Cumulative Incidence of Venous Thromboembolic Events In-Hospital, and at 1, 3, 6, and 12 Months After Metabolic and Bariatric Surgery: Systematic Review of 87 Studies and Meta-analysis of 2,731,797 Patients

Abstract Systematic review/meta-analysis of cumulative incidences of venous thromboembolic events (VTE) after metabolic and bariatric surgery (MBS). Electronic databases were searched for original studies. Proportional meta-analysis assessed cumulative VTE incidences. (PROSPERO ID:CRD42020184529). A total of 3066 records, and 87 studies were included (N patients = 4,991,683). Pooled in-hospital VTE of mainly laparoscopic studies = 0.15% (95% CI = 0.13–0.18%); pooled cumulative incidence increased to 0.50% (95% CI = 0.33–0.70%); 0.51% (95% CI = 0.38–0.65%); 0.72% (95% CI = 0.13–1.52%); 0.78% (95% CI = 0–3.49%) at 30 days and 3, 6, and 12 months, respectively. Studies using predominantly open approach exhibited higher incidence than laparoscopic studies. Within the first month, 60% of VTE occurred after discharge. North American and earlier studies had higher incidence than non-North American and more recent studies. This study is the first to generate detailed estimates of the incidence and patterns of VTE after MBS over time. The incidence of VTE after MBS is low. Improved estimates and time variations of VTE require longer-term designs, non-aggregated reporting of characteristics, and must consider many factors and the use of data registries. Extended surveillance of VTE after MBS is required. Graphical Abstract


Introduction
Metabolic and bariatric surgery (MBS) is an effective approach for achieving weight loss and resolution of obesityassociated medical problems among patients with obesity [1,2].As with any surgical procedure, there is the risk of postoperative venous thromboembolic events (VTE) after MBS, which is exacerbated by the underlying obesity [3].With the increasing global frequency of MBS [4], VTE presents a particularly serious complication with significant effects on readmission rates and mortality [3,5,6].
The literature has shown wide variations in the incidence of VTE after MBS [7,8].For instance, previous studies have reported 30-day incidences ranging from 0 to 5.66% [9,10].Although the number of meta-analyses on many MBS topics continues to increase [11][12][13], there are no systematic reviews/meta-analyses of the incidence of VTE after MBS.This is despite the available literature that could be metaanalyzed to generate high-quality estimates [5,7,10,[14][15][16][17].To date, global estimates of VTE at different timepoints after MBS remain uncertain, despite the calls for more accurate estimates [7].The current study is the first to bridge this knowledge gap.

Aim of the Study
The present study aimed to review and synthesize the evidence on the incidence of VTE after MBS.The objectives were to (1) compute global cumulative incidence of VTE at five timepoints after surgery (in-hospital, at 30 days and 3, 6, and 12 months) for studies that utilized mainly laparoscopic approach and those that used predominantly open surgical approach and (2) for the first 30 days investigate the proportions of VTE that occurred in-hospital vs post-discharge.In addition, we subgrouped the studies at each timepoint by two variables, namely, geographic origin and study age (final year of data acquisition) to explore potential sources of heterogeneity, and appraise whether such factors influenced the incidence of VTE.Evaluation of procedure-or patientrelated risk factors or prophylaxis and their associations with VTE were not within the scope of this review.

Outcomes
The outcomes were cumulative incidence of VTE at five timepoints (in-hospital, 30 days, and 3, 6, 12 months) and for the first 30 days, the proportions of VTE that occurred in-hospital vs post-discharge.

Risk of Bias Assessment
Potential risk of bias was assessed using the tool by Loney et al. [19], which was developed specifically for studies of prevalence/incidence.The tool was selected for its comprehensiveness and applicability to the study objectives.It comprises eight equally weighted items yielding a maximum score of eight, with higher scores indicating lower risk of bias.Included studies were scored by ML, and then 10% was randomly selected and scored by WEA.Mean percentage agreement across the eight individual items was reported.

