Abstract
Summary
Patients with low-energy hip fractures do not follow the obesity paradox as previously reported. In datasets where injury mechanism is not available, the use of age >50 years (as opposed to commonly used >65 years) as a surrogate for a low-energy hip fracture patients may be a more robust inclusion criterion.
Purpose
In elderly patients with a hip fracture, limited data suggests that obese patients counterintuitively have improved survival compared to normal-weight patients. This "obesity paradox" may be the byproduct of selection bias. We hypothesized that the obesity paradox would not apply to elderly hip fracture patients.
Methods
The National Surgical Quality Improvement Project dataset identified 71,685 hip fracture patients ≥50 years-of-age with complete body mass index (BMI) data that underwent surgery. Patients were stratified into under and over 75-year-old cohorts (n=18,956 and 52,729, respectively). Within each age group, patients were stratified by BMI class and compared with respect to preoperative characteristics and 30-day mortality. Significant univariate characteristics (p<0.1) were included in multivariate analysis to determine the independent effect of obesity class on 30-day mortality (p<0.05).
Results
Multivariate analysis of <75-year-old patients with class-III obesity were more likely to die within 30-days than similarly aged normal-weight patients (OR 1.91, CI 1.06-3.42, p=0.030). Multivariate analysis of ≥75-year-old overweight (OR 0.69, CI 0.62-0.77, p<0.001), class-I obese (OR 0.62, CI 0.51–0.74, p<0.001), or class-II obese (OR=0.69, CI 0.50–0.95, p=0.022) patients were less likely to die within 30-days when compared to similarly aged normal-weight patients.
Conclusions
Our data suggest that obesity is a risk factor for mortality in low-energy hip fracture patients, but the appearance of the "obesity paradox" in elderly hip fracture patients results from statistical bias that is only evident upon subgroup analysis.
Keywords: Obesity, Hip fracture, Obesity paradox, Mortality, Risk factor
Introduction
After a hip fracture, patients with advanced age, male gender, diabetes, chronic renal disease, and low preoperative functional status are at increased risk of perioperative morbidity and mortality.[1-4] Interestingly, obesity also reduces the risk of early perioperative mortality after a hip fracture compared to normal-weight patients.[5] This protective effect of obesity on mortality is despite the well-known association between obesity and its other closely associated comorbid diseases with mortality in the general population. [6-8] Hence, obesity conferring a survival advantage in hip fractures is termed the "obesity paradox." [9]
This paradox is also well described in congestive heart failure, renal failure, and diabetes. [5, 10, 11] However, multiple reports challenge the statistical and clinical validity of the obesity paradox. [12, 13] The obesity paradox is termed a paradox because it defies intuition, which would conclude that mortality would be elevated in obese patients due to the well-known health risks of increased body mass. The prevailing explanation of the paradox is selection bias as large population-based studies demonstrate that younger obese patients are at elevated risk of mortality compared to normal-weight patients. [12, 14, 15] Thus, in a dataset with a skewed age distribution, the obesity paradox may be explained by reverse-causation or survival bias. The sickest obese patients die before sustaining a hip fracture or at a relatively younger chronologic age excluding them from meaningful inclusion in whole population analysis. [16, 17] This concentrates the healthier individuals that survive the selection event and overlooks those that did not, leading to potentially erroneous conclusions.
We sought to understand the etiology of the obesity paradox in elderly patients that have surgery for a hip fracture. [5, 14, 15, 18] We hypothesized that the relatively younger, obese patients (50-to-74 years-of-age) will not demonstrate the obesity paradox and will be at an increased perioperative mortality risk compared to similarly aged normal-weight patients.
Methods
The 2015–2018 National Surgical Quality Improvement Project (NSQIP) database was utilized for this study. The database contains prospectively collected preoperative and 30-day postoperative outcomes for randomly selected patients. Data is entered by trained reviewers and has 98% interobserver reliability.[19] The NSQIP was queried for patients undergoing fixation (plate or intramedullary nail) or arthroplasty (hemi or total hip arthroplasty) (CPT 27125, 27130, 27236, 27244, and 27245) for acute, closed, nononcologic intracapsular or peritrochanteric hip fracture (Appendix 1). Patients were excluded if there was incomplete information regarding body mass index (BMI) or <50 years-of-age to identify those sustaining elderly, osteoporotic type, low-energy hip fractures (Appendix 2). [20, 21] Patients were stratified based on BMI into categories: underweight (<18.5 kg/m2), normal weight (18.5–24.9 kg/m2), overweight (25–29.9 kg/m2), class-I obese (30–34.9 kg/m2), class-II obese (35–39.9 kg/m2), class-III obese (40–49.9 kg/m2), or super obese (≥50 kg/m2). [22, 23] The latter classification was performed given that super obesity may be a separate clinical entity with a different underlying etiology than other forms of obesity.[24]
Given the significant effect of age on mortality, an inflection point in the obesity paradox at 75 years-of-age, and the possibility of survival bias to explain the obesity paradox in large population-based health studies, the analysis was performed independently for patients <75 and ≥75 years-of-age.[25-27] Within each age category, patients were stratified based on BMI classification and compared based on 15 preoperative characteristics: male gender, tobacco use, hypertension requiring medication, bleeding disorder, dependent functional status for activities of daily living, age, congestive heart failure, dialysis dependence, insulin-dependent diabetes mellitus, dyspnea with exertion, preoperative hematocrit, albumin <3.0 g/dL, chronic obstructive pulmonary disease, transfusion, and surgical procedure. [28-30] Univariate analysis of each BMI category revealed the significant risk factors for 30-day mortality (p<0.1). A multivariate analysis of the significant mortality-related covariates for each of the age groups was performed to determine the effect of obesity class on perioperative mortality. The mean-variance inflation factors were 4.30 and 4.98 for the <75 years-of-age and ≥75 years-of-age cohorts, respectively. Significance was defined as a p-value of <0.05 on multivariate analysis.
