We included studies that should have compared the deficiency rates and complication rates of 4 nutrients (vitamin B12, iron, vitamin D, and calcium) in people with obesity after different BS surgical approach manuscript Formatting. Prior to the start of the study, it was registered in the International Prospective Register of Systematic Reviews (PROSPERO) (CRD42022351956).

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (12). Our search strategy aimed to identify all studies examining the impact of different BS surgical approaches on people with obesity nutrient deficiencies. We used Mesh and did our best to include all synonymous medical terms in the search to avoid omissions. Two co-first authors collaborated with a Ph.D. (the second author of this paper) to identify search terms, as well as relevant search terms and keywords for our primary outcome, nutrient deficiency. We also identified search criteria and strategies and searched PubMed, Embase, Scopus, Google Scholar, and the Cochrane Clinical Trials Register for eligible studies. To ensure comprehensiveness, we also searched the cited literature of relevant reviews on an article-by-article basis.

Several inclusion criteria were used in determining study eligibility. a.

Study design: non-review, non-letters, and other non-research articles.

All three reviewers were involved, and many abstracts were identified for full-text review. All reviewers received training on study eligibility criteria and data extraction. A panel of three reviewers reads and assessed all included full-text articles to determine whether they met the eligibility criteria and, if not, to explain the reasons for exclusion. One reviewer independently extracted the following data elements from each article that met the eligibility criteria table developed by the study team: first author's name, year of publication, journal name, location, BS program model of the study, nutrient deficiency status, study population (including disease), selection of study cohort, sample size, study purpose, primary outcome, and statistical significance (if applicable) (13).

The study used the gemtc package in R (R Studio 2022.02.1 Build 461-Inc version) (https://posit.co) to analyze the data (14, 15). We performed a network meta-analysis (Markov chain Monte Carlo approach in a Bayesian framework) in order to study nutrient deficiencies due to different surgical modalities, pooling the different BS surgical modalities and nutrient deficiencies included in the studies (16). Not all studies could be included in the network meta-analysis, as some of the included studies did not include nutrient deficiency as a dichotomous variable and they only provided data for calculating odds ratios (ORs). We first developed a consistency model using forest plots to describe the comparison of ORs and 95% confidence intervals (CrI) between different BS surgical modalities. The difference in effect between interventions was judged to be statistically significant by whether the Crl of the interventions compared with each other crossed 1 or not. We also performed nodal analysis methods and consistency tests for all included studies to look for sources of heterogeneity and to ensure the assumption of homogeneity in the model.

In the study, we plotted trajectory and density plots to assess the volatility and overlap of MCMC chains and the convergence of the model (17). From the trajectory plots, we know whether the MCMC chain fluctuations are stable and have good overlap when the number of iterations reaches 5,000 or more; from the density plots, we know whether Bandwidth tends to 0 and reaches stability when the number of iterations reaches 20,000. When the requirements are satisfied, the synthesis indicates that the model converges well. We also plotted the Brooks-Gelman-Rubin diagnostic and calculated the potential size reduction factor (PSRF).

A satisfactory convergence model needs to satisfy two conditions at the same time: a.

97.5% of the scaling factor and the median value of the scaling factor converge to 1 and reach stability after n iterations of calculation.

Final conclusions will be drawn from the following three sources: a.

from the forest plot as to which surgical modality was used with the highest risk of outcome events.

3,312 potential studies were identified in our literature review (Figure 1).

We reviewed the citations, titles, and abstracts of 3,312 studies and identified 277 articles for inclusion. After a full-text review, 20 of these articles were included (Table 1). Studies were typically excluded because there was no comparison of surgical modalities or nutritional deficiencies were not the primary outcome. We began by mapping the network relationships in preparation for the network meta-analysis (Figure 2). While SG was directly compared with RYGB and LAGB due to the network linkages between trials in Figure 2, comparisons between the remaining two surgical modalities (i.e., RYGB and LAGB) were directly linked through at least two trials.

We included 13 trials (n = 2408) that reported vitamin B12 deficiency after BS (Table 1). A network meta-analysis was used to rank the BS surgical modalities in order of frequency of postoperative vitamin B12 deficiency (from A–C). The forest plot in Figure 3 depicts A for SG, B for RYGB, and C for LAGB. We show the ORs and 95% CrI for the comparison by forest plot, where the OR for the outcome of intervention B compared with A was 2.90 and the OR for the outcome of C compared with A was 2.00, indicating that the risk of outcome events was 2.9 times higher for surgical modality B than for intervention modality A and 2 times higher for C. The 95% Crl for ORs were (1.70–5.60) and (0.68–5.80), respectively. Since the Crl of the intervention crossed 1 for C compared to A, the difference in effect between them was not statistically significant. The trajectory and density plots show that the model converged well and satisfied the required conditions.

