric surgery outcomes and tended to be worse for severe hypoalbuminemia compared with marginal hypoalbuminemia. These findings suggest that serum albumin may be a useful biomarker to screen for negative bariatric surgery outcomes.

Obesity is a major public health crisis in the United States [1]. The age-adjusted prevalence of obesity increased from 30.5% in 2000 to 42.4% in 2018 [2] and is projected to increase to 47.1% by 2030 [3]. Consequently, the incidence rate of bariatric surgery increased from 45 per 100,000 patients in 2006 to 71 per 100,000 patients in 2015 [4]. Bariatric surgery is extremely effective in facilitating weight loss in both adults and adolescents, and is currently accepted as a treatment for diabetes among individuals with obesity [5–8]. Laparoscopic sleeve gastrectomy (LSG) is the most common type of bariatric surgery performed, followed by Roux-en-Y gastric bypass (RYGB) and laparoscopic adjustable gastric banding (LAGB) procedures [9]. As an elective surgery, it is particularly important to avoid deaths and minimize the risk of negative postoperative outcomes. Owing to the increasing demand for bariatric surgery, it is imperative to identify risk factors associated with postoperative death and complications [10].

A proposed biomarker that could be predictive of bariatric surgery postoperative complications is serum albumin. Albumin is the largest component of serum and is vital for maintaining oncotic pressure and other physiologic processes. Low albumin levels are associated with a variety of conditions including mortality among older adults [11,12]. In addition, albumin is an important indicator of inflammation (formerly thought to be a biomarker of malnutrition) and is commonly evaluated in the hospital setting [13]. Among hospitalized patients, serum albumin level is predictive of both short-term and long-term mortality following discharge [14]. Additionally, low preoperative serum albumin is associated with increased risk of postoperative morbidity and mortality in multiple other surgical fields such as urologic, oncologic, and orthopedic [15–18].

Hypoalbuminemia is common among individuals with obesity [19]. A study using data from National Surgery Quality Improvement Program (NSQIP) found that patients with obesity and hypoalbuminemia receiving elective general surgery were at an increased risk of morbidity and mortality [20]. Furthermore, a high prevalence of preoperative hypoalbuminemia has been observed among bariatric surgery candidates [21,22], and has been associated with negative bariatric surgery outcomes in preliminary studies [23–26]. Therefore, the objective of this study is to evaluate the association of preoperative serum albumin concentration with risk of 30-day mortality and other outcomes using data from the Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP).

Methods

Database

This retrospective analysis used the MBSAQIP Participant Use File (PUF) data set, which contains deidentified information on 135 Health Insurance Portability and Accountability Act–compliant variables for bariatric surgery cases collected from 868 participating centers in the United States and Canada. The MBSAQIP PUF contains data on over 90% of all bariatric surgeries performed in the United States and Canada [27]. These data are abstracted and deidentified from medical records by certified metabolic and bariatric surgical clinical reviewers who received database-specific trainings and pass a yearly certification exam [28]. In addition, preoperative and demographic variables, co-morbidities, information about the procedure, and 30-day outcomes are collected. Data integrity is ensured by an audit process requiring a disagreement rate of <5%.

Study population

In the present study, PUFs from 2015–2019 were used for analysis. Data were available on 964,459 adults ≥18 years of age who underwent metabolic or bariatric surgery. After exclusion of patients with missing information on preoperative serum albumin (n = 269,539), those with serum albumin. >5.5 g/dL (n = 568), and those who did not have a reported Current Procedural Terminology (CPT) code for bariatric surgery (n = 61,341) data were available on 633,011 adult patients and included in the present study.

Exposure assessment

Preoperative serum albumin measured within 90 days before the bariatric procedure was classified into 2 categories: hypoalbuminemia (<3.5 g/dL) and normoalbuminemia (3.5–5.5 g/dL). To further evaluate the effect of hypoalbuminemia on bariatric surgery outcomes, the study sample was further stratified into 3 categories: severe hypoalbuminemia (<2.5 g/dL), marginal hypoalbuminemia (2.5 to <3.5 g/dL), and normoalbuminemia (3.5–5.5 g/dL).