Continuous variables are reported as mean and standard deviation and categorical variables as number and person. Comparisons were conducted in univariate analysis via two-sided t-tests or analysis of variance, or chi-square tests, respectively. Results from multivariate logistic regression analysis are reported as odds ratio (OR) and 95% confidence interval (CI). All statistical computing was performed using Stata 15 (StataCorp.2017. College Station, TX: StataCorp LLC).
Results
There were 18,956 patients <75 years-of-age with hip fractures (Table 1). Their mean age was 66.2 ± 6.1 years, and BMI was 26.1 ± 6.5kg/m2. The 30-day mortality was 2.1% (n=405). On univariate analysis, the incidence of all preoperative characteristics and 30-day mortality was significantly different between the BMI classes (p < 0.001). Compared to normal-weight patients, obese patients had a higher incidence of diabetes, dyspnea on exertion, limited functional status, chronic obstructive pulmonary disease, congestive heart failure, hypertension, dialysis dependence, bleeding disorders, and intramedullary fixation (p < 0.001). On multivariate analysis, patients with class-III obesity were 91% more likely to die within 30 days of surgery than similarly aged normal-weight patients (OR 1.91, CI 1.06–3.42, p=0.030, Table 2). Additionally, when compared to similarly aged patients, class-III obese patients were more likely to die within 30-days when compared to those patients that were overweight (OR 1.82, CI 0.30–0.99, p=0.045), class-I obese (OR 2.09, CI 0.25–0.90, p=0.023), or class-II obese (OR 2.42, CI 0.19–0.89, p=0.024).
Table 1.
Univariate analysis of surgically relevant characteristics in the under 75-year-old patients that underwent surgery for a hip fracture stratified by body mass index
| Category | Body Mass Index (kg/m2) |
p-value | ||||||
|---|---|---|---|---|---|---|---|---|
| <18.5 | 18.5-24.9 | 25-29.9 | 30-34.9 | 35-39.9 | 40-49.9 | ≥50 | ||
| Number of Patients, n (%) | 1505 (7.9) | 7751 (40.9) | 5447 (28.7) | 2613 (13.8) | 1002 (5.3) | 510 (2.7) | 128 (0.7) | |
| Age (mean ± deviation) | 65.5 (6.2) | 65.9 (6.3) | 66.6 (6.0) | 66.6 (5.9) | 66.3 (5.9) | 65.4 (6.0) | 64.3 (6.3) | <0.001 |
| Male, n (%) | 404 (26.8) | 2847 (36.7) | 2336 (42.9) | 1060 (40.6) | 324 (32.3) | 143 (28.0) | 26 (20.3) | <0.001 |
| Diabetes, n (%) | 82 (5.5) | 656 (8.5) | 710 (13.0) | 549 (21.0) | 272 (27.2) | 145 (28.4) | 34 (26.6) | <0.001 |
| Smoking, n (%) | 795 (52.8) | 2727 (35.2) | 1248 (22.9) | 500 (19.1) | 160 (16.0) | 87 (17.1) | 28 (21.9) | <0.001 |
| Dyspnea, Exertion, n (%) | 142 (9.4) | 463 (6.0) | 319 (5.9) | 180 (6.9) | 86 (8.6) | 50 (9.8) | 21 (16.4) | <0.001 |
| Dyspnea, Rest, n (%) | 39 (2.6) | 116 (1.5) | 50 (0.9) | 34 (1.3) | 18 (1.8) | 14 (2.8) | 2 (1.6) | |
| Limited Function, n (%) | 172 (11.6) | 817 (10.6) | 534 (9.9) | 238 (9.2) | 113 (11.4) | 57 (11.3) | 20 (15.6) | <0.001 |
| Dependent Function, n (%) | 29 (2.0) | 168 (2.2) | 78 (1.4) | 34 (1.3) | 15 (1.5) | 9 (1.8) | 0 (0) | |
| Missing Function, n (%) | 16 (1.1) | 53 (0.7) | 36 (0.7) | 14 (0.5) | 9 (0.9) | 6 (1.2) | 0 (0) | 0.285 |
| Chronic Pulmonary Disease, n (%) | 382 (25.4) | 1086 (14.0) | 575 (10.6) | 308 (11.8) | 157 (15.7) | 90 (17.7) | 24 (18.8) | <0.001 |
| Fieart Failure, n (%) | 27 (1.8) | 134 (1.7) | 121 (2.2) | 81 (3.1) | 55 (5.5) | 31 (6.1) | 5 (3.9) | <0.001 |
| Hypertension, n (%) | 612 (40.7) | 3726 (48.1) | 3231 (59.3) | 1775 (67.9) | 737 (73.6) | 389 (76.2) | 98 (76.6) | <0.001 |
| Hemodialysis, n (%) | 27 (1.8) | 245 (3.2) | 207 (3.8) | 96 (3.7) | 46 (4.6) | 19 (3.7) | 4 (3.1) | <0.001 |
| Bleeding Disorder, n (%) | 142 (9.4) | 852 (11.0) | 736 (13.5) | 409 (15.7) | 179 (17.9) | 94 (18.4) | 13 (10.2) | <0.001 |
| Transfusion, n (%) | 78 (5.2) | 256 (3.3) | 123 (2.3) | 61 (2.3) | 29 (2.9) | 22 (4.3) | 2 (1.6) | <0.001 |
| Hematocrit (mean ± deviation) | 34.3 ± 5.7 | 35.7 ± 5.4 | 36.6 ± 5.4 | 36.8 ± 5.4 | 36.5 ± 5.4 | 36.0 ± 5.8 | 36.1 ±5.3 | <0.001 |
| Hypoalbuminemia, n (%) | 283 (26.6) | 836 (16.4) | 513 (14.4) | 228 (13.4) | 95 (14.4) | 55 (16.1) | 17 (20.0) | <0.001 |
| Missing Data, n (%) | 441 (29.3) | 2650 (34.2) | 1894 (34.8) | 912 (34.9) | 342 (34.1) | 168 (32.9) | 43 (33.6) | 0.006 |
| Non-Femoral Neck Fracture, n (%) | 712 (47.3) | 3093 (39.9) | 2141 (39.3) | 1149 (44.0) | 498 (49.7) | 275 (53.9) | 69 (53.9) | <0.001 |
| Death within 30 days, n (%) | 55 (3.7) | 165 (2.1) | 97 (1.8) | 48 (1.8) | 20 (2.0) | 19 (3.7) | 1 (0.8) | <0.001 |
Table 2.
