Revisiting the Role of BMI in the Guidelines for Bariatric Surgery

The high and growing prevalence of obesity worldwide, combined with the corresponding increase in the prevalence of diabetes, and further combined with the relatively poor long-term success of conservative obesity treatment, led to a sharp escalation in the use of bariatric surgery, the prevalence of which has tripled in the past decade (1). A total of 468,609 bariatric procedures were performed worldwide in the year 2013 according to the International Federation for the Surgery of Obesity and Metabolic Disorders (IFSO) survey (1), and this extraordinary acceptance, by both patients and physicians, stems from the fact that these procedures are not only efficient for weight loss but also effective in treating the metabolic syndrome in people with obesity. Bariatric surgery, as its name suggests, was originally designed to reduce weight, with current eligibility criteria being BMI ≥40 or ≥35 kg/m² with comorbidities, while lower BMIs are excluded. However, its significant beneficial impact on several components of the metabolic syndrome, particularly uncontrolled diabetes, led to the introduction of bariatric surgery as a means to treat diabetes two to three decades ago (2–4), yet only in obese individuals. Since then, along with the improvement in surgical technique, bariatric surgery performed in morbidly obese individuals established itself not only as the most effective means of treating diabetes but also as a “cure” to the disease in a significant proportion of the cases (5–7). The excellent results of bariatric surgery in improving glucose homeostasis in obese patients with type 2 diabetes led to the coinage of the term “metabolic surgery” (8), wherein amelioration of hyperglycemia in obese patients with type 2 diabetes has been observed even before a significant weight loss has taken place (9). Several mechanisms were offered to explain this dramatic impact of bariatric surgery on glucose homeostasis in patients with diabetes, including changes in the secretion of glucagon-like peptide 1, fibroblast growth factor 19, and bile acid as well as in the gut microbiota (10–13). Although the exact mechanisms that lead to postsurgical remission of type 2 diabetes are not fully understood, the fact that the metabolic improvement seems to be, at least partially, weight loss independent provides sufficient basis to consider “bariatric” surgery for nonmorbidly obese or even nonobese metabolically impaired patients.

Before we enter a debate over bariatric surgery as a treatment for the metabolic syndrome, especially uncontrolled diabetes, in less obese individuals, it is important to review the factual and observational insights stemming from two decades of accumulating experience.

In a large meta-analysis summarizing 621 studies with morbidly obese patients with diabetes undergoing bariatric surgery, a dramatic impact of surgery on diabetes outcomes with an average 2.1% decrease in HbA_1c and a diabetes remission rate of 78.1% has been found (14). However, this work is based on cohorts lacking a long-term follow-up period. In trials where the follow-up period is extended to ≥5 years, diabetes remission rate is lower, and a significant percent of the patients in remission eventually relapse (15–17). The Swedish Obese Subjects (SOS) study, providing 15 years of follow-up (17), has shown a high rate of diabetes relapse after an initial remission (occurring in >50% of the patients) and has also shown that the impact of bariatric surgery on long-term diabetes remission is reserved for patients with a short duration of disease. However, both microvascular and macrovascular complications in this cohort were significantly lower in operated patients compared with the control group, suggesting that the benefits of bariatric surgery may encompass important advantages beyond diabetes remission.

Hence, it is safe to say at this point in time that bariatric surgery in eligible patients improves diabetes control and seems to improve both micro- and macrovascular disease, whereas long-term remission, or “cure” of diabetes, is only partial and temporary.

What happens when bariatric surgery is performed in “noneligible” patients, who are leaner than prescribed by guidelines? Are the metabolic advantages of the procedure outlined above still present in the nonobese population? Does an obese patient with diabetes suffer from the same disease as a lean patient with diabetes? Is it justified, at this point in time, to overlook BMI as the major determinant in deciding which patient to operate? Or should metabolic surgery still be reserved only for obese individuals?