Outcome measures

All outcomes evaluated occur either intraoperatively or postoperatively. The primary outcome in the present study is 30-day postoperative mortality. Secondary outcomes include pulmonary embolism, urinary tract infection, acute renal failure, vein thrombosis requiring therapy, intra- or postoperative cardiac arrest requiring cardiopulmonary resuscitation (CPR), pneumonia, intervention therapeutic endoscopy, an operative drain still present at 30 days postoperative, failure to be discharged by day 30, unplanned ICU admission, and reoperation within 30 days. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines [29].

Statistical analysis

χ² tests and analysis of variance were used to compare baseline categorical variables and continuous variables, respectively. Logistic regression was utilized to estimate odds ratios (ORs) of 30-day outcomes according to preoperative serum albumin categories. All models were adjusted for age, sex, race/ethnicity, smoking status within 1 year of surgery, preoperative body mass index (BMI), diabetes, hypertension, hyperlipidemia, obstructive sleep apnea, chronic obstructive pulmonary disease (COPD), percutaneous coronary intervention (PCI), previous cardiac surgery, renal insufficiency, requiring or on dialysis, CPT principal operative procedure, and for revisional/conversional bariatric surgery procedures. Participants with missing covariate information were coded as an additional category for missing.

Because metabolic and bariatric surgery for individuals with BMIs <35.0 kg/m² is restricted to those unable to achieve sustained weight loss or those unable to control co-morbidities with nonsurgical methods [30], we performed a sensitivity analysis excluding individuals with preoperative BMIs <35.0 kg/m². In addition, because revisional/conversional procedures may have additional risks or reasons for the surgery, we performed a sensitivity analysis excluding these types of procedures. Furthermore, because the different procedure types may lead to differing negative outcomes, we performed an analysis stratified by procedure type.

All statistical analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC, USA). The level of statistical significance (α) was set at .05.

Results

The final sample consisted of 633,011 adult patients (85.1% women) with a mean (standard deviation) age of 44.8 (12.0) years. Preoperative serum samples were collected for albumin analysis with a mean (SD) of 17.3 (17.5) days before bariatric surgery. The median preoperative serum albumin concentration was 4.1 g/dL (interquartile range: 3.8–4.3), and the prevalence of hypoalbuminemia was 6.13%. Hypoalbuminemia was more common among non-Hispanic Black patients, women, current smokers, patients with a BMI >45.0 kg/m², patients receiving RYGB and other procedures, patients receiving a revisional bariatric surgery procedure, and patients with co-morbidities other than hyper-lipidemia and obstructive sleep apnea (Table 1).

Table 1.

Demographic characteristics according to serum albumin levels within 90 days prior to bariatric surgery.¹

	Serum albumin (g/dL)
Variables	<3.5	3.5–5.5	P
No. of patients	38792	594219
Mean Age (SD), years	44.3 (11.8)	44.8 (12.0)	<0.0001
Female, %	86.2	79.5	<0.0001
Race/ethnicity, %
Non-Hispanic White	51.4	59.3	<0.0001
Non-Hispanic Black	25.3	15.8
Hispanic	12.3	13.3
Other2	2.5	2.8
Missing	8.6	8.9
Current smoker within one year	9.5	8.5	<0.0001
Pre-op BMI closest to bariatric surgery
<30.0	1.6	1.0	<0.0001
30.0–34.9	2.2	4.2
35.0–39.9	13.6	23.4
40.0–44.9	23.9	30.6
45.0–49.9	21.9	19.8
50.0–54.9	16.1	11.0
55.0–59.9	9.9	5.5
≥60	10.8	4.5
CPT principal operative procedure, %
Sleeve (LSG)	64.5	69.6	<0.0001
RYGB (LRYGB/open RYGB)	31.5	27.2
Band (LAGB)	1.9	1.8
Other3	2.2	1.4
Revisional/conversional procedure, %	10.5	7.9	<0.0001
Diabetes, %	31.2	25.3	<0.0001
Hypertension, %	49.9	47.6	<0.0001
Hyperlipidemia, %	23.1	23.7	0.008
Obstructive sleep apnea, %	35.5	37.2	<0.0001
COPD, %	2.4	1.6	<0.0001
Percutaneous coronary intervention, %	2.6	1.9	<0.0001
Previous cardiac surgery, %	1.4	1.1	<0.0001
Renal insufficiency, %	1.9	0.6	<0.0001
Requiring or on dialysis, %	0.8	0.3	<0.0001

¹Data are shown as mean (SD) or percentages.

