Efficacy of Teduglutide for Parenteral Support Reduction in Patients with Short Bowel Syndrome: A Systematic Review and Meta-Analysis

Teduglutide has been described as an effective treatment for parenteral support (PS) reduction in patients with short bowel syndrome (SBS). However, a quantitative summary of the available evidence is still lacking. PubMed/Medline, EMBASE, Cochrane library, OVID, and CINAHL databases were systematically searched up to July 2021 for studies reporting the rate of response (defined as a ≥20% reduction in PS) to teduglutide among PS-dependent adult patients. The rate of weaning (defined as the achievement of PS independence) was also evaluated as a secondary end-point. Ten studies were finally considered in the meta-analysis. Pooled data show a response rate of 64% at 6 months, 77% at 1 year and, 82% at ≥2 years; on the other hand, the weaning rate could be estimated as 11% at 6 months, 17% at 1 year, and 21% at ≥2 years. The presence of colon in continuity reduced the response rate (−17%, 95%CI: (−31%, −3%)), but was associated with a higher weaning rate (+16%, 95%CI: (+6%, +25%)). SBS etiology, on the contrary, was not found to be a significant predictor of these outcomes, although a nonsignificant trend towards both higher response rates (+9%, 95%CI: (−8%, +27%)) and higher weaning rates (+7%, 95%CI: (−14%, +28%)) could be observed in patients with Crohn’s disease. This was the first meta-analysis that specifically assessed the efficacy of teduglutide in adult patients with SBS. Our results provide pooled estimates of response and weaning rates over time and identify intestinal anatomy as a significant predictor of these outcomes.


Introduction
Short bowel syndrome (SBS), i.e., the condition following surgical removal of a portion of the small bowel until the intestinal length is shorter than 200 cm, could be caused by different underlying diseases such as mesenteric ischemia, Crohn's disease, radiation enteritis, and surgical complications, among others [1][2][3][4][5]. SBS is the most frequent etiology of chronic intestinal failure (CIF) [1][2][3][4][5]. CIF is defined as the reduction in gut function below the minimum necessary for the absorption of macronutrients and/or water and electrolytes, such that intravenous supplementation is required to maintain health [1][2][3]; until 1970 [6], CIF was a condition incompatible with life; this is still true in some countries nowadays. Parenteral support (PS), i.e., parenteral nutrition (PN) and/or intravenous fluid administration, revolutionized the prognosis of patients with CIF [7,8]. Nevertheless, as defined in the following subsection. Post hoc analyses of included studies were assessed for relevance and employed for data extraction if they contained additional information for the assessment of the outcomes of interest. In case of patient overlap between studies to be pooled in the same analysis, the one with the largest sample size was considered.

Outcomes
The primary outcome of interest was the proportion of patients that achieved PS reduction, defined as a ≥20% reduction in PS volume with respect to baseline; these data will be referred to as "response rate" throughout the rest of the manuscript. The secondary outcome was the proportion of patients that achieved complete PS independence; these data will be referred to as "weaning rate" throughout the rest of the manuscript. Comparisons of response and weaning rates according to intestinal anatomy and SBS etiology were performed based on crude rate differences.

Data Extraction
Three authors (C.D.E., F.B., V.P.) independently examined and extracted data from papers which met the inclusion criteria using prespecified data extraction templates. For each eligible study, the following information was collected: (a) first author and publication year; (b) study design; (c) number of subjects; (d) patients' characteristics in terms of demographic data and SBS-related data; (e) time points at which the efficacy of teduglutide was assessed; (f) observed proportion of patients that achieved a ≥20% reduction in PS volume at each considered time-point (response rate); (g) observed proportion of patients that achieved a complete independence from PS at each considered time-point (weaning rate); (h) stratification of response rate and weaning rate according to intestinal anatomy, if available; and (i) stratification of response rate and weaning rate according to SBS etiology, if available.

Risk of Bias Assessment
The risk of bias was independently assessed for each included study by two authors (F.B., S.B.) using the RoB 2 (Risk of Bias) tool for randomized controlled trials [39] and the ROBINS-I (Risk Of Bias In Non-randomized Studies of Intervention) tool for observational studies [40].
The first tool evaluated five domains, which addressed (a) bias arising from the randomization process, (b) bias due to deviations from intended interventions, (c) bias due to missing outcome data, (d) bias in measurement of the outcome, and (e) bias in selection of the reported result. An additional evaluation of overall risk of bias was also carried out as a summary measure. The options for a domain-level risk-of-bias judgement were 'Low', 'Some concerns' or 'High'. The second tool evaluated seven domains, which addressed (a) bias due to confounding, (b) bias in selection of participants for the study, (c) bias in measurement classification of interventions, (d) bias due to deviations from intended interventions, (e) bias due to missing data, (f) bias in measurement of outcomes, (g) bias in selection of the reported result. An additional evaluation of overall risk of bias was also carried out as a summary measure. The options for a domain-level risk-of-bias judgement were 'Low', 'Moderate', 'Serious' or 'Critical', with an additional option of 'Unknown' if sufficient information for judgement was lacking.