Several studies have attempted to evaluate the effect of bariatric surgery on metabolic syndrome in patients with stage 1 obesity (BMI 30–35 kg/m²) and in the nonobese population (BMI <30 kg/m²). Studies in which Roux-en-Y gastric bypass (RYGB) was performed in patients with diabetes with stage 1 obesity show a significant, albeit inconsistent, diabetes remission rate ranging from 25 to 88% in similar populations (5,18). This ability to surgically cure diabetes declines even further where nonobese patients with diabetes are concerned. While a large meta-analysis on bariatric surgery performed in morbidly obese patients with diabetes shows an average decrease in HbA_1c of 2.1% and a diabetes remission rate of 78.1% (14), a smaller meta-analysis, looking at 290 patients with diabetes with a BMI <30 kg/m², is much less impressive, with an average decrease in HbA_1c of 1.88% and a remission rate of only 42% (19). However, it should be noted that these two meta-analyses entailing different procedures, with diverse definitions of remission, are hard to compare.

Only a small number of studies have been directed to the outcome of metabolic surgeries in the nonobese population. These studies were performed mostly on small cohorts of patients with uncontrolled diabetes, using diverse surgical techniques and allowing a relatively short follow-up period, with inconsistent results (20–24). Studies with a follow-up period of at least 12 months are summarized in Table 1.

An ileal interposition was performed along with a diverted sleeve gastrectomy in 69 patients with diabetes with a BMI of 21–29 kg/m² (25). In this cohort, average HbA_1c decreased from 8.7 to 5.9%, with 65% of the cohort achieving “remission” of diabetes after an average 21-month follow-up period—marginally lower than the remission rate seen in the obese population. An Australian randomized study using the gastric banding operation versus medical treatment in overweight patients with diabetes (BMI 25–30 kg/m²) showed a diabetes remission rate of 52% in the surgical group compared with 8% in the medical group after 2 years of follow-up (26). RYGB or a mini–gastric bypass, performed in 103 patients with uncontrolled diabetes with an average BMI of 26 kg/m², resulted in a somewhat lower remission rate of only 30% after a 12-month follow-up period (27). A 2-year follow-up after a biliopancreatic diversion (BPD) procedure performed in 15 nonobese (BMI 25–30 kg/m²) and 15 obese (BMI 30–35 kg/m²) subjects with diabetes revealed a vast difference in diabetes remission rate between the two groups, reaching 27 and 67% in the nonobese and obese groups, respectively (28); this observation suggests that diabetes occurring at a lower BMI is more resistant to bariatric treatment. A metabolic surgery was performed in 80 nonobese and 432 obese (BMI >30 kg/m²) subjects with diabetes in a single medical institution in Taiwan (29). In this study, the diabetes remission rate observed in the nonobese group was 25%, which was significantly lower than the remission rate observed in subjects with a higher BMI (49.5 and 79% in subjects with a BMI of 30–35 and >35 kg/m², respectively).

Taken together, studies performed in subjects with diabetes with lower BMIs undergoing bariatric surgery demonstrate that these procedures are much less effective in these patients and raise the question: why do patients with diabetes with a lower BMI react differently to surgery?

Type 2 diabetes is a heterogeneous group of diseases, with diverse degrees of insulin sensitivity and β-cell function. Subtypes of type 2 diabetes include conditions such as ketosis-resistant diabetes of the young, a lean type 2 diabetes variant, and the like. Lean type 2 diabetes is not a rare condition in Western society; in a study in a large Chicago cohort, comprising 18,000 subjects with diabetes, 13% had a BMI of ≤25 kg/m² (30). This study compared 1,784 normal-weight patients with diabetes (average BMI 23 kg/m²) with 8,630 obese patients with diabetes (average BMI 35.6 kg/m²) and found that nonobese diabetes is more prevalent in men and that subjects have a higher insulin-use rate and a lower triglyceride-to-HDL ratio, suggesting a state of increased β-cell failure as opposed to insulin resistance, the latter being more prevalent in subjects with diabetes with a higher BMI. Furthermore, genetic predispositions, affecting insulin secretion, may also contribute to diabetes occurring at a relatively low BMI (31). These observations provide additional evidence suggesting that nonobese individuals with type 2 diabetes represent a unique subgroup that does not respond to metabolic surgery as robustly as does the obese population with diabetes.