We ranked the probability of vitamin B12 deficiency following BS for each of the three surgical procedures (see Figure 4A). Therefore, postoperative SG is better in terms of vitamin B12 absorption than the other two surgical approaches (i.e., RYGB and LAGB). Comparisons of RYGB with LAGB are not statistically significant.

Six of our included trials (n = 987) reported calcium deficiency after BS (Table 1). We ranked the frequency of postoperative calcium deficiency (from A–C). We show the ORs and 95% CrI for comparison by forest plot, with an outcome OR of 2.80 for intervention B compared to A and 0.89 for C compared to A. This indicates that the risk of an outcome event using surgical modality B is 2.8 times greater than the risk of an outcome event using intervention A and 0.89 times greater than the risk of an outcome event using intervention C. The 95% Crl for ORs was (0.97–11.00), (0.096–8.20). Since the Crl comparisons for the interventions all crossed 1, the difference in effect between them was not statistically significant. It can be seen from the trajectory and density plots that the model converged well and met the required conditions.

We compared each intervention A–C to each other and computed the logarithm of the index outcomes. For each of the three surgical procedures, the probability of postoperative calcium deficiency was ranked (see Figure 4B). As a result, we could only conclude from the ranked results that RYGB had a higher risk of postoperative calcium deficiency symptoms than SG and LAGB.

We ranked the frequency of postoperative iron deficiency (from A–C). We show the ORs and 95% CrI of the comparisons by forest plots, with an OR of 2.50 for the outcome of comparison of intervention B with A and an OR of 1.10 for the outcome of comparison of C with A. This indicates that the risk of occurrence of an outcome event with surgical modality B is 2.5 times greater than the risk of occurrence of an outcome event with intervention A and 1.1 times greater with C. The 95% Crl of the ORs are (1.30–4.40), (0.43–2.60). Since the Crl for the comparison of A and C crossed 1, the difference in effect between them was not statistically significant. The incidence of RYGB is significantly higher than that of SG and LAGB as can be observed by the forest plot. it is clear from the trajectory and density plots that the model converges well and satisfies the required conditions.

The probability of postoperative calcium deficiency was ranked for each of the three surgical procedures (see Figure 4C). Direct comparison between interventions The I² between pair-wise comparisons (pair-wise) was 0%, and the I² between the results of network comparisons (network) was also 0%, then there was no heterogeneity between them and the homogeneity assumption was satisfied. Therefore, it can be judged that RYGB is more likely to cause iron deficiency symptoms in the postoperative period compared to SG and LAGB.

We ranked the frequency of postoperative vitamin D deficiency (from A–C). We showed the ORs and 95% CrI of the comparisons by forest plots, with an OR of 2.90 for the outcome of intervention B compared with A and 2.00 for the outcome of C compared with A. This indicates that the risk of an outcome event in surgical approach B is 2.9 times greater than the risk of an outcome event in intervention A and 2.0 times greater in C. The 95% Crl of the ORs was (1.7–5.60), (0.69–5.90). Since the Crl for the comparison of A and C crossed 1, the difference in effect between them was not statistically significant. From the forest plot, it can be seen that the incidence of RYGB is higher than SG and LAGB. from the trajectory and density plots, it can be seen that the model converges well and satisfies the required conditions.

The probability of postoperative calcium deficiency was ranked for each of the three surgical procedures (see Figure 4D). The direct comparison between interventions (pair-wise) is 0%, and the I² between network comparison (network) results is also 0%, then there is no heterogeneity between them and the homogeneity assumption is satisfied. Therefore, it can be judged that RYGB is more likely to cause postoperative Vitamin D deficiency symptoms compared with SG and LAGB.

The goal of choosing the most appropriate surgical option is always to create the best balance between weight loss, complications, and micronutrient deficiencies. In clinical practice, BS is often performed only with SG, RYGB, and LAGB (38). Current data suggest that RYGB is more likely to cause micronutrient deficiencies than SG and LAGB. Although previous studies have compared the three most commonly used procedures, there is still a lack of more accurate evidence to choose the most scientific procedure.

Compared with RYGB, LAGB has a higher risk of slip/dilation and surgical reversal/conversion, but a lower risk of stricture, ulceration, and hernia compared with RYGB (39). LAGB has a shorter duration of stay compared to RYGB; SG has a higher risk of reoperation compared to RYGB (40). Compared with RYGB, SG and LAGB are associated with a higher failure rate or secondary surgery, which may be the main reason for the current high clinical use of RYGB.