Data Synthesis and Statistical Analysis
All studies were cross-checked for duplicated use of data by verifying their data sources (hospital or national/ regional registries), sampling timeframe, and included procedures.Where duplicate use of patient data was suspected, only the studies that minimized any overlap were included in the meta-analyses.As many of the included studies were undertaken using large administrative datasets such as NSQIP, MBSAQIP, or NIS, multiple studies included in the same year/s of data from the same registry were meticulously securitized for their procedures, patient samples, and recruitment years, in order to check, confirm, and exclude any potential duplicate use of the data of the same patients.If there was any remaining doubt, the research team undertook the extra step of contacting the authors of such papers for more verification.
Random effects proportional meta-analyses of VTE at the five timepoints were conducted using MetaXL (Epi-Gear, international Pty Ltd., Queensland, Australia) for Microsoft Excel.Data were transformed using the double arcsine method.This allows inclusion of zero-case studies, stabilizes variance, and has demonstrated advantages [20].Additionally, categorical meta-analysis assessed pre-versus post-discharge 30-day incidence and was expressed as a proportion of the total number of cases.
Results were presented by surgical approach as pooling both (laparoscopic and open) approaches was deemed inappropriate because most procedures are currently undertaken laparoscopically.Most studies reported a mix of laparoscopic and open approaches; hence, cut-offs were required.As approximately half of the studies that used a majority open approach reported it for 50-80% of their procedures, and almost all studies that used a majority laparoscopic approach reported it for > 80% of their procedures, we subgrouped studies into " > 80% laparoscopic approach" vs " > 50% open approach."Furthermore, subgroup analyses were conducted on cumulative incidences at each timepoint to identify any influence of the subgroups on the pooled estimates, and to assess sources of heterogeneity.In terms of study age, we categorized studies into those with data collected up to the end of 2010 vs after 2010, as an earlier cut-off was not feasible due to a lack of relevant studies.Geographically, it was only feasible to subgroup studies into North America vs "other" countries, as roughly 70% of studies were from North America.This latter comparison was limited to studies with > 80% laparoscopic approach to minimize possible confounding due to surgical approach.Since small samples have potentially lower sensitivity to capture VTE, sensitivity analysis was undertaken excluding the small studies (n < 2000 patients) to assess its influence on pooled incidence of the geographical subgroups.

Heterogeneity
Heterogeneity was measured using Higgin's I 2 , Cochrane's Q, and Chi 2 .Given the nature of incidence data, high heterogeneity was expected due to large sample sizes and low variance.Therefore, thresholds for heterogeneity [21] were interpreted conservatively in line with recommendations regarding proportional meta-analysis [22].

Publication Bias
We used funnel plots based on sample size (rather than standard error) as they have been shown to be a valid alternative for assessing publication bias in proportional metaanalysis [23], since traditional funnel plots may indicate asymmetry when no publication bias is present [22,23].In addition, recent guidelines recommend qualitative methods to appraise publication bias of incidence data [22].Hence, we assessed publication bias using a combination of both.

Search Results
The PRISMA diagram (Fig. 1) shows that of 3066 retrieved articles, 87 were included in the review [5,6,9,10,[14][15][16][17], of which 68 were meta-analyzed.The studies excluded at full text and their reasons, as well as the included studies, and their subgroupings are available in Supplementary File 2.