Odds of mortality within 30 days of surgery for a hip fracture stratified by body mass index after multivariate regression
|
Body Mass
Index |
<75 years-of-age Odds Ratio (Confidence Interval) |
p-
value |
≥75 years-of-age Odds Ratio (Confidence Interval) |
p-
value |
|---|---|---|---|---|
| <18.5 kg/m2 | 1.54 (1.06-2.24) | 0.023 | 1.35 (1.17-1.55) | <0.001 |
| 18.5 −24.9 kg/m2 | Reference | Reference | ||
| 25 - 29.9 kg/m2 | 1.05 (0.78-1.41) | 0.767 | 0.69 (0.62-0.77) | <0.001 |
| 30 - 34.9 kg/m2 | 0.91 (0.61-1.37) | 0.663 | 0.62 (0.51-0.74) | <0.001 |
| 35 - 39.9 kg/m2 | 0.79 (0.43-1.44) | 0.437 | 0.69 (0.50-0.95) | 0.022 |
| 40 - 49.9 kg/m2 | 1.91 (1.06-3.42) | 0.030 | 0.65 (0.38-1.13) | 0.126 |
| ≥50 kg/m2 | 0.39 (0.05-2.98) | 0.364 | 1.15 (0.51-2.60) | 0.729 |
There were 52,729 patients ≥75 years-of-age (Table 3). Compared to the <75-year-old cohort, the ≥75 years-of-age cohort had a significantly lower mean BMI (24.5 ± 5.1 kg/m2, p<0.001). The 30-day mortality was 6.3% (n=3,330). Univariate analysis revealed that the incidence of all characteristics and 30-day mortality was significantly different between the BMI classes (p < 0.001). Compared to normal-weight patients, obese patients had a higher incidence of diabetes, dyspnea on exertion, chronic obstructive pulmonary disease, congestive heart failure, hypertension, dialysis dependence, bleeding disorders, and use of intramedullary fixation (p < 0.001). On multivariate analysis, overweight (OR 0.69, CI 0.62–0.77, p < 0.001, Table 2), class-I obese (OR 0.62, CI 0.51–0.74, p < 0.001, Table 2), or class-II obese (OR=0.69, CI 0.50–0.95, p=0.022) patients were less likely to die within 30-days of surgery for a hip fracture when compared to similarly aged normal-weight patients. Class-III (OR=0.65, CI 0.38–1.13, p = 0.126), and super (OR=1.15, CI 0.51–2.60, p = 0.729), obesity did not confer any survival advantage when compared to similarly aged normal-weight patients.
Table 3.