²Includes native Hawaiian or other Pacific islander, Asian, and unknown.

³Includes vertical banded gastroplasty, biliopancreatic diversion, biliopancreatic diversion with duodenal switch, and other open gastric restriction surgery.

Abbreviations: BMI, body mass index; COPD, chronic obstructive pulmonary disease; CPT, current procedural terminology; LSG, laparoscopic sleeve gastrectomy; LRYGB, laparoscopic Roux-en-Y gastric bypass; LAGB, laparoscopic adjustable gastric banding.

There was a significant association of hypoalbuminemia with most outcomes evaluated in this population. Compared with normoalbuminemia, the OR (95% confidence interval [CI]) for 30-day mortality after bariatric surgery was 1.42 (1.10, 1.82) for hypoalbuminemia (Table 2). The associations for hypoalbuminemia were significant among all other outcomes except for pulmonary embolism, urinary tract infection, and cardiac arrest requiring CPR. The remaining OR (95% CIs) for hypoalbuminemia ranged from 1.14 (1.05, 1.24) for reoperation within 30 days to 2.32 (1.66, 3.25) for failure to be discharged by day 30.

Table 2.

Associations of pre-operative serum albumin levels with 30-day post-operative outcomes

	Serum albumin levels (g/dL)				Per 1 unit decrease
	<3.5		3.5–5.5
Post-op outcomes	NO. of cases	OR (95% CI)	NO. of cases	OR (95% CI)	OR (95% CI)
Death	73	1.42 (1.10, 1.82)	557	1.0 (Reference)	1.37 (1.13, 1.67)
Pulmonary embolism	68	1.26 (0.98, 1.63)	668	1.0 (Reference)	1.42 (1.18, 1.71)
Urinary tract infection	185	1.15 (0.99, 1.34)	2111	1.0 (Reference)	1.11 (0.99, 1.23)
Acute renal failure	68	1.72 (1.32, 2.42)	402	1.0 (Reference)	1.79 (1.45, 2.21)
Vein thrombosis requiring therapy	129	1.87 (1.55, 2.26)	1028	1.0 (Reference)	1.82 (1.58, 2.10)
Intra- or post-op cardiac arrest requiring CPR	23	1.03 (0.67, 1.60)	243	1.0 (Reference)	1.33 (0.98, 1.79)
Pneumonia	155	1.56 (1.32, 1.85)	1252	1.0 (Reference)	1.49 (1.31, 1.69)
Intervention therapeutic endoscopy	238	1.35 (1.18, 1.55)	2315	1.0 (Reference)	1.33 (1.20, 1.47)
An operative drain still present 30 days post-op	139	1.29 (1.08, 1.54)	1449	1.0 (Reference)	1.31 (1.16, 1.49)
Failure to be discharged by day 30	44	2.32 (1.66, 3.25)	209	1.0 (Reference)	2.76 (2.15, 3.54)
Unplanned ICU admission	460	1.27 (1.15, 1.40)	4352	1.0 (Reference)	1.27 (1.19, 1.37)
Reoperation within 30 days	675	1.14 (1.05, 1.24)	8160	1.0 (Reference)	1.12 (1.06, 1.18)

Data are shown as odds ratios and 95% confidence intervals.

All models were adjusted for age, sex, race/ethnicity, smoking status, BMI, diabetes, hypertension, hyperlipidemia, obstructive sleep apnea, COPD, percutaneous coronary intervention, previous cardiac surgery, renal insufficiency, requiring or on dialysis, and if it was a revisional or conversional procedure.

Several associations were stronger after further stratification of the sample to include severe and marginal hypoalbuminemia. The ORs (95% CIs) for death were 4.76 (1.91, 11.9) for severe and 1.35 (1.04, 1.75) for marginal hypoalbuminemia (Table 3). For severe hypoalbuminemia, the associations for vein thrombosis and intervention endoscopy were attenuated, but these associations remained significant in the marginal hypoalbuminemia group. For all other outcomes with significant associations in the primary analysis, the associations were stronger in the severe hypoalbuminemia group with ORs (95% CI) ranging from 2.28 (1.57, 3.30) for reoperation within 30 days to 16.5 (7.97, 34.3) for failure to be discharged by day 30. Most associations remained significant for the marginal hypoalbuminemia group with ORs (95% CIs) ranging from 1.11 (1.03, 1.21) for reoperation within 30 days to 1.92 (1.33, 2.78) for failure to be discharged by day 30; however, the association for an operative drain still present 30 days after surgery was not significant with an OR (95% CI) of 1.18 (.98, 1.42). Contrary to the primary analysis, the association for urinary tract infection was significant for the marginal hypoalbuminemia group (OR 5 1.17, 95% CI: 1.01, 1.37).

