Systematic Review and Meta-Analysis of Bariatric Surgery for Pediatric Obesity

Jonathan R. Treadwell, PhD; Fang Sun, MD, PhD; Karen Schoelles, MD, SM

Disclosures

Annals of Surgery. 2008;248(5):763-776. 

In This Article

Results

The study identification process is depicted in Figure 1. Of the 18 unique studies that met the inclusion criteria, 8 investigated laparoscopic adjustable gastric banding (LAGB), 5 investigated RYGB, one investigated both RYGB and vertical banded gastroplasty (VBG), 2 VBG only, one BPD only, and one the banded bypass (BB) procedure. The total patient enrollment was 641; the procedure-specific totals were 352 for LAGB, 131 for RYGB, 71 for VBG, 68 for BPD, and 19 for BB.

Study attrition diagram depicts article counts at each stage of the review process.

Key characteristics about enrolled patients and the procedures performed are listed in Table 1 . For LAGB, the earliest procedures on pediatric patients were performed in 1996; 6 of the 8 LAGB studies used the LAP-BAND® (Inamed Health, Santa Barbara, CA) one used the Swedish Adjustable Gastric Band (SAGB; Ethicon Endo-Surgery, Cincinnati, OH), and one used the SAGB in 74% of patients and the LAP-BAND® in the remaining 26%. In the United States, the LAP-BAND is F.D.A.-approved only in adults (although a trial in adolescents is ongoing); the SAGB is still in the FDA approval process. The mean age ranged from 15.6 to 18 (weighted average 16.7; overall age range 9-20). The percentage of females ranged from 50% to 81% (weighted average 70%). The mean baseline BMI ranged from 42.4 to 50.5 (weighted average 46; overall BMI range 31-76.6).

For RYGB, the earliest procedures on pediatric patients were performed in the 1970s. Of the 6 RYGB studies, 2 used a laparoscopic approach, 3 used an open approach, and one used an open approach for 94% of procedures and a laparoscopic approach for the remaining 6%. The mean age ranged from 15.7 to 17.57 (weighted average 16.8; overall age range 11-21). The percentage female ranged from 57% to 79% (weighted average 66%). The mean baseline BMI ranged from 47 to 56.5 (weighted average 51.8; overall BMI range 38-95.5). All other procedures were performed using an open approach; the patients undergoing these procedures were similar to those undergoing LAGB or RYGB.

Fourteen of 18 studies explicitly stated that, prior to surgery, nonsurgical methods of weight loss had been attempted and were unsuccessful in all patients. Eleven of 18 studies were conducted in the USA; the other 7 were conducted in Israel (two studies), Italy (two studies), Australia, Austria, or Saudi Arabia. Six non-USA studies investigated LAGB, and the seventh investigated BPD. Fourteen studies reported data from a single surgical center, whereas the other 4 were from 2 or more surgical centers.

One of the 18 studies reported a control group of patients who were not treated with bariatric surgery (the Lawson study of RYGB).[62] This control group included 12 patients who had completed one year in a nonsurgical pediatric weight management program. However, patients in the control group were much different from those who received surgery. Specifically, the control group patients weighed statistically significantly less at baseline (mean BMI 47) than surgical patients (mean BMI 56.5), and the study did not report any medical comorbidities among control group patients, as compared with surgical patients who had several comorbidities at baseline. These factors mean that the groups were not well-matched at baseline, consequently we excluded the data from this control group, and included only the data from the surgical group. Thus, for the purpose of our review, the studies were all case series.

Only one study was clearly conducted prospectively (the Nadler study). Retrospective design may introduce bias because at the point when authors decide to publish the data, they are armed with the knowledge of the favorable (or unfavorable) outcomes experienced by patients. One strategy to counteract this potential bias is to include all eligible patients consecutively. This was performed in 14 studies, not performed in 3 studies, and unclear in the remaining study. Consecutive inclusions helps reduce selection for favorable outcomes. A related quality factor is study completion rate: studies would ideally report long-term outcome data on all patients who received the treatment. However, with any long-term follow-up, there will be patients whose outcomes are not known or patients who have not reached longer timepoints. Usually it is unclear whether dropouts experienced similar outcomes as those remaining in the study, or whether recently-treated patients will eventually experience similar outcomes.