Study Characteristics
Table 1 outlines the studies included in the review (N = 4,991,683 patients).A total of 2,259,886 patients were subsequently excluded from meta-analyses due to data overlap or aggregated data.Data of the remaining 2,731,797 patients were meta-analyzed.The largest study included 540,959 patients [31] and the smallest comprised 39 patients [60].
Seventy-three studies (83.9%) reported the sex distribution of their sample, with females comprising a mean of 77.6%.Fourteen (16.1%) did not report sex distribution.Seventy-two studies (82.8%) reported the mean age of Lins 2015 66  Mabeza 2022 67  Magee 2010 68  Masoomi 2011 69  McCullough 2006 70  Miller 2004 71   Thereaux 2018 97  Thereaux 2014 98  Westling 2002 99  Winegar 2011 100 * Woo 2013 101  Young 2015 102 * Due to space limitations, only the first author is cited; m: months; d: days; Grey shaded boxes indicate that data was explicitly provided or was able to be calculated; * Study was excluded from all meta-analyses of this time point due to significant overlap in data with other studies; † Study excluded from at least one metaanalysis for this timepoint due to overlap or insufficient information about surgical approach or sampling timeframe, but included in others; ‡ Patients were scanned 2 to 3 days post-surgery, data was included as inhospital; § Timeframe for incidence data was 15-months, but was included in the 12-month analysis their sample, with an average of 42.9 ± 3.02 years, while seven (8%) provided the median age for the sample, and eight studies (9.2%) did not report age.Sixty-four studies (73.6%) reported mean BMI (46.1 ± 6.0 kg/m 2 across all studies), six (6.9%) provided their median BMIs, while 17 studies (19.5%) did not report BMI.
Table 2 shows the time point/s of incidence provided by each included study (i.e., the specific meta-analysis/es that each study contributed to, as well as the studies that were excluded from primary (not sub-grouped) meta-analysis of any given time point due to significant overlap in data with other studies.Data sources of each included study are outlined in Supplementary File 3.

Risk of Bias Appraisal
The mean risk of bias score was 5.82 ± 1.43, with a range of 3-8 (Supplementary File 4).Fifty studies (57.5%) scored six or higher, indicating a low risk of bias.The 20 studies with lower scores of 3-4 were mainly due to small sample

Meta-analysis
The summary of findings of the meta-analyses at the different time points and their subgroupings is depicted in Table 3. Below, we detail the findings at each time point individually.

Incidence of VTE Within 30 Days: In-hospital vs Post-Discharge
Meta-analysis of 11 studies that reported both in-hospital and 30-day incidence (Supplementary File 5) showed that 60% (95% CI 57-63%; I 2 = 88.16%) of the 30-day VTE occurred after discharge, based on 1073 events.

VTE Over Time
Cumulative incidence of VTE over time is depicted in Fig. 7. Incidence generally increased up to the last timepoint examined (12 months post-MBS).Incidence for > 80% laparoscopic approach was consistently lower compared to the > 50% open approach (Fig. 7A).Subgroup analyses displayed variations across time; incidence from North American studies was higher for most timepoints (based on > 80% laparoscopic procedures only) (Fig. 7C).Sensitivity analysis removing studies with sample sizes < 2000 patients increased the incidence for both subgroups and largely accounted for differences at 30 days (North America 0.43% vs other 0.31%) and 3 months (North America 0.49% vs other 0.51%), but not at 6 months (North America 1.83% vs other 0.48%).Sensitivity analysis was not possible for 12-month data.

Publication Bias
Figure 8 depicts the funnel plots of cumulative incidence of VTE.At some timepoints, more studies reported lower incidence (Fig. 8A-C), and there was a relative paucity of Fig. 3 Thirty-day cumulative incidence of venous thromboembolic events.Forest plot showing: A > 80% laparoscopic and > 50% open; B pooled results by two subgroupings-country (North America vs other countries, limited to studies comprising > 80% laparoscopic surgical approach to minimize confounding from surgical approach) and year (last year of data inclusion before and including 2010 vs after 2010, not limited by surgical approach).Square data points: incidence from individual studies; diamond-shaped data points: pooled values from subgroups; hexagonal data points: pooled values from all studies that reported relevant data ◂ studies of moderate sample sizes; hence, studies clustered at the upper (larger sample sizes) and lower (smaller sample sizes) ends of the Y axis (Figs.8C-E).Qualitatively, countries outside of North America were underrepresented.Roughly three quarters of the studies reported North American data, with many using data registries.For instance, 28 (40.58%) of 69 studies reporting 30-day incidence used North American registry data, introducing considerable overlap of patient data across studies.Figure 8 B shows an unusual 'stacking' pattern of very similar incidences of VTE suggesting the duplicate use of patient data by different studies.