Univariate analysis of surgically relevant characteristics in the over 75-year-old patients that underwent surgery for a hip fracture stratified by body mass index
| Category | Body Mass Index (kg/m2) |
p-value | ||||||
|---|---|---|---|---|---|---|---|---|
| <18.5 | 18.5-24.9 | 25-29.9 | 30-34.9 | 35-39.9 | 40-49.9 | ≥50 | ||
| Number of Patients, n (%) | 4743 (9.0) | 26620 (50.5) | 14777 (28.0) | 4846 (9.2) | 1208 (2.3) | 411 (0.8) | 124 (0.2) | |
| Age (mean ± deviation) | 85.4 ± 4.7 | 85.4 ± 4.7 | 84.5 ± 4.8 | 83.5 ± 4.9 | 82.6 ± 5.0 | 82.2 ± 5.0 | 84.5 ± 4.6 | <0.001 |
| Male, n (%) | 717 (15.1) | 7025 (26.4) | 5119 (34.6) | 1489 (30.7) | 327 (27.1) | 77 (18.7) | 44 (35.5) | <0.001 |
| Diabetes, n (%) | 117 (2.5) | 1135 (4.3) | 1139 (7.7) | 579 (12.0) | 218 (18.1) | 81 (19.7) | 12 (9.7) | <0.001 |
| Smoking, n (%) | 582 (12.3) | 1752 (6.6) | 678 (4.6) | 230 (4.8) | 55 (4.6) | 12 (2.9) | 9 (7.3) | <0.001 |
| Dyspnea, Exertion, n (%) | 306 (6.5) | 1532 (5.8) | 943 (6.4) | 355 (7.3) | 121 (10.0) | 49 (11.9) | 17 (13.7) | <0.001 |
| Dyspnea, Rest, n (%) | 72 (1.5) | 242 (0.9) | 136 (0.9) | 65 (1.3) | 28 (2.3) | 5 (1.2) | 2 (1.2) | |
| Limited Function, n (%) | 1132 (24.0) | 5517 (20.9) | 2706 (18.5) | 829 (17.2) | 210 (17.6) | 83 (20) | 21 (17.1) | <0.001 |
| Dependent Function, n (%) | 262 (5.6) | 1034 (3.9) | 436 (3.0) | 139 (2.9) | 27 (2.3) | 15 (3.7) | 4 (3.3) | |
| Missing Function, n (%) | 33 (0.7) | 224 (0.8) | 119 (0.8) | 38 (0.8) | 14 (1.2) | 3 (0.7) | 1 (0.8) | 0.837 |
| Chronic Pulmonary Disease, n (%) | 729 (15.4) | 2646 (9.9) | 1346 (9.1) | 516 (10.7) | 173 (14.3) | 55 (13.4) | 14 (11.3) | <0.001 |
| Heart Failure, n (%) | 165 (3.5) | 980 (3.7) | 639 (4.3) | 279 (5.8) | 76 (6.3) | 28 (6.8) | 5 (4.0) | <0.001 |
| Hypertension, n (%) | 2933 (61.8) | 18175 (68.3) | 11027 (74.6) | 3889 (80.3) | 995 (82.4) | 328 (79.8) | 103 (83.1) | <0.001 |
| Hemodialysis, n (%) | 4743 (0.9) | 333 (1.3) | 242 (1.6) | 71 (1.5) | 20 (1.7) | 7 (1.7) | 3 (2.4) | 0.001 |
| Bleeding Disorder, n (%) | 557 (11.7) | 4222 (15.9) | 2875 (19.5) | 1066 (22.0) | 284 (23.5) | 97 (23.6) | 24 (19.4) | <0.001 |
| Transfusion, n (%) | 255 (5.4) | 1167 (4.4) | 555 (3.8) | 202 (4.2) | 57 (4.7) | 16 (3.9) | 6 (4.8) | <0.001 |
| Hematocrit (mean ± deviation) | 33.9 ± 5.2 | 34.7 ± 5.2 | 35.5 ± 5.3 | 35.5 ± 5.3 | 35.6 ± 5.4 | 35.3 ± 5.4 | 35.2 ± 4.4 | <0.001 |
| Hypoalbunrinenria, n (%) | 624 (19.4) | 2355 (13.7) | 1200 (12.7) | 409 (13.2) | 106 (13.6) | 39 (14.5) | 11 (13.8) | <0.001 |
| Missing Data, n (%) | 1522 (32.1) | 9435 (35.4) | 5312 (36.0) | 1743 (36.0) | 428 (35.4) | 142 (35.6) | 44 (35.5) | <0.001 |
| Non-Fenroral Neck Fracture, n (%) | 2133 (45.0) | 11273 (42.4) | 6340 (42.9) | 2284 (47.1) | 618 (51.2) | 222 (54.0) | 57 (46.0) | <0.001 |
| Death within 30 days, n (%) | 439 (9.3) | 1831 (6.9) | 746 (5.1) | 222 (4.6) | 62 (5.1) | 19 (4.6) | 11 (8.9) | <0.001 |
Given that class-III obesity was only a risk factor for mortality in relatively younger patients (<75 years-of-age), the patients with class-III obesity of each cohort were compared to determine whether this disparate effect may be the product of a survival bias (Table 4). Indeed, the patients <75 years-of-age with class-III obesity were sicker than the class-III obese patients ≥75-years-of-age. This is demonstrated by a higher incidence of comorbidities known to be associated with mortality in patients with hip fracture, including male sex (28.0%), insulin-dependent diabetes mellitus (28.4%), and active smoking (17.1%) in class-III obese patients <75 years-of-age when compared to class-III obese patients >75 years-of-age (male: 18.7%, p<0.001; diabetes: 19.7%, p<0.001; smoking 2.9% p<0.001). The patients >75 years-of-age with class-III obesity, as expected with advancing age, had a significantly higher incidence of dependent functional status (3.7%) compared to the patients that were <75 years-of-age with class-III obesity (dependent functional status 1.8%, p<0.001). The absolute risk of 30-day mortality between the patients under and over 75 years-of-age with class-III obesity was not significantly different (3.7% v. 4.6%, p=0.497).
Table 4.