Other aspects of the metabolic syndrome that seem to be less affected by surgery in the nonobese population should also be considered. A meta-analysis on nonobese patients with diabetes undergoing metabolic surgery showed that hypertension had been resolved in 89% of the subjects; however, total cholesterol and triglyceride levels were not significantly changed after surgery (19). Interestingly, one of the articles included in this study reported an unexpected increase in triglyceride levels, which was still detected 2 years after a BPD operation (28). Although this unfavorable outcome in terms of triglycerides and HDL is based on a small population and may be related specifically to the BPD operation, there is no evidence that bariatric surgery in nonobese subjects improves lipid profile. This outcome should be considered by practitioners referring patients to surgery in an attempt to improve lipid profile.

Cardiovascular events and survival rate, which are affected by bariatric procedures, now enter the equation. Studies underlying the increased survival and decreased cardiovascular event rates after bariatric surgery are not randomized and were all performed on morbidly obese individuals (32–36). To date, there is no substantiating evidence based on randomized controlled studies that bariatric surgery has a beneficial impact on hard end points such as mortality and/or cardiovascular events in obese subjects, let alone in nonobese subjects.

Bone density is yet another aspect that has to be taken into consideration. Several studies looking at bone metabolism after bariatric surgery have shown that surgery may induce a significant amount of bone loss (37–40). There are several explanations for decrease in bone mass, which include the reduction of mechanical load on the skeleton, secondary hyperparathyroidism due to reduced absorption of calcium and/or vitamin D, and changes in gut peptides (40–42). It stands to reason that nonobese individuals, not sharing the same extent of mechanical load protection as obese individuals, will be more prone to bone loss, especially after malabsorptive procedures, which are relatively prevalent as metabolic surgeries.

One of the most vital factors affected by bariatric surgery is quality of life (QOL). Both obesity-related and disease-related QOL seem to improve after bariatric procedures in obese individuals (43). This comes as no surprise, since for a morbidly obese individual, losing a significant amount of weight may be a life-changing event. However, obesity-related QOL is irrelevant to the nonobese population, whereas metabolic diseases are only partially resolved in this population. Only limited data pertaining to QOL are available for the nonobese population, with one study demonstrating an improvement in physical (but not mental) well-being after gastric banding (26).

Finally, complication rate is a major consideration and is relatively high in nonobese metabolic surgery compared with bariatric surgery. In an above-mentioned meta-analysis, a 6.2% major surgical complication rate, including intestinal obstruction, intestinal perforation, and intra-abdominal bleeding, was reported, as well as a 1.7% reoperation rate (19). This rate is much higher than the 3.2% complication rate reported in patients with a somewhat higher BMI (44).

In summary, studies performed in nonobese subjects with diabetes undergoing bariatric surgery, compared with the same in obese subjects, reveal that surgery in this population is less effective as a means to control both diabetes and dyslipidemia, its ability to reduce cardiovascular disease as well as reduce mortality rate is a mere speculation, it may entail an increased risk of bone loss, and it involves a notable increase in the rate of complications. Even the well-appreciated improvement in QOL seen in morbidly obese individuals who undergo surgery does not seem to be as impressive where nonobese individuals are concerned. Given the above, it is apparent that bariatric surgery should not be routinely recommended in nonobese individuals.