Discussion
Patients with obesity are at risk of VTE in the post-MBS period [8,100], and those who develop VTE have an increased risk of mortality [5,6].Despite this, no previous B pooled results by two subgroupings-country (North America vs other countries, limited to studies comprising > 80% laparoscopic surgical approach to minimize confounding from surgical approach) and year (last year of data inclusion before and including 2010 vs after 2010, not limited by surgical approach).Square data points: incidence from individual studies; diamond-shaped data points: pooled values from subgroups; hexagonal data points: pooled values from all studies that reported relevant data study has meta-analyzed the incidence of VTE after MBS.The present systematic review and meta-analysis presented high-quality cumulative incidences of VTE pooled from nearly 5 million patients worldwide.The in-hospital, 30-day, and 3-, 6-and 12-month incidences provide clinically relevant and meaningful information regarding the timing and patterns of VTE, to guide the follow-up, detection, and prevention of this life-threatening complication.The review also explored the influence of surgical approach, geographical origin, and study age on incidence of VTE.To our knowledge, this is the first study to undertake such a task.In terms of the incidence of VTE at the five timepoints under examination, our observed cumulative incidence of VTE exhibited an increasing trend in-hospital, and at 30 days and 3, 6 and 12 months, for the > 80% laparoscopic approach (0.15%, 0.50%, 0.51%, 0.72%, and 0.78% respectively) and for the > 50% open approach (0.43%, 2.02%, 2.14%, 2.36%, and 3.38% respectively).Such increasing pattern is consistent with two studies that reported incidences of VTE after MBS of 0.88% in-hospital , 2.17% 1 month , and 2.99% 6 month [93] and 0.3% 7 days , 1.9% 30 days , 2.1% 3 months , and 2.1% 6 months [43].Therefore, MBS patients require clinical vigilance to continue for an extended period, in order to identify VTE and reduce the risk of morbidity and mortality.
Individual studies reported a wide range of incidences at each timepoint.Such variations could be due to patient features such as age, BMI, or comorbidity [103]  B pooled results by two subgroupings-country (North America vs other countries, limited to studies comprising > 80% laparoscopic surgical approach to minimize confounding from surgical approach) and year (last year of data inclusion before and including 2010 vs after 2010, not limited by surgical approach).Square data points: incidence from individual studies; diamond-shaped data points: pooled values from subgroups; hexagonal data points: pooled values from all studies that reported relevant data characteristics, such as operative time [103], MBS procedure, or surgical approach [100,104,105]; or study characteristics, such as study design, years of data acquisition, and sample size.Across studies that reported both in-hospital and 30-day VTE, 60% of events occurred after discharge, concurring with previous reports where up to 80% of VTE occurred after discharge [52,93,100].Higher post-discharge incidence of VTE might be partly attributed to short in-hospital stays of only a few days [106,107], compared to longer post-discharge periods.Similarly, we found that most VTE occurred within the first 30 days, consistent with observations from some of the included studies [43,52].This further highlights the importance of vigilance during this period.
The present study noted that cumulative incidence across the > 80% laparoscopic studies was consistently lower than the > 50% open approach for all timepoints, consistent with previous findings at 30 days [105], 90 days [100], and 5 years [104].Notwithstanding, some literature has demonstrated no differences in VTE outcomes between laparoscopic vs open approaches [108][109][110].
As for the subgroup analyses, we explored the effects of study age and geographical origin.
Studies using data up to the end of 2010 demonstrated higher incidences at most timepoints, compared to more recent studies, likely due to the larger proportion of > 50% open approach studies in the former subgroup.Factors that have contributed to the reduction in VTE since the turn of the century include the shift from open to laparoscopic approaches, MBS technical advancements, pre-/post-surgery thromboprophylaxis, and enhanced recovery regimens [68,[111][112][113][114][115].
To explore geographical differences, we compared North American studies to other countries.Despite limiting this to > 80% laparoscopic studies to minimize confounding from surgical approach, incidence from North American studies open; B pooled results by two subgroupings-country (North America vs other countries, limited to studies comprising > 80% laparoscopic surgical approach to minimize confounding from surgical approach) and year (last year of data inclusion before and including 2010 vs after 2010, not limited by surgical approach).Square data points: incidence from individual studies; diamond-shaped data points: pooled values from subgroups; hexagonal data points: pooled values from all studies that reported relevant data was higher for most timepoints.Sensitivity analysis removing studies with less than 2000 patients increased the incidence of the other countries group closer to that of North American studies at 30 days and 3 months, the timepoints where both subgroups used large samples from registry data, indicating the influence of sample size.