Patient weight is easily measured objectively, and of the 13 studies included for BMI data, 9 stated that patients attended follow-up visits in the clinic, suggesting that the weight data were based on actual weight measurement rather than patients' self-reporting. Three other studies did not report sufficient information on how the weight data were measured, and the other study (the study by Greenstein and Rabner[68]) based weight data on patient self-reporting. Two other quality criteria that we applied involved the independence of outcome assessors and the study funding source. None of the 18 studies used independent outcome assessment (eg, weights recorded at a center independent from the surgeon and surgical staff), which raises the possibility of outcome recording bias. Also, only one of the 18 studies reported the study funding source. However, studies were generally conducted by bariatric surgeons, who do have financial interest in the performance of surgery.

In the Results sections below, we discuss the outcome data that met the inclusion criteria, separately for different bariatric procedures. This separation is important because different bariatric procedures would be expected to result in different amounts of weight loss, different rates of comorbidity resolution, and different types of complications.

Reduction in BMI. Of the 8 studies, 2 studies' data did not meet inclusion criteria for this outcome because authors either did not report the number of patients followed for 1+ years (Al-Qahtani et al),[50] or did not report 1+ year data for at least 50% of patients (Nadler et al),[51] Thus, our BMI inclusion criteria were met by 6 of the 8 LAGB studies.

We conducted a meta-analysis of BMI change at longest follow-up after LAGB (top half of Fig. 2). The length of follow-up in the 5 studies ranged from 1 to 3 years, and the percentage of surgical patients who were included in each study ranged from 58% to 100%. Our meta-analysis indicated substantial heterogeneity (I2=56%). Given the imputation of a prepost correlation for 4 of the 6 studies (Dillard et al,[48] Yitzhak et al,[52] Silberhumer et al,[53] and Angrisani et al[55]) we did not attempt to explain this heterogeneity via meta-regression. The 95% confidence interval of the random-effects summary statistic ranged from -13.7 to -10.6 BMI units, indicating substantial weight loss after adjustable gastric banding. This interval compares favorably to the minimum weight loss considered clinically significant (the dashed line in the figure at 3.4 BMI units, which corresponds to 7% of body weight in these LAGB patients). The finding of clinically significant weight loss after bariatric surgery persisted through all of our sensitivity analyses, indicating a robust finding for BMI reduction at longest follow-up.

Meta-analyses of BMI reduction at longest follow-up after bariatric surgery. Forest plots of random-effects meta-analyses of the change in BMI at longest follow-up, separately for laparoscopic adjustable gastric banding (top half) and Roux-en-Y gastric bypass (bottom half). The data for Fielding et al[82] and Collins et al[56] are based on earlier publications from those centers (Dolan et al[83] and Stanford et al[84], respectively) because the more recent publications did not report data on at least 50% of patients at one year or longer.

Four of the 8 LAGB studies reported comorbidity data that met inclusion criteria. The pertinent data on comorbidity resolution after LAGB appear in the upper section of Table 2 . The mean length of follow-up in these 4 studies ranged from 1.3 to 2.9 years. For diabetes, 2 studies[50,53] reported resolution rates of 100% (7/7) and 80% (4/5). For hypertension, 3 studies[50,52,53] reported resolution rates of 50% (6/12), 100% (6/6), and 100% (3/3). For other comorbidities, no more than one study reported resolution rates.

All 8 LAGB studies were included for complications data ( Table 3 ). No in-hospital or postoperative death was reported in any LAGB study. Reoperations were performed on 8% of the patients (28/352) to correct various complications such as band slippage, gastric dilation, intragastric band migration, psychologic intolerance of band, hiatal hernia, cholecystitis, and tubing crack. Overall, band slippage was the most frequently reported specific complication (3%; 12/352). Eight of the 12 cases occurred in one center using SAGB, while the other 4 cases occurred in 3 centers using LAP-BAND. In addition, 8 cases of iron deficiency and 5 cases of mild hair loss were reported; the remaining reported complications had a case number equal to or less than three. No studies reported data on the impact of surgery on growth or development.

Reduction in BMI. Of the 6 studies, 1 (Barnett et al)[67] did not meet inclusion criteria for this outcome because authors did not report the length of follow-up of patients receiving specific procedures. Another study (Rand et al)[29] was excluded from consideration because it was very low quality because of nonconsecutive patient inclusion and the possibility that BMI data were based on patient recall rather than objective weight measurement. Thus, our BMI inclusion criteria were met by 4 of the 6 RYGB studies.