Profile of patients with Class III obesity undergoing surgery for a hip fracture
| Category | <75 years | ≥75 years | p-value |
|---|---|---|---|
| Age (mean ± deviation) | 65.4 ± 6.0 | 82.2 ±5.0 | <0.001 |
| Male (%) | 143 (28.0) | 77 (18.7) | <0.001 |
| Diabetes (%) | 145 (28.4) | 81 (19.7) | <0.001 |
| Smoking (%) | 87 (17.1) | 12 (2.9) | <0.001 |
| Dyspnea, Exertion (%) | 50 (9.8) | 49 (11.9) | 0.16 |
| Dyspnea, Rest (%) | 14 (2.8) | 5 (1.2) | |
| Limited Function (%) | 57 (11.3) | 83 (20.3) | <0.001 |
| Dependent Function (%) | 9 (1.8) | 15 (3.7) | |
| Missing Function (%) | 6 (1.2) | 3 (0.7) | 0.488 |
| Chronic Pulmonary Disease (%) | 90 (17.7) | 55 (13.4) | 0.076 |
| Heart Failure (%) | 31 (6.1) | 28 (6.8) | 0.652 |
| Hypertension (%) | 389 (76.3) | 328 (79.8) | 0.199 |
| Hemodialysis (%) | 19 (3.7) | 7 (1.7) | 0.059 |
| Bleeding Disorder (%) | 94 (18.4) | 97 (23.6) | 0.055 |
| Transfusion (%) | 22 (4.3) | 16 (3.9) | 0.749 |
| Hematocrit (mean ± deviation) | 36.0 ± 5.8 | 35.3 ± 5.4 | 0.072 |
| Hypoalbuminemia (%) | 55 (16.1) | 39 (14.5) | 0.590 |
| Missing Data (%) | 168 (32.9) | 142 (34.6) | 0.608 |
| Non-Femoral Neck Fracture (%) | 275 (54) | 222 (54) | 0.002 |
| Body Mass Index (mean ± deviation) | 43.5 ± 2.7 | 43.3 ± 2.6 | 0.196 |
| Death within 30 days (%) | 19 (3.7) | 19 (4.6) | 0.497 |
Discussion
The obesity paradox is not universally true for all low-energy hip fractures. In relatively younger hip fracture patients (50-to-74 years-of-age), morbid obesity is a risk factor for early perioperative mortality. Additionally, class-III obese patients <75 years-of-age demonstrated a statistically significant increase in 30-day mortality risk compared to similarly aged, normal-weight, overweight, class-I obese, or class-II obese patients. The obesity paradox is only true for a select group of patients who are ≥75 years-of-age and either overweight, class-I, or class-II obese.
While our data may demonstrate that the prior protective appearance of obesity on mortality in all low-energy hip fracture patients is not true, there are multiple possible explanations for the appearance of the obesity paradox, including survivor bias. [5] Obesity alone is a risk factor for early mortality in the absence of a hip fracture. [31] Moreover, the conditions that commonly co-exist with obesity, including diabetes, heart failure, and renal disease, are also independent risk factors for mortality. [8, 32, 33] As such, patients with obesity are at risk of dying earlier than normal or overweight patients. [31] This mortality difference is demonstrated in our data by the greater incidence of severe comorbidities in the relatively younger (50-to-74 years-of-age) patients when compared to the older patients (≥75 years-of-age) with class-III obesity (Table 4). Additionally, a low-energy hip fracture may be associated with advanced physiologic age due to impaired cognitive or functional status. [34] While most individuals have equivalent chronologic and physiologic age, patients with obesity or other closely associated comorbid conditions may be chronologically younger than their physiologic age. Therefore, the sickest patients with class-III obesity are either more likely to die before sustaining a hip fracture or sustain hip fractures at a relatively earlier chronological age. Thus, this group of chronologically young but physiologically old obese patients is likely excluded from meaningful analysis hip fracture studies where >65 years-of-age is used to select low-energy hip fracture patients. [5]
Survival bias as an explanation for the obesity paradox in this population is supported in our data by the distribution of comorbidities seen among the BMI categories. For example, in the ≥75 years-of-age cohort, the rates of comorbid conditions independently associated with mortality, such as smoking, dependent functional status, and preoperative transfusion, were paradoxically higher in the normal-weight population compared to obese patients. [4, 29, 34] Hence, patients with obesity that survived past 75 years-of-age likely represent a healthier cohort than the entire population sustaining a hip fracture. The skewed distribution of the age data in our study (73.6% of the population >75 years-of-age) allows relatively more healthy, older, patients with obesity to dominate the analysis and therefore derive a seemingly protective effect of obesity on mortality for the entire population, where this only applies to patients ≥75-years-of-age. Our study's age distribution coincides with the rightward shift in the age of patients sustaining a hip fracture. [35]
Prior studies evaluating hip fractures included patients only over 65 years-of-age. [5] However, data suggests that the incidence of low-energy hip fractures double per decade after 50 year-of-age. [5, 36-40] Additionally, our data suggests the use of chronological age equal to physiologic age may result in a selection bias. The exclusion criteria of Modig et al. allowed only the relatively healthier obese patients to be studied and consequently demonstrated an obesity-related survival advantage despite the sobering statistics related to obesity and death occurring independently of hip fracture. [5] The population in Modig et al. approximates our over 75 years-of-age years population. Additionally, Table 4 also demonstrates that the young versus elderly obese population have significantly different physiologic profiles. [5] Thus, prior finding realted to the obesity paradox in hip fractures when analyzed as a single aggregate population may be an example of Simpson's paradox. Simpson's paradox is present when the subgroup analysis outcomes are completely reversed when the subgroups are combined and analyzed as a whole. In this case, when patients <75 years-of-age are analyzed alone, obesity is a risk factor for mortality. When the patients <75 years-of-age are included with those ≥75 years-of-age for whole population analysis, the outcome is wholly reversed, and obesity is seemingly protective against mortality for all patients. This complete reversal in outcome is due to the inability of whole population analysis to account for specific confounding variables addressed with subgroup analysis. In this instance, the whole population analysis does not account for the intricate relationship of age and obesity on mortality in patients with a hip fracture.