We used radar plots to show the combined ranking results of nutrient deficiencies (Figure 5). In Figure 5, there are three groups (i.e., SG, RYGB, and LAGB) with four dimensions (i.e., Vitamin B12, calcium, Vitamin D, and Iron). The farther from the center point means the higher risk of deficiency of that nutrient. The results showed that RYGB was more likely to cause vitamin B12 deficiency compared to SG, while LAGB was less pronounced, but LAGB was less deficient than RYGB (41). This may be because RYGB is a malabsorptive procedure, compared with SG and LAGB, Insufficient secretion of endogenous factors due to reduced gastric capacity and intestinal rearrangement, resulting in reduced absorption of vitamin B12, so we recommend postoperative supplementation and long-term, regular laboratory monitoring should have adhered to (42).

The duodenum and proximal jejunum are the best sites for dietary calcium absorption, where vitamin D is involved. Therefore, calcium deficiency can be further exacerbated if the diet is deficient in vitamin D (43). Food does not pass through the duodenum after RYGB, bypassing these parts of the small intestine, which reduces calcium and vitamin D intake, ultimately leading to calcium deficiency (44). Bone mineral density typically decreases after BS due to altered mechanical loading of the bone, whereas parathyroid hormone production increases. If parathyroid hormone continues to rise and hyperthyroidism occurs, it can lead to osteopenia and osteoporosis. Therefore, metabolic bone disease is a long-term risk factor after BS bypass surgery (45). In contrast, SG and LAGB did not alter duodenal traits, so calcium and magnesium deficiencies were less frequent.

After RYGB, iron deficiency was significantly higher in people with obesity than with SG and LAGB (46). This is because iron needs to be absorbed in the duodenum and proximal jejunum, with the participation of gastric acid (47). Divalent iron is better absorbed than trivalent iron when supplementing with iron after surgery (48). This is because ferrous iron is usually found in animal foods, while ferric iron is found in plant foods (49). Ferric iron needs to be reduced to ferrous iron under the action of gastric acid before it can be absorbed by the intestinal tract (50). Therefore, it is recommended to consume animal food after BS. RYGB is also a procedure with a high rate of vitamin D deficiency because it bypasses part of the duodenum and ileum, which is the main site of vitamin D absorption (51). RYGB is more likely to cause vitamin D deficiency than SG, while the defect in LAGB is not significant. For people with obesity, vitamin D deficiency should be detected early and supplemented in time to avoid the deterioration of the disease.

Regardless of the type of BS, it is inevitable that nutritional deficiencies will result. Our primary goals are to find the most appropriate procedure for people with obesity, minimize complications and nutritional losses, and monitor and supplement postoperatively. When selecting a procedure, people with obesity and doctor must also consider specific differences in postoperative nutritional needs and care (52).

In this study, we found that nutritional deficiencies after BS are widespread, although data from large, robust long-term RCTs are lacking. More studies are needed, such as direct comparisons of the effects of different BS procedures on clinically relevant nutritional deficiencies during the long-term, follow-up-especially for newer procedure types. Studying postoperative nutritional deficiencies in different subgroups (e.g., those with other diseases, special populations, or weaker social determinants of health) of BS should also be a high research priority.

Because our goal was to compare nutritional deficiencies after three commonly used bariatric surgeries at the same time (which can only be accomplished by constructing a network meta-analysis), we included surgical options in our network meta-analysis to help clinicians reduce the incidence of nutritional deficiencies in patients. We included all studies that fulfilled the inclusion criterion for the absence of risk. Small studies, to be sure, can introduce a lot of bias since there is greater uncertainty in the estimations and the outcomes are more likely to be random.

However, the quantity and quality of the identified evidence limit the strength of our conclusions. Only a few studies had sufficiently long follow-ups; however, we also note that the double-blind requirement may have artificially reduced the quality of the studies, which is rarely performed in trials evaluating surgical interventions. In addition, only a few studies reported patient undernutrition at baseline, but we attempted to address this limitation by performing a sensitivity analysis of these studies. Because this was an interim sensitivity analysis, the results should be interpreted cautiously. We included this part of the study in the heterogeneity study. For this study, we first noticed that there was no heterogeneity in these trials (I²= 0%) and that patients had nearly the same rate of the nutritional deficit as the group reporting baseline values. Secondly, In these trials, the follow-up time was so extensive that researchers may have thought that baseline levels could be ignored during long-term follow-up. As a result of these considerations, we believe that incorporating these trials in the network meta-analysis would have no detrimental consequences.

We followed PRISMA, developed a review protocol in advance, and mapped Brooks-Gelman-Rubin diagnostics charts. We used Mesh, a comprehensive literature search strategy, quality assessment and data extraction by repeat reviewers, and rigorous statistical methods to reduce potential bias. Given that the medical tools used to perform BS have advanced over time, there is very little literature on direct comparisons between different surgical approaches. However, we continue to use state-of-the-art, rigorous natural comparison methods to estimate the relative effectiveness of each procedure compared to others. Therefore, we are confident that our results are valid.