The current review identified only one study that assessed outcomes beyond the first few years after MBS [16].This study found that over a median of 10.7 years post-surgery, MBS patients exhibited significantly less VTE compared to non-MBS patients matched for sex, age, and baseline BMI [16].This suggests that despite our observed shorter-term incidence of VTE, MBS appears to offer "protection" (e.g., decreases in BMI), resulting in lower long-term risk of VTE [7].Future research should include longer-term assessment of VTE after MBS.Collectively, the above suggests that a deeper understanding of the variations in VTE across time must consider the interrelationships between surgical approach (and hence study age) and sample size (and hence the use of data registries and geographical origin), amongst other factors.In terms of the quality of estimates, risk of bias within and across studies and heterogeneity, slightly more than half of the included studies exhibited low risk of bias.Some of the studies displaying higher risk of bias were due to small sample sizes, potentially biased sampling frames, or poor reporting.North American studies were over-represented, with many utilizing large national/regional registries.This led to considerable overlap of patient data, which increased our efforts to identify and exclude overlapping data to ensure the validity of the metaanalysis.Heterogeneity in the overall meta-analyses was high at all timepoints.Subgrouping reduced some heterogeneity; however, it remained high for the > 80% laparoscopic approach and the North American subgroups, both of which included the studies with the largest sample sizes and lowest variance.This is consistent with others who noted that measures of heterogeneity such as Higgin's I 2 may indicate high heterogeneity in proportional meta-analysis, even when data are consistent [22].
This review has some limitations.Many studies reported 30-day incidence, while others reported inconsistent timepoints, rendering interpretations of incidence across individual studies difficult.However, this variation enabled us to assess cumulative incidence and its patterns over time.Additionally, as most studies were retrospective, based on patient charts/records, pooled incidences are likely to reflect symptomatic VTE.As it was only possible to use the > 80% laparoscopic and > 50% open subgroups, this would have resulted in some contamination within the subgroups, suggesting that our observed VTE differences between surgical approaches could be underestimated.It would have been beneficial to include elements of the prophylaxis undertaken as well as operative time in the analysis.However, the extent of non-reporting, aggregated or undetailed reporting of these items, and in the case of prophylaxis, the numerous and wide variations in the chemo/mechanical prophylaxis protocols used singly or in combination at different times and durations of administration, with or without inferior vena cava filters, transfusions, or stoppage of chemical thromboprophylaxis where required would result in countless combinations thereof, which mitigated against a meaningful analysis.Notwithstanding, some of the included studies reported that duration of surgery for patients who experienced VTE after MBS was longer than that of matched control patients [29], and that operative time was significantly longer in patients who experienced a post-operative VTE [52] and a significant predictor of or associated with of post-operative VTE [40,44].
Future studies would benefit from prospective designs, better (non-aggregated) reporting of sample/procedure characteristics and timeframes, assessment of longer-term VTE, and greater representation from outside of North America.Future meta-analyses should be aware of studies utilizing large national/regional registries that could lead to considerable overlap of patient data.Future researchers should be mindful of the differences across data registries when conducting research to ensure that significant proportions of events are not missed.The current study clearly demonstrated that the time course of VTE post-surgery is dynamic.As such, researchers presenting primary research on such complications need to clearly relate reported incidences to a given timeframe postsurgery, and those synthesizing such studies should be careful not to aggregate incidences related to different timeframes, since this would render any reported values meaningless.
This study has many strengths.We assessed the pooled incidence of VTE after MBS at five timepoints.Subgroup analysis included surgical approach, geographical origin of the studies, and study age.We meticulously identified potential overlap of patient data, including that from large registries, and excluded such studies from the meta-analysis, enhancing the internal validity [116].The extremely large number of patients worldwide enhances the external validity and generalizability of the findings.To our knowledge, this is the most extensive and comprehensive systematic review/meta-analysis of VTE after MBS over several timepoints that has been undertaken, and probably the largest systematic review/meta-analysis conducted to date in the field of surgery/health in general in terms of the number of patients.