Table 3.

Associations of pre-operative serum albumin levels with 30-day post-operative outcomes stratified to include severe (n = 691) and marginal (n = 38,101) hypoalbuminemia.

	Serum albumin levels (g/dL)
	<2.5		2.5–3.5		3.5–5.5
Post-op outcomes	NO. of cases	OR (95% CI)	NO. of cases	OR (95% CI)	NO. of cases	OR (95% CI)
Death	5	4.76 (1.91, 11.9)	68	1.35 (1.04, 1.75)	557	1.0 (reference)
Pulmonary embolism	3	3.06 (0.98, 9.59)	65	1.23 (0.95, 1.59)	668	1.0 (reference)
Urinary tract infection	0	NA	185	1.17 (1.01, 1.37)	2111	1.0 (reference)
Acute renal failure	4	3.70 (1.31, 10.5)	64	1.67 (1.27, 2.19)	402	1.0 (reference)
Vein thrombosis requiring therapy	3	2.33 (0.75, 7.27)	126	1.86 (1.54, 2.25)	1028	1.0 (reference)
Intra- or post-op cardiac arrest requiring CPR	4	7.54 (2.67, 21.3)	19	0.88 (0.55, 1.41)	243	1.0 (reference)
Pneumonia	13	6.11 (3.47, 10.8)	142	1.46 (1.23, 1.75)	1252	1.0 (reference)
Intervention therapeutic endoscopy	7	1.98 (0.94, 4.21)	231	1.34 (1.17, 1.54)	2315	1.0 (reference)
An operative drain still present 30 days post-op	15	6.13 (3.61, 10.4)	124	1.18 (0.98, 1.42)	1449	1.0 (reference)
Failure to be discharged by day 30	9	16.5 (7.97, 34.3)	35	1.92 (1.33, 2.78)	209	1.0 (reference)
Unplanned ICU admission	22	2.79 (1.79, 4.34)	438	1.24 (1.12, 1.37)	4352	1.0 (reference)
Reoperation within 30 days	31	2.28 (1.57, 3.30)	644	1.11 (1.03, 1.21)	8160	1.0 (reference)

Data are shown as odds ratios and 95% confidence intervals.

All models were adjusted for age, sex, race/ethnicity, smoking status, BMI, diabetes, hypertension, hyperlipidemia, obstructive sleep apnea, COPD, percutaneous coronary intervention, previous cardiac surgery, renal insufficiency, requiring or on dialysis, and if it was a revisional or conversional procedure.

The results of a sensitivity analysis were similar after the exclusion of patients with preoperative BMI <35.0 kg/m² (Supplemental Table 1). In another sensitivity analysis restricted to nonrevisional/conversional procedures, most associations remained significant (Supplemental Table 2). In analyses stratified by procedure type, the association of hypoalbuminemia with death was attenuated and no longer significant in the LSG and for the combined LAGB plus other procedure types group; however, this association remained significant for the RYGB group (OR = 1.72; 95% CI: 1.21, 2.44; Supplemental Table 3). Overall, the associations with secondary outcomes were attenuated owing to reduced sample size, but many associations remained significant across procedure types.

Discussion

In the present study, hypoalbuminemia was associated with several 30-day post–bariatric surgery complications including death, after adjustment for age, sex, race/ethnicity, smoking status, preoperative BMI, baseline co-morbidities, procedure type, and whether the procedure was revisional. In secondary analysis, severe hypoalbuminemia was associated with higher odds of negative outcomes compared with marginal hypoalbuminemia.

Albumin is a well-accepted biomarker of inflammation that may be abnormally low during chronic disease [13]. In the present study, patients with preoperative COPD, PCI, previous cardiac surgery, renal insufficiency, or requiring or on dialysis were more likely to have hypoalbuminemia. The overall prevalence of hypoalbuminemia (defined as serum albumin <3.5 g/dL) in this study was 6.1%, which is slightly lower than the 7.7%–12.5% prevalence rates reported in other studies of bariatric surgery candidates [21,22].