Our observations are also supported by the fact that obese patients undergoing all-cause orthopaedic surgery do not follow the paradox.[20, 41] Hence, our study derives a second important finding in addition to casting doubt on the obesity paradox holding univeral truth for all hip fractures. The ubiquitous use of age ≥65-years-of-age as a surrogate in large datasets and registries to define the low-energy hip fractures in the elderly population may result in substantial selection bias and false associations. Where injury mechanism (low versus high energy) is not included, and age is used as a surrogate marker, age >50-years may be a more robust and appropriate limit to define this low-energy fracture population, since low-energy type hip fractures begin to double in incidence after 50 years-of-age. The presumption of chronologic age equaling physiologic age obscures findings where physiologic differences are more significant than chronologic age – one such example is obesity. [42, 43]
Our data also demonstrates that underweight patients are at significantly elevated risk of mortality across the age spectrum compared to normal-weight patients (Table 2). Though our data does not specifically address frailty as a comorbidity, which is a known risk factor for mortality, this health parameter may indicate frailty in this population. These outcomes are consistent with prior published data on underweight patients being at increased risk for mortality in orthopedic surgery.[5, 20, 23] The association of BMI extremes, both low and high, with mortality in this population further highlights the clinically relevant nature of this parameter in patient outcomes following hip fracture surgery.[20, 23]
There are several limitations to our study. First, our study is retrospective and is unable to report causality or lack of association. Second, our mortality outcomes are limited to 30-days post-operatively. Third, the NSQIP dataset only accounts for those patients undergoing surgery and does not account for the small proportion of patients too sick to undergo surgery for a hip fracture. Fourth, our registry data may be subject to coding inaccuracy and missing information. However, data in the NSQIP dataset is entered by trained reviewers with interobserver reliability >98%. Additionally, less than 11.3% of the population was excluded for incomplete BMI data. Our use of list-wise deletion produces the most accurate estimates in the NSQIP dataset. [44] Fifth, our data does not account specifically for frailty indices, which are associated with mortality in this population. Finally, our analysis of patients <75 years-of-age does not demonstrate an effect of class-I or II obesity on mortality relative to normal-weight patients. This finding may be expected in hip fracture patients due to substantial comorbidity, where BMI likely has a threshold effect on mortality instead of a continuous gradient of risk.
Conclusion
The obesity paradox is not universally true for all low-energy hip fractures patients. In relatively younger patients that have surgery for a hip fracture, class-III obesity is an independent risk factor for mortality. In older patients, the obesity paradox appears but may be the result of survivor bias. Our data suggest that obesity should not be viewed as a clinically protective risk factor in the setting of 30-day outcomes following hip fracture surgery. Further study is necessary to identify the longer-term associations between BMI and mortality. Our finding coincides with other scientific data suggesting that obesity is a risk factor for poor perioperative outcomes.
Supplementary Material
Funding
Dr. Amanatullah is funded by NIH-NCATS (KL2TR003143) and Orthopaedic Research and Education Foundation (OREF) grants.
Footnotes
Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s00198-021-06046-7.
Ethics approval This study utilizes de-identified patient information and does not require Institutional Review Board approval.
Conflicts of interest Raj Amin, Micheal Raad, Sandesh Rao, Farah Musharbash, Matthew Best and Derek Amanatullah declare that they have no conflicts of interest.
Data availability
NSQIP is publicly available to participating institutions. The dataset for this manuscript is available for review.
References
- 1.Guzon-Illescas O, Perez Fernandez E, Crespí Villarias N, Quirós Donate FJ, Peña M, Alonso-Blas C, García-Vadillo A, Mazzucchelli R (2019) Mortality after osteoporotic hip fracture: incidence, trends, and associated factors. J Orthop Surg Res 14(1 ):l–9. 10.1186/s13018-019-1226-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Sheehan KJ, Sobolev B, Guy P (2017) Mortality By Timing of Hip Fracture Surgery. J Bone Jt Surg Am Vol 99(106): 1–9. 10.2106/JBJS.F.01426 [DOI] [PubMed] [Google Scholar]
- 3.Frisch NB, Wessell N, Jildeh TR, Greenstein A, Trent Guthrie S (2018) Early-Stage Chronic Kidney Disease and Hip Fracture Mortality. J Surg Orthop Adv 27(3):226–230 [Online], Available: http://europepmc.org/abstract/MED/30489248 [PubMed] [Google Scholar]