Conclusion
We pooled a large number of studies and patients worldwide to provide high-quality estimates of VTE and valuable insights into its patterns over time.For studies that utilized a mainly laparoscopic approach, in-hospital incidence of VTE and cumulative incidence at 30 days and 3, 6 and 12 months were 0.15%, 0.50%, 0.51%, 0.72%, and 0.78% respectively.Most VTE occurred in the first 30 days, of which 60% was after discharge, although we observed some VTE up to our last timepoint.Incidence was consistently lower for laparoscopic compared to open MBS.Lower incidences from studies outside of North America were largely due to smaller sample sizes.Deeper understanding of the variations in VTE across time must consider the interrelationships between surgical approach, geographical origin, study age, and sample size, amongst other factors.Post-operative surveillance needs to be particularly vigilant after discharge and continue thereafter for an extended period to detect VTE and reduce the risk of associated morbidity and mortality.These findings provide clinically relevant estimates of VTE to inform policy, clinical practice, and research.collection and acquisition of datasets.ML and WEA provided the data analysis.WEA, ML, and KE-A were responsible for writing and revising the manuscript.AE-M and AAA edited the manuscript.The authors also critically reviewed and approved the final version of this paper.
Funding Open Access funding provided by the Qatar National Library.

Fig. 1
Fig. 1 PRISMA flowchart of search and screening results

Fig. 5
Fig.5 Six-month cumulative incidence of venous thromboembolic events.Forest plot showing: A > 80% laparoscopic and > 50% open; B pooled results by two subgroupings-country (North America vs other countries, limited to studies comprising > 80% laparoscopic surgical approach to minimize confounding from surgical approach) and

Fig. 7
Fig. 7 Total and sub-grouped cumulative incidence of VTE after metabolic and bariatric surgery across time: A by surgical approach (> 50% open vs > 80% laparoscopic), B by study age (up to and

Fig. 8
Fig. 8 Panel of Funnel plots of all included studies presenting data for cumulative incidence of venous thromboembolic events: A in-hospital, B 30-day, C 3-month, D 6-month, and E 12-month.Solid black

Table 1
Characteristics of the 87 original studies included in the current review(4,991,683 patients) Due to space limitations, only the first author is cited BPD biliopancreatic diversion, D duration, DS duodenal switch, GB gastric bypass, GBn gastric banding, L laparoscopic, BMS bariatric/metabolic surgery, N number of patients, O open approach, RYGB Roux-en-Y GB, SG sleeve gastrectomy, y years, M multiple, Md median, R robotic, SIPS Stomach Intestinal Pylorus-Sparing Surgery, -not explicitly reported/ cannot be computed a Surgical approach not explicitly reported, assumed to be mostly open due to the sampling time frame being 1977-1993 b Time frame not explicitly reported, assumed to be before 2010 as publication date was before 2010

Table 2
Timeframe of incidence provided by each included study

Table 3
Summary of findings Items 2 and 4 had the lowest number of studies receiving a score for them (50.57% and 42.53%, respectively).Average inter-rater agreement for the 10% of the studies randomly selected was 79.17% ± 17.25 across the nine items.