Our study represents the largest to date that examines the association between preoperative serum albumin and 30-day postoperative outcomes after bariatric surgery. Preliminary evidence from previous studies suggests hypoalbuminemia is among the strongest predictors of 30-day mortality among bariatric surgery patients [24] and is strongly associated with risk of additional operations [23]. Consistent with our findings, Husain et al. [25] observed that the OR (95% CI) for the association of hypoalbuminemia with postoperative complications was 7.45 (2.63–21.1) among individuals receiving bariatric surgery. Similarly, a previous analysis of the MBSAQIP PUF found an association between preoperative hypoalbuminemia and increased risk of postoperative gastrointestinal leak with an OR (95% CI) of 1.66 (1.08–2.57) [26].

In secondary analysis in the present study, higher odds of negative bariatric surgery outcomes were observed for severe hypoalbuminemia suggesting a dose-response relationship between preoperative serum albumin levels and postoperative bariatric surgery outcomes. Owing to the small sample size of participants with severe hypoalbuminemia in the present study, these results need to be interpreted with caution; however, a dose-response relationship has been observed for preoperative serum albumin and surgical outcomes among patients with gastric and ovarian cancer [31,32]. Future studies are needed to determine if there is dose-response relationship between preoperative serum albumin and postoperative bariatric surgery outcomes.

While the data from previous studies on the association of hypoalbuminemia and bariatric surgery outcomes are limited, this association is present in several other surgical fields. An analysis of the American College of Surgeons NSQIP PUFs found that preoperative hypoalbuminemia is associated with negative postoperative outcomes for nearly all of the 16 major procedures evaluated [18]. In addition, a retrospective analysis of elderly adults undergoing hip fracture surgical procedures found an adjusted risk ratio (95% CI) of 1.52 (1.37–1.70) for the association of preoperative hypoalbuminemia and 30-day postoperative mortality [15]. Furthermore, hypoalbuminemia has also been shown to be a risk factor for postoperative complications for several oncology procedures [16,17,31,32].

The mechanism by which hypoalbuminemia is associated with higher risk of negative 30-day postoperative bariatric surgery outcomes remains elusive [33]. Early postoperative interventions with intravenous albumin infusions have not improved postoperative outcomes [34,35], suggesting that the increased risk of negative postoperative outcomes is not due to the functions of albumin but rather the underlying pathogenesis of hypoalbuminemia. Serum albumin is negatively associated with C-reactive protein, and albumin concentrations decrease in the inflammatory state for a variety of reasons [33] including synthesis inhibition by proinflammatory cytokines such as tumor necrosis factor and interleukin-1 [36,37]. Thus, underlying chronic inflammation likely drives the association between hypoalbuminemia and negative postoperative bariatric surgery outcomes.

The strengths of this analysis include the use of the large MBSAQIP PUF data set, which contains demographic and clinical information on nearly all metabolic and bariatric surgeries performed in the United States and Canada and allows for the generalization of findings to a broad population of bariatric surgery candidates. In addition, results from sensitivity analyses excluding participants with BMI <35.0 kg/m² or receiving revisional/conversional procedures indicate that the relationship between preoperative hypoalbuminemia and postoperative bariatric surgery outcomes is not majorly influenced by these participants who may require bariatric surgery for additional pathologies. There are also several limitations to the present study. First, the MBSAQIP is a collection of clinical outcomes from over 135 independent institutions; thus, it is possible that site-to-site variation may be present in the data owing to variation in analysis and reporting procedures. However, site-to-site variation is likely minimal owing to the MBSAQIP accreditation procedure for bariatric surgery centers. Second, although we have controlled for a variety of potential confounders based on rich data collected in the MBSAQIP, we could not rule out the possibility of residual confounding by unmeasured confounders such as inflammation, which is a known risk factor for hypoalbuminemia.

Conclusion

In this large retrospective cohort of bariatric surgery patient outcome data, significant associations were observed between hypoalbuminemia and 30-day postoperative complications including death. Secondary analysis suggests that severe hypoalbuminemia is associated with higher odds of negative outcomes. These findings are consistent with previous studies and suggest that serum albumin may be a useful biomarker to screen for negative bariatric surgery outcomes.