- 4.Nkanang B, Parker M, Parker E, Griffiths R (2017) Perioperative mortality for patients with a hip fracture. Injury 48(10):2180–2183. 10.1016/j.injury.2017.07.007 [DOI] [PubMed] [Google Scholar]
- 5.Modig K, Erdefelt A, Mellner C, Cederholm T, Talback M, Hedstrom M (2019) ‘Obesity Paradox’ Holds True for Patients with Hip Fracture. J Bone Jt Surg Am Vol 101 (A):888–895 [DOI] [PubMed] [Google Scholar]
- 6.Hu FB et al. (2001) Diet, Lifestyle and the Risk of Type 2 Diabetes Mellitus in Women. New 345(11 ):790–797 [DOI] [PubMed] [Google Scholar]
- 7.Câmara NOS, Iseki K, Kramer H, Liu ZH, Sharma K (2017) Kidney disease and obesity: epidemiology, mechanisms and treatment. Nat Rev Nephrol 13(3):181–190. 10.1038/nrneph.2016.191 [DOI] [PubMed] [Google Scholar]
- 8.Van Gaal LF, Mertens IL, De Block CE (2006) Mechanisms linking obesity with cardiovascular disease. Nature 444:875–880 [DOI] [PubMed] [Google Scholar]
- 9.Hainer V, Aldhoon-Hainerová I (2013) Obesity paradox does exist. Diabetes Care 36(SUPPL.2). 10.2337/dcS13-2023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Gupta PP, Fonarow GC, Horwich TB (2015) Obesity and the Obesity Paradox in Heart Failure. Can J Cardiol 31 (2):195–202. 10.1016/j.cjca.2014.08.004 [DOI] [PubMed] [Google Scholar]
- 11.Vashistha T, Mehrotra R, Park J, Streja E, Dukkipati R, Nissenson AR, Ma JZ, Kovesdy CP, Kalantar-Zadeh K (2014) Effect of age and dialysis vintage on obesity paradox in long-term hemodialysis patients. Am J Kidney Dis 63(4):612–622. 10.1053/j.ajkd.2013.07.021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Banack HR, Kaufman JS (2013) The ‘obesity paradox’ explained. Epidemiology 24(3):461–462. 10.1097/EDE.0b013e31828c776c [DOI] [PubMed] [Google Scholar]
- 13.Stovitz SD, Banack HR, Kaufman JS (2018) Structural Bias in Studies of Cardiovascular Disease: Let’s Not Be Fooled by the ‘Obesity Paradox. Can J Cardiol 34(5):540–542. 10.1016/j.cjca.2017.10.025 [DOI] [PubMed] [Google Scholar]
- 14.Franks PW, Hanson RL, Knowler WC, Sievers ML, Bennet PH, Looker HC (2010) Childhood obesity, other cardiovascular risk factors, and premature death. N Engl J Med 362(6):485–493. 10.1056/NEJMoa0904130.Childhood [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Twig G, Yaniv G, Levine H, Leiba A, Goldberger N, Derazne E, Ben-Ami Shor D, Tzur D, Afek A, Shamiss A, Haklai Z, Kark JD (2016) Body-mass index in 2.3 million adolescents and cardiovascular death in adulthood. N Engl J Med 374(25):2430–2440. 10.1056/NEJMoa1503840 [DOI] [PubMed] [Google Scholar]
- 16.Tobias DK, Pan A, Jackson CL, O'Reilly EJ, Ding EL, Willett WC, Manson JAE, Hu FB (2014) Body-Mass Index and Mortality among Adults with Incident Type 2 Diabetes. N Engl J Med 370(3):233–244 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Di Angelantonio E et al. (2016) Body-mass index and all-cause mortality: individual-participant-data meta-analysis of 239 prospective studies in four continents. Lancet 388(10046):776–786. 10.1016/S0140-6736(16)30175-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Kusnezov N, Bader J, Blair JA (2017) Predictors of inpatient mortality and systemic complications in acetabular fractures requiring operative treatment. Orthopedics 40(2):e223–e228. 10.3928/01477447-20161202-03 [DOI] [PubMed] [Google Scholar]
- 19.Shiloach M, Frencher SK Jr, Steeger JE, Rowell KS, Bartzokis K, Tomeh MG, Richards KE, Ko CY Hall BL (2010) Toward Robust Information: Data Quality and Inter-Rater Reliability in the American College of Surgeons National Surgical Quality Improvement Program. J Am Coll Surg 210(1):6–16. 10.1016/j.jamcollsurg.2009.09.031 [DOI] [PubMed] [Google Scholar]
- 20.Zhang JC, Matelski J, Gandhi R, Jackson T, Urbach D, Cram P (2018) Can patient selection explain the obesity paradox in orthopaedic hip surgery? an analysis of the acs-nsqip registry. Clin Orthop Relat Res 476(5):964–973. 10.1007/s11999.0000000000000218 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Zuckerman JD (1996) Hip Fracture. N Engl J Med 334(23):1519–1525 [DOI] [PubMed] [Google Scholar]
- 22.Augustin T, Moslirn MA, Brethauer S, Aminian A, Kroh M, Schneider E, Walsh RM (2017) Obesity and its implications for morbidity and mortality after cholecystectomy: a matched NSQIP analysis. Am J Surg 213(3):539–543. 10.1016/j.amjsurg.2016.11.037 [DOI] [PubMed] [Google Scholar]
- 23.Ottesen TD, Malpani R, Galivanche AR, Zogg CK, Varthi AG, Grauer JN (2020) Underweight patients are at just as much risk as super morbidly obese patients when undergoing anterior cervical spine surgery. Spine J 20(7):1085–1095. 10.1016/j.spinee.2020.03.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Froguel P, Blakemore AIF (2008) The power of the extreme in elucidating obesity. N Engl J Med 359(9):891–893. 10.1056/NEJMp0805396 [DOI] [PubMed] [Google Scholar]
- 25.Skinner JS, Abel WM, McCoy K, Wilkins CH (2017) Exploring the ‘Obesity Paradox’ As a Correlate of Cognitive and Physical Function in Community-Dwelling Black and White Older Adults. Cardiovasc Dis Risk Factors 27(4):387–394. 10.1016/j.virol.2016.10.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lazzeri C, Valente S, Chiostri M, Attanà P, Picariello C, Sorini Dini C, Gensini GF (2013) Impact of age on the prognostic value of body mass index in ST-Elevation myocardial infarction. Nutr Metab Cardiovasc Dis 23(3):205–211. 10.1016/j.numecd.2012.05.013 [DOI] [PubMed] [Google Scholar]