As with LAGB, we conducted a meta-analysis of BMI change at longest follow-up after RYGB (bottom half of Fig. 2). The mean length of follow-up in the 4 studies ranged from 1.0 years to 6.3 years, and the percentage of surgical patients who were included in each study ranged from 61% to 90%. Our meta-analysis indicated no heterogeneity (I 2=0%). The 95% confidence interval of the random-effects summary statistic ranged from -17.8 to -22.3 BMI units, indicating substantial weight loss after Roux-en-Y gastric bypass. As with LAGB, this interval compares favorably to the minimum weight loss considered clinically significant (the dashed line in the figure at 4.1 BMI units, which corresponds to 7% of body weight in these RYGB patients). The finding of weight loss after bariatric surgery persisted through all of our sensitivity analyses, indicating a robust finding for BMI reduction at longest follow-up.

Four of 6 RYGB studies reported comorbidity data that met inclusion criteria. The included data appear in the middle section of Table 2 . Among these 4 studies, the mean length of follow-up ranged from 5 months to 2.7 years. For hypertension, 3 studies[27,28,60] reported resolution rates of 50% (3/6), 82% (9/11), and 100% (3/3). For sleep apnea, 2 studies[27,62] each reported resolution rates of 100% (6/6 and 10/10). For other comorbidities, no more than one study reported resolution rates.

All 6 RYGB studies were included for complications data ( Table 3 ). No in-hospital death was reported. One patient in the Lawson et al study.[62,63,64] died 9 months after surgery, because of severe Clostridium difficile colitis, severe diarrhea, an extended period of profound hypovolemia, and multiple organ failure. Three additional patients died of causes that were unlikely to be directly related to the bariatric surgeries (one patient in the Barnett study[67] died 4 years after surgery; 2 patients in the Sugerman et al study[27] died 2 years and 6 years after surgery).

Reported postoperative complications of RYGB included some potentially life-threatening conditions such as shock, pulmonary embolism, severe malnutrition, immediate postoperative bleeding, and gastrointestinal obstruction. The most frequently reported type of complication involved protein-calorie malnutrition and micronutrient deficiency. Inconsistencies in reporting precluded calculation of an overall reoperation rate after RYGB.

Regarding physical maturation, Rand et al[29] reported patients' preoperative and postoperative heights and concluded that there was no evidence of growth retardation after surgery (at an average follow-up of 6 years). However, the authors of the study also stated that the question as to whether these adolescents achieved their expected growth could not be extracted from data available.

For weight or BMI data, three[68,69,71] of the 5 studies of other procedures met the inclusion criteria, however, all 3 of these studies were of very low quality. Unique quality problems with these 3 studies involved nonconsecutive patient inclusion in 2 of the 3 studies; weights based on patient reporting or unreported method of obtaining weight data; and <85% completion in 2 of the 3 studies.

For comorbidity resolution, 3 of the 5 studies reported data, but Barnett et al[67] only reported data combined across procedures, and Greenstein and Rabner[68] included no more than 2 patients for any given comorbidity, therefore these data did not meet inclusion criteria. The reported comorbidity results of the BPD study appear at the bottom of Table 2 .

The complications data appear in the lower sections of Table 3 . No in-hospital deaths were reported, and 3 follow-up deaths were reported (all after BPD; the causes were protein malnutrition, pulmonary edema, and acute necrotizing pancreatitis).[70] In the VBG studies, recurrent gastric ulceration (in 2 patients), enlarged pouches (in 2 patients) and staple line disruption (in one patient) were reported. The study of BPD (a malabsorptive procedure) reported 11 cases of protein malnutrition. In the banded bypass study, 2 revisions for gastro-gastric fistula, one cholecystectomy, one recurrent marginal ulcer requiring antacids, and 3 plastic surgeries for excess skin were reported as postsurgery complications. None of the 5 studies reported any data on the potential impact of physical growth.

Our strength-of-evidence ratings for each outcome of each bariatric procedure are shown in Table 4 . The strength ratings reflect a balance of various factors including low/moderate quality, limited quantity (especially for comorbidity resolution), large magnitude of reported effects, and consistency of results across studies. We did not draw precise quantitative conclusions for any outcomes because of the imputation of prepost correlations (for BMI data) or limited numbers of studies (comorbidity resolution).

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.

processing....