Introduction

As the global prevalence of obesity continues to rise, it is expected that by 2050, nearly one in three adults over the age of 25, approximately 1.95 billion individuals, will be living with obesity.1. Bariatric surgery is a well-established and effective long-term treatment for individuals with severe obesity2. However, nearly 20% of patients show a suboptimal clinical response3,4. Although bariatric procedures are minimally invasive with comparable risks to common operations like cholecystectomy and appendectomy5, they may lead to significant physical and mental consequences6,7,8,9. Patients undergoing bariatric surgery endure the challenges with the expectation of achieving significant weight loss or improvement of comorbidities. Therefore, ensuring that patients achieve an optimal clinical response is critical. To better define optimal response after bariatric surgery, the International Federation for Surgery of Obesity and Metabolic Disorders (IFSO) recently proposed a consensus definition for suboptimal clinical response: a total body weight loss (TWL) of less than 20%, or insufficient improvement in an obesity-related complication that was a key indication for surgery10.

Although adherence to dietary guidelines, regular physical activity, and regular long-term follow-ups are associated with better weight loss outcomes after surgery, some patients still experience a suboptimal clinical response11. In this context, total weight loss (TWL) percentile charts have been presented as valuable tools for bariatric surgery, offering extensive information on patients’ typical weight loss12,13,14,15. These charts enable a quick evaluation of a patient’s weight loss at each follow-up solely on the TWL from that visit and establish neutral benchmarks for determining optimal weight loss over time15. However, TWL percentiles have not yet been investigated for their ability to predict suboptimal clinical response. By visualizing patients’ TWL percentiles during early follow-ups, these charts may help us identify patients with suboptimal clinical response earlier. Tracking a patient’s position on these charts during early follow-ups may allow for earlier detection of suboptimal clinical response, enabling timely assessments, closer monitoring, and targeted interventions to improve long-term results.

This study aims to examine the relationship between TWL percentiles and suboptimal clinical response and evaluate the predictive performance of TWL percentiles for suboptimal clinical response after one-anastomosis gastric bypass (OAGB), Roux-en-Y gastric bypass (RYGB), and sleeve gastrectomy (SG), using our previously developed TWL percentile charts12,13. Determining the best percentile thresholds for predicting suboptimal clinical response at each follow-up visit and identifying the optimal time for prediction helps us to identify suboptimal clinical response earlier and intervene appropriately.

Material and methods

This retrospective cohort study included individuals who underwent OAGB, RYGB, or SG from 2008 to 2020. All patients aged 18–70 years old with a BMI ≥ 35 kg/m2 who underwent OAGB, RYGB, or SG as primary bariatric surgery by two expert surgeons and were followed up for two to three years were included in this study. Data was retrieved from the Iran National Obesity Surgery Database16. A multidisciplinary team, including general practitioners, nutritionists, psychologists, sports medicine specialists, and surgeons, followed up the patients 10 days, 1, 3, 6, 9, 12, 18, and 24 months after surgery.

The primary outcome of this study was to assess the %TWL percentile of individuals with suboptimal clinical response following bariatric surgery, using our previously presented total weight loss (TWL) percentile charts for SG13, OAGB, and RYGB12. Our secondary outcome was to present a percentile threshold that could help predict suboptimal clinical responses early after different types of surgery to enhance the timely identification of these patients.

Based on the recent IFSO consensus on definitions and clinical practice guidelines for obesity management, suboptimal clinical response is considered a total body weight loss of less than 20% or inadequate improvement in an obesity complication that was a significant indication for surgery10. In the current study, suboptimal clinical response was defined as a %TWL < 20 at the first follow-up visit occurring between 24 and 36 months after surgery. %TWL was calculated as (initial weight − post-op weight) * 100/(initial weight).

Statistical analysis

Descriptive analyses were conducted to summarize the baseline characteristics of participants. Continuous variables were reported as means with standard deviations, and categorical variables as frequencies and percentages. Group comparisons were performed using chi-square tests for categorical variables and analysis of variance (ANOVA) for continuous variables.

We implemented an iterative logistic regression approach to evaluate the predictive performance of TWL percentiles for suboptimal clinical response. At each post-surgical visit, we defined the exposure variable as being below a given TWL percentile threshold. In contrast, the outcome variable was suboptimal clinical response at the end of the study period. Separate logistic regression models were fitted for each percentile threshold at each time point, and the odds ratios (OR) with 95% confidence intervals (CI) were reported to quantify the association between early weight loss and suboptimal clinical response risk.