- 27.Bucholz H, Beckman EM, Krumholz AL, Krumholz HA (2016) Excess weight and life expectancy after acute myocardial infarction: the obesity paradox reexamined. Am Heart J 172:173–181. 10.1016/j.ahj.2015.10.024.Excess [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Nelson CL, Karnath AF, Elkassabany NM, Guo Z, Liu J (2019) The serum albumin threshold for increased perioperative complications after total hip arthroplasty is 3.0 g/dL. HIP Int 29(2):166–171. 10.1177/1120700018808704 [DOI] [PubMed] [Google Scholar]
- 29.Bohl DD, Shen MR, Hannon CP, Fillingham YA, Darrith B, Valle CJD (2017) Serum albumin predicts survival and postoperative course following surgery for geriatric hip fracture. J Bone Jt Surg Am Vol 99(24):2110–2118. 10.2106/JBJS.16.01620 [DOI] [PubMed] [Google Scholar]
- 30.Arshi A, Rezzadeh K, Stavrakis AI, Bukata SV, Zeegen EN (2019) Standardized Hospital-Based Care Programs Improve Geriatric Hip Fracture Outcomes: An Analysis of the ACS NSQIP Targeted Hip Fracture Series. J Orthop Trauma 33(6):E223–E228. 10.1097/BOT.0000000000001443 [DOI] [PubMed] [Google Scholar]
- 31.Engin AB and Engin A (2018) The definition and prevalence of obesity and metabolic syndrome. In: Obesity and Lipotoxicity. Springer International Publishing, pp. 1–17 [Google Scholar]
- 32.Gregg EW, Cheng YJ, Srinivasan M, Lin J, Geiss LS, Albright AL, Imperatore G (2018) Trends in cause-specific mortality among adults with and without diagnosed diabetes in the USA: an epidemiological analysis of linked national survey and vital statistics data. Lancet 391(10138):2430–2440. 10.1016/S0140-6736(18)30314-3 [DOI] [PubMed] [Google Scholar]
- 33.Benjamin E et al. (2019) Heart Disease and Stroke Statistics-2019 Update: A Report From the American Heart Association. Circulation 139(10):e56–e528 [DOI] [PubMed] [Google Scholar]
- 34.Yornbi JC, Putineanu DC, Cornu O, Lavand’homme P, Cornette P, Castanares-Zapatero D (2019) Low haemoglobin at admission is associated with mortality after hip fractures in elderly patients. Bone Jt J 101-B(9):1122–1128. 10.1302/0301-620X.101B9.BJJ-2019-0526.R1 [DOI] [PubMed] [Google Scholar]
- 35.Bergström U, Jonsson H, Gustafson Y, Pettersson U, Stenlund H, Svensson O (2009) The hip fracture incidence curve is shifting to the right: a forecast of the age-quake. Acta Orthop 80(5):520–524. 10.3109/17453670903278282 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Malik AT, Quatman CE, Phieffer LS, Ly TV, Wiseman J, Khan SN (2019) The impact of metabolic syndrome on 30-day outcomes in geriatric hip fracture surgeries. Eur J Orthop Surg Traumatol 29(2):427–433. 10.1007/s00590-018-2298-4 [DOI] [PubMed] [Google Scholar]
- 37.Aprato A, Audisio A, Santoro A, Grosso E, Devivo S, Berardino M, Massè A (2018) Fascia-iliaca compartment block vs intra-articular hip injection for preoperative pain management in intracapsular hip fractures: a blind, randomized, controlled trial. Injury 49(12):2203–2208. 10.1016/j.injury.2018.09.042 [DOI] [PubMed] [Google Scholar]
- 38.Bellas N et al. (2020) Impact of Preoperative Specialty Consults on Hospitalist Comanagement of Hip Fracture Patients. J Hosp Med 15(1):16–21. 10.12788/jhm.3264 [DOI] [PubMed] [Google Scholar]
- 39.Friedman SM, Mendelson DA, Bingham KW, Kates SL (2009) Impact of a comanaged geriatric fracture center on short-term hip fracture outcomes. Arch Intern Med 169(18):1712–1717. 10.1001/archinternmed.2009.321 [DOI] [PubMed] [Google Scholar]
- 40.Bhandari M, Swiontkowski M (2017) Management of acute hip fracture. N Engl J Med 377(21):2053–2062. 10.1056/NEJMcp1611090 [DOI] [PubMed] [Google Scholar]
- 41.Tohidi M, Brogly SB, Lajkosz K, Harrison MM, Campbell AR, VanDenKerkhof E, Mann SM (2019) Ten-year risk of complication and mortality after total hip arthroplasty in morbidly obese patients: a population study. Can J Surg 62(6):442–149. 10.1503/cjs.017318 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Banack HR, Kaufman JS, Wactawski-Wende J, Troen BR, Stovitz SD (2019) Investigating and Remediating Selection Bias in Geriatrics Research: The Selection Bias Toolkit. J Am Geriatr Soc 67(9):1970–1976. 10.1111/jgs.16022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Banack HR, Harper S, Kaufman JS (2018) Accounting for Selection Bias in Studies of Acute Cardiac Events. Can J Cardiol 34(6):709–716. 10.1016/j.cjca.2018.01.013 [DOI] [PubMed] [Google Scholar]
- 44.Aziz KT, Best MJ, Shi BY, Srikumaran U (2020) Missing Data in the National Surgical Quality Improvement Program Database: How Does It Affect the Identification of Risk Factors for Shoulder Surgery Complications? Arthrosc - J Arthrosc Relat Surg 36:1–7. 10.1016/j.arthro.2019.12.028 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
NSQIP is publicly available to participating institutions. The dataset for this manuscript is available for review.