Current BMI cutoffs for bariatric surgery stem from the National Institutes of Health consensus statement, based on a consensus development conference dating back to 1991 (45), where the authors themselves state that “there are insufficient data on which to base recommendation for patient selection using objective clinical features alone.” Despite significant improvements in surgical techniques and abundant data that have accumulated, the same golden “BMI ≥40 or ≥35 kg/m² with comorbidities” standard is still practiced as originally recommended. In the past few years, quite a few studies, nearly all performed in patients with diabetes, have challenged the “BMI ≥35 kg/m²” threshold for bariatric surgery, while adopting the more appropriate designation “metabolic surgery.” Many of these studies provide data suggesting that the extraordinary beneficial impact these procedures have on the control of diabetes, as well as on other obesity-related metabolic conditions, still holds when the preoperative BMI is <35 kg/m² (5,46–48).

The realization that patients with diabetes, who are not necessarily morbidly obese, may dramatically improve glucose homeostasis via bariatric surgery raised the inevitable question of whether BMI was the most appropriate patient selection tool for metabolic surgery. It has been shown that BMI, determined by both fat and lean body mass, does not necessarily predict metabolic status. Although there is much overlap between obese individuals (per BMI) and those having a high degree of adiposity, a significant number of obese individuals do not suffer from metabolic syndrome (49,50) and are now referred to as metabolically healthy obese. On the other hand, less obese, or nonobese, individuals may be found to be more metabolically impaired. Individuals who share the same BMI do not necessarily share the same degree of fatness and may have a very different fat mass percentage as well as a different fat distribution, with diverse metabolic consequences. Visceral adiposity and intramyocellular lipids as well as triglyceride content in the liver are more strongly correlated to metabolic abnormalities, including insulin resistance, compared with mere high BMI (51–54).

It is therefore not surprising that BMI, being a relatively poor indicator of metabolic status, was not able to predict the metabolic outcomes of bariatric surgery. The SOS cohort provided data showing that patients undergoing bariatric surgery have a lower chance of developing diabetes, cardiovascular events, and cancer, as well as overall mortality (34,35,55–58); however, in neither of these studies has baseline BMI been established as a predicting factor. This result underscores the limited significance of presurgery BMI to the outcomes of metabolic surgery. In another study based on the SOS cohort, subjects were retrospectively divided into groups deemed eligible or noneligible for bariatric surgery according to current guidelines and were then assessed for the development of diabetes and cardiovascular outcomes (57). After a 10-year observation period, subjects qualifying for surgery under current guidelines had a similar rate of diabetes remission compared with those who did not qualify (73 and 67%, respectively). Briefly, this means that some patients who would not be eligible for bariatric surgery, according to the currently used guidelines, could still have benefited from the surgery. A meta-analysis looking at 94 studies including 4,944 patients with type 2 diabetes showed that bariatric surgery led to similar rates of diabetes remission in patients who had a baseline BMI <35 kg/m² (72%) and in patients who had a baseline BMI ≥35 kg/m² (71%) (59).

The IFSO recently released a position statement regarding bariatric surgery in class 1 obesity, where the authors refer to BMI as “an inaccurate index of adiposity and a poor health risk predictor” and state that in patients with class 1 obesity associated with significant obesity-related comorbidities, “access to bariatric surgery should not be denied . . . simply on the basis of the BMI level” (43), suggesting that the “BMI-centric” attitude should be replaced by a more individualized, metabolically oriented approach in this population.

Increased waist circumference is a well-established predictor for development of type 2 diabetes (60–62), and several studies have shown that it is a superior predictor for diabetes development compared with BMI (63–65). In the above-mentioned meta-analysis (59), baseline waist circumference was the only parameter that was correlated with HbA_1c reduction after surgery; counterintuitively, subjects starting at a lower baseline waist circumference benefited from a higher rate of HbA_1c reduction. The authors suggest that this counterintuitive result may reflect a study selection bias for patients with higher HbA_1c. It should be noted that those patients with lower baseline waist circumference, still undergoing bariatric surgery, most likely had a more advanced disease accompanied by a higher HbA_1c. Nonetheless, this study clearly shows that bariatric surgery is an equally effective diabetes treatment in both the morbidly obese and nonmorbidly obese population with diabetes.