We constructed receiver operating characteristic (ROC) curves to assess discriminative ability and calculated the area under the curve (AUC) for each model. The AUC values and their corresponding 95% CI were used to determine the percentile-time combination with the highest predictive performance. The optimal percentile threshold at each time point was identified as yielding the highest AUC. All statistical analyses were performed using R software (version 4.2.2). A two-tailed p-value < 0.05 was considered statistically significant.

Ethics approval

This study was approved by the Research Ethics Committees of Iran University of Medical Sciences (IR.IUMS.REC.1397.464). This study was conducted in accordance with the Declaration of Helsinki. All participants signed informed consent for their data on the Iran National Obesity Surgery Database to be used in future studies.

Results

We included 5,780 individuals in our study, of whom 2,579 underwent OAGB, 1,943 RYGB, and 1,258 SG. The mean age of participants was 37 years for SG and 40 years for both OAGB and RYGB. Patients’ baseline characteristics are summarized in Table 1. The majority were female, with the proportion of women being highest in the RYGB group (84%), followed by SG (76%) and OAGB (75%). Most patients had a baseline BMI in the 40–50 kg/m2 range, with the mean baseline BMI being highest in the OAGB group (46.9 kg/m2), followed by RYGB (43.9 kg/m2) and SG (43.8 kg/m2). Similarly, the highest prevalence of type 2 diabetes was observed in the OAGB group (23%), followed by RYGB (15%) and SG (7.4%).

Table 1 Baseline characteristics of patients.
Full size table

The prevalence of suboptimal clinical responses was 1.83% for SG, 1.39% for RYGB, and 1.05% for OAGB. As Fig. 1 illustrates, patients with suboptimal clinical response were more concentrated below the 25th percentile from three months onward, following SG, RYGB, and OAGB. However, some cases fell at even higher percentiles early in the follow-up period, becoming fewer as time progressed to later follow-ups. In all three procedures, weight loss follows a rapid decline within the first 6–12 months, after which the rate of change stabilizes. The weight loss trajectory for patients undergoing SG appeared more gradual than those undergoing bypass procedures. Following RYGB, the weight loss trajectory followed a structured decline before stabilizing, but there was slightly more variability in weight loss trajectories for OAGB. This graphical presentation provides an overview of the relationship between TWL percentiles and suboptimal clinical response. However, determining the best percentile-time threshold for predicting suboptimal clinical response requires a more detailed statistical approach. To address this, we conducted a more specific analysis using logistic regression models and ROC curves, as described in the next section.

Fig. 1
figure 1

Total Weight Loss percentiles over time in patients experiencing a suboptimal clinical response following Sleeve Gastrectomy, Roux-en-Y Gastric Bypass, and One-Anastomosis Gastric Bypass. (Red dots indicate individuals with suboptimal clinical response).

Full size image

Predictive performance of TWL percentiles for suboptimal clinical response

To identify the most effective percentile-time thresholds for predicting suboptimal clinical response, we conducted an iterative analysis using logistic regression models, where the exposure variable was below a given TWL percentile at a specific follow-up visit and the outcome was suboptimal clinical response at the end of the study. The predictive power of these thresholds was assessed using ROC curves and AUC values, as shown in Figs. 2–4 for SG, RYGB, and OAGB, respectively.

Fig. 2
figure 2

AUC values for predicting suboptimal clinical response based on Total Weight Loss percentiles at different follow-ups after Sleeve Gastrectomy.

Full size image

Figure 2 presents the AUC values for different TWL percentiles at various post-surgical time points in SG patients. The predictive performance of TWL percentiles improved over time, with AUC values generally increasing from early visits to later follow-ups. At 10 days and 1 month post-surgery, AUC values were relatively low, indicating limited predictive ability. However, by 6 months, the predictive performance increased substantially, with AUC values peaking at 9 months and 1 year. This suggests that weight loss percentiles at these time points provide the most informative thresholds for suboptimal clinical response prediction. Similarly, Figs. 3 and 4 display the predictive performance of TWL percentiles for suboptimal clinical response in RYGB and OAGB patients, showing a similar pattern. As with SG, the predictive ability of early TWL percentiles for RYGB and OAGB was initially limited but improved progressively over time.

Fig. 3
figure 3

AUC values for predicting suboptimal clinical response based on Total Weight Loss percentiles at different follow-ups after Roux-en-Y Gastric Bypass.

Full size image
Fig. 4
figure 4

AUC values for predicting suboptimal clinical response based on Total Weight Loss percentiles at different follow-ups after One-Anastomosis Gastric Bypass.