Taken together, these data suggest that BMI is an outdated and probably poor indicator for metabolic improvement after bariatric surgery and call for finding more reliable indicators for selecting patients for bariatric surgery based on metabolic status.

The Edmonton Obesity Staging System (EOSS) has been proposed as a patient-selection tool more capable of predicting the outcome of bariatric surgery compared with anthropometric measures such as BMI or waist circumference. This system takes into account the severity of the obesity-related comorbidities, as well as the functional and mental status of the obese patient considering bariatric surgery (66). According to this system, the patient is assigned a stage from 0 to 4, wherein stage 0 describes a patient with no apparent obesity-related comorbidities, no physical symptoms or psychopathology, and no functional limitations or impaired well-being related to obesity and stage 4 describes a patient with severe disabilities due to obesity-related comorbidities, severe psychopathology, physical limitations, and/or severe impairment of well-being. Using the criteria of the EOSS, Kuk et al. (67) found that obese subjects assigned to stage 0–1 did not show an increase in mortality risk, whereas subjects assigned to stage 2 or 3 had an increased all-cause as well as cardiovascular-related mortality risk.

Another study assessed the capacity of the EOSS to predict mortality analyzed data from the National Health and Nutrition Examination Surveys (NHANES) (68). In this cohort, patients assigned to stages 2 and 3 had an increased mortality risk (with a hazard ratio of 1.57 and 2.69, respectively) compared with patients assigned to stages 0 and 1, even after adjustment for BMI and the metabolic syndrome; this was true also in a cohort eligible for bariatric surgery. These studies show that patients sharing the same BMI may exhibit diverse cardiometabolic risks and may suggest that care providers, in selecting their treatment regimens, should prioritize patients at risk over merely obese individuals. Interestingly, in a study applying the criteria of the EOSS to a cohort of 5,787 patients attending a weight-management clinic, it was found that the stage approach could predict one’s ability to lose weight (69). Patients assigned to stage 0–1 lost weight more easily and rapidly than those assigned to stages 2 and 3, thereby providing a weight-independent tool for the prediction of weight loss. This observation led to the notion that the EOSS may replace BMI as a better tool for selecting patients who can benefit from bariatric surgery (70).

Lastly, like other surgeries, bariatric surgery also entails complications, and that risk should be taken into account on an individual basis while selecting the most suitable candidates for surgery. The King’s College criteria offer a holistic method for assisting perioperative complications by gathering all physical, mental, social, and financial information (71); however, this approach may suffer from practical limitations stemming from inherent difficulties in data collection.

The ultimate bariatric eligibility scale, enabling an optimal selection of candidates for bariatric surgery, based on validated hard outcomes, such as cardiovascular disease, complication rate, and mortality, is yet to be established. However, a compilation of a robust bariatric eligibility scale that will be incorporated into the metabolic surgery preoperative assessment and replace the BMI criteria seems to gather momentum.

It is the authors’ opinion, based on current available information, that the following criteria should be used for selecting patients for bariatric surgery: 1) BMI >40 kg/m², 2) BMI >35 kg/m² plus comorbidities, 3) BMI 30–35 kg/m² and uncontrolled diabetes, 4) BMI 30–35 kg/m² plus EOSS stage >2 subject to the practitioner’s clinical judgment, and 5) exclusion of BMI <30 kg/m² subject to further research.

It is important to emphasize that at present there are no guidelines recommending bariatric surgery in nonobese patients, and this procedure is considered investigational and preliminary in this population.

We hope for elucidation of the mechanisms acting in bringing about the remission of diabetes after bariatric surgery, which will hopefully enable further refinement of the selection process of candidates for bariatric surgery.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.