Full size image

At each follow-up, following all procedures, as TWL percentiles increase, the AUC increases before reaching a peak. The decline in AUC after reaching its peak may be due to the stabilization of weight loss, reducing the variability in TWL percentiles and making them less predictive of long-term suboptimal clinical response risk. The pattern observed in the AUC curves, an initial rise as TWL percentiles increase, followed by a decline, reflects the evolving nature of weight loss trajectories after bariatric surgery. In the early postoperative period, weight loss is highly variable and not yet strongly associated with long-term outcomes, limiting predictive accuracy. As time progresses and inter-individual differences in weight loss become more pronounced, TWL percentiles gain predictive value, resulting in higher AUCs. However, once weight stabilizes and most cases of suboptimal clinical response have emerged, the added predictive value of percentiles diminishes, leading to a decline in AUC.

Optimal percentile thresholds for predicting suboptimal clinical response

To determine the most effective percentile thresholds for predicting suboptimal clinical response at each post-surgical visit, we identified the percentile with the highest predictive power at each time point using logistic regression models. Table 2 presents the optimal percentile thresholds, their associated odds ratios (ORs) with 95% confidence intervals (CIs), and the AUC values for SG, RYGB, and OAGB patients.

Table 2 Optimal Total Weight Loss (TWL) percentile thresholds for predicting suboptimal clinical response at different post-surgical visits across Sleeve Gastrectomy, Roux-en-Y Gastric Bypass, and One-Anastomosis Gastric Bypass.
Full size table

At early post-surgical visits, the optimal percentile thresholds varied across procedures, with modest AUC values indicating limited predictive power. At 10 days, the 41st percentile was the best predictor for SG, the 33rd percentile for RYGB, and the 25th percentile for OAGB, though their predictive strength was relatively low, with AUC values near 0.6. By 1 month, the predictive ability improved slightly, with the 44th and 43rd percentiles emerging as the optimal threshold for SG and RYGB (AUC 0.65), while it remained weak for OAGB (AUC 0.57).

At 3 months, the optimal percentiles were 15th for SG, 56th for RYGB, and 30th for OAGB, indicating that patients with TWL below these levels were at significantly higher risk of suboptimal clinical response. The ORs for SG, RYGB, and OAGB at this time point were 5.16, 4.73, and 2.42, respectively, with AUC values ranging from 0.6 to 0.65, suggesting moderate predictive power.

By six months post-surgery, the predictive performance of TWL percentiles improved substantially, with AUC values ranging from 0.71 to 0.79. The optimal thresholds at this time were the 23rd percentile for SG, the 29th for RYGB, and the 33rd for OAGB. SG demonstrated the strongest association with suboptimal clinical response (OR 16.08), followed by RYGB (OR 7.9) and OAGB (OR 5.85). Predictive power peaked at nine months, with AUCs of 0.89 for SG, 0.78 for RYGB, and 0.83 for OAGB. At this point, the most informative percentiles were the 23rd for SG, 16th for RYGB, and 24th for OAGB. The corresponding odds ratios were highest for SG (OR 66.41), followed by OAGB (OR 27.87) and RYGB (OR 14.88), indicating strong predictive value across all procedures.

At one-year post-surgery, the 23rd percentile remained the optimal threshold for SG, while the 20th percentile emerged for RYGB and the 18th for OAGB, with AUC values ranging from 0.85 to 0.86. The ORs remained high, with SG (OR 40.49), RYGB (OR 32.85), and OAGB (OR 37.42), indicating a strong association between 1-year weight loss percentiles and suboptimal clinical response risk.

Discussion

Individuals undergoing bariatric surgery do so with the expectation of achieving an optimal clinical response, despite the inherent challenges and potential complications. Therefore, early identification of patients at risk for suboptimal clinical response is critical to enable timely intervention and improve outcomes. This study evaluated the association between TWL percentiles and suboptimal clinical response across OAGB, RYGB, and SG. Our findings show that, from three months onward, patients with suboptimal clinical response were consistently concentrated below the 25th percentile, suggesting this threshold may serve as a general early warning sign. Six months post-surgery, the 23rd percentile for SG, 29th for RYGB, and 33rd for OAGB emerged as the most predictive thresholds. These findings highlight the potential of TWL percentiles to guide early identification and targeted intervention for patients at elevated risk.

Previous studies have underscored early postoperative weight loss as a critical factor in predicting long-term suboptimal response17. Weight loss within the first 3–6 months is significantly associated with 2-year outcomes18, with over half of the maximum TWL typically achieved within the first 6 months19. A multi-center prospective study in the Netherlands showed that patients with lower TWL at 3 months had lower weight loss trajectories over 5 years20. Additionally, patients who did not reach 66% of their predicted TWL by 6 months failed to achieve their 1-year target weight21. Our findings emphasize the importance of the 6-month follow-up as a critical time point for identifying patients at risk of suboptimal clinical response. By 6 months, the predictive performance of TWL percentiles improved substantially, with the 23rd percentile emerging as the best predictor for SG, the 29th for RYGB, and the 33rd for OAGB.

During the first few weeks after surgery, patients follow a restricted diet, with solid foods started only after one month. This dietary transition can be challenging, and food tolerance varies. By 6 months, most patients’ postoperative symptoms and food tolerances improve, allowing them to eat larger meals again22,23. Thus, adherence to dietary and behavioral recommendations becomes critical in this period. A study on food cravings after bariatric surgery reported a reduction in total food cravings, high-fat cravings, sweet cravings, carb cravings, and consumption of craved foods at 3 months, followed by an increase from 3 to 6 months24. Notably, by 6 months, consumption of high-fat foods did not differ from the pre-operation phase. Therefore, 6 months becomes an important time point for monitoring patients’ diet and providing appropriate interventions through nutritional counseling and behavioral support.

Close observation of patients falling below the percentile thresholds in early follow-ups by a multidisciplinary team may facilitate early detection and management of contributing factors. A wide range of factors may contribute to suboptimal clinical response, including psychological factors, such as depression, anxiety, alcohol and substance abuse, personality disorder, and low self-esteem11,25,26, eating behaviors including binge eating, loss of control eating, and emotional eating11,27,28, low socio-economic status, low social support, low education, physical inactivity, hormonal factors, genetic factors, anatomical and surgical factors11,26, biochemical factors such as low albumin levels and elevated HbA1c29, and older age30. Early identification and effective intervention for these risk factors are essential for enhancing long-term outcomes.

Beyond TWL percentile thresholds, the pattern of weight loss percentiles over time, whether increasing, stable, or decreasing, can provide valuable information on patients’ progress. A downward trend in an individual’s TWL percentile might indicate adverse changes in behavior, lifestyle, adherence, or overall health, potentially indicating the onset of a suboptimal clinical response. Even patients initially above threshold percentiles may be at risk if their percentile consistently declines across follow-up visits. Therefore, it is important to monitor not only TWL percentiles but also patterns of change. We recommend that clinicians pay close attention to patients with high initial percentiles who show a downward trend, as early intervention may help prevent suboptimal clinical responses.

Regular follow-up, especially within the first year after surgery, is important for timely screening of patients and long-term success31. We should educate patients before the operation about the importance of early postoperative weight loss in predicting long-term outcomes and the significance of regular follow-up with a multidisciplinary team for timely detection.

In our study, the predictive performance of TWL percentiles increased over time, with AUC values generally rising from early postoperative visits to later follow-ups. Predictive ability was limited at 10 days and 1 month, but improved significantly by 6 months, peaking at 9 months and 1 year. It’s clear that as weight loss progresses and differences between patients become more apparent, the prediction ability of TWL percentiles improves. However, at later follow-ups (1 year), most patients with suboptimal clinical response have already been identified, reducing the added value of percentiles at that stage. Therefore, 6 months appears to be an optimal time point for earlier identification of suboptimal clinical response, balancing predictive power with clinical utility.

A key strength of our study is the identification of TWL percentile thresholds for predicting suboptimal clinical response among three common bariatric procedures. While most previous studies focus on RYGB or SG, data on OAGB remain limited. Therefore, our findings on OAGB add valuable insight to the literature. However, a limitation of our study is that we were not able to account for all potential confounding variables, such as psychological factors, eating behaviors, and adherence to postoperative protocols, which influence weight loss outcomes. Future studies should incorporate these variables to validate and refine TWL percentile thresholds and present a more comprehensive model. Another limitation is that this study did not include patients without 2-year follow-up data. It is more plausible that individuals with suboptimal clinical responses who were unsatisfied with their outcomes were less likely to return for follow-up visits. This may have resulted in an underrepresentation of suboptimal clinical responses and a lower observed prevalence in our study. Future research should aim to include more diverse and representative populations with more patients experiencing suboptimal clinical response.

Conclusion

In conclusion, TWL percentiles offer a valuable tool for predicting suboptimal clinical response after bariatric surgery. Incorporating TWL percentile monitoring into routine postoperative care may support the timely detection of suboptimal clinical response, allowing for earlier intervention with personalized care strategies, ultimately ensuring better long-term outcomes. Patients with suboptimal clinical response were more frequently below the 25th percentile from three months onward, following all three procedures, indicating this threshold as a general early warning sign. The 6-month follow-up emerged as a critical time point, with the 23rd percentile threshold for SG, 29th for RYGB, and 33rd for OAGB as key predictors of suboptimal clinical response. Patients falling below these thresholds or with consistent declines in their TWL percentile should be closely monitored and considered for early intervention strategies to reduce the risk of suboptimal clinical response. Future research should validate these thresholds in larger, more diverse populations and explore targeted strategies to address modifiable risk factors.