Statistical Analysis
Continuous variables and categorical variables were reported as numbers and percentages, respectively. The proportion of patients that achieved the outcomes of interest was assessed as an absolute rate and compared between groups as a rate difference. A random-effect model with inverse-variance weighting was adopted for statistical pooling of the retrieved data [41]. Higgins I 2 statistics and Cochran Q-test were used to assess heterogeneity between studies [42]. Publication bias was quantitatively assessed by Begg's test.
Statistical analysis was performed using R 4.0.3 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria) and STATA 17 (StataCorp, College Station, TX, USA).

Search Results
A total of 4906 records were identified in the initial literature search (accounting for possible duplicates within each electronic database if a record was obtained in multiple search strings). Removal of duplicates led to an overall pool of 435 studies. An accurate title or abstract revision was sufficient to exclude 401 articles as not pertinent or not fulfilling our prespecified inclusion or exclusion criteria. The remaining 34 studies were assessed in full-text for eligibility; 10 of them finally met all criteria for being included in the final analysis [27,28,[43][44][45][46][47][48][49][50]; 2 additional post hoc analyses were employed for data extraction [35,37], as they contained additional and relevant information for the assessment of the outcomes of 3 among the 10 pertinent papers [28,45,49]. The process of study selection is summarized in Figure 1. random-effect model with inverse-variance weighting was adopted for statistical pooling of the retrieved data [41]. Higgins I 2 statistics and Cochran Q-test were used to assess heterogeneity between studies [42]. Publication bias was quantitatively assessed by Begg's test. Statistical analysis was performed using R 4.0.3 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria) and STATA 17 (StataCorp, College Station, TX, USA).

Search Results
A total of 4906 records were identified in the initial literature search (accounting for possible duplicates within each electronic database if a record was obtained in multiple search strings). Removal of duplicates led to an overall pool of 435 studies. An accurate title or abstract revision was sufficient to exclude 401 articles as not pertinent or not fulfilling our prespecified inclusion or exclusion criteria. The remaining 34 studies were assessed in full-text for eligibility; 10 of them finally met all criteria for being included in the final analysis [27,28,[43][44][45][46][47][48][49][50]; 2 additional post hoc analyses were employed for data extraction [35,37], as they contained additional and relevant information for the assessment of the outcomes of 3 among the 10 pertinent papers [28,45,49]. The process of study selection is summarized in Figure 1.  Table 1 summarizes the basic study characteristics. Two studies had a randomized, placebo-controlled, double-blinded design [27,28]; the remaining eight studies were characterized by an observational design [43][44][45][46][47][48][49][50]. Among these, one was actually designed as a randomized controlled trial (RCT) [45], but the randomization and the blindness of treatment assignment was performed between different teduglutide dosing schedules rather than between the teduglutide and placebo; therefore, to the scope of our analyses, this study falls in the group of "observational" ones, since both the investigators and the patients were aware that a treatment with the drug was actually in place.  Table 1 summarizes the basic study characteristics. Two studies had a randomized, placebo-controlled, double-blinded design [27,28]; the remaining eight studies were characterized by an observational design [43][44][45][46][47][48][49][50]. Among these, one was actually designed as a randomized controlled trial (RCT) [45], but the randomization and the blindness of treatment assignment was performed between different teduglutide dosing schedules rather than between the teduglutide and placebo; therefore, to the scope of our analyses, this study falls in the group of "observational" ones, since both the investigators and the patients were aware that a treatment with the drug was actually in place. The teduglutide dose was 0.05 mg/kg/day in most studies, with the only exception of two studies, in which the 0.1 mg/kg/day dose was also used [27,45]. As for the considered time points at which the efficacy of the teduglutide treatment was assessed, both the available RCTs evaluated their end-points after 6 months of treatment [27,28]; an additional four observational studies provided efficacy data at this time-point [43,46,48,50]. Five studies, all observational, evaluated teduglutide efficacy after 1 year of treatment [44][45][46][47][48]; three studies provided data about teduglutide efficacy after ≥2 years of treatment [46,47,49] ( Table 2).

Relationship between Teduglutide Efficacy and Intestinal Anatomy
Three studies [28,43,44] reported sufficient data to compare the response rate amo patients with or without colon in continuity; two of them evaluated this outcome a months [28,43], and one at 1 year [44]; the latter [44], however, had to be excluded fr the analysis due to partial patient overlap with one of the previous two [43]. The respo rate was significantly lower (-17%, 95%CI: (-31%, -3%)) among patients with colon continuity ( Figure 6A). Six studies [43,[45][46][47]49,50] reported sufficient data to compare weaning rate among those with or without colon in continuity; one of them evaluated t outcome at 6 months [43], one at 1 year [45], one at ≥2 years [49], and three at the available follow-up [46,47,50]. The weaning rate was significantly higher (+16%, 95% (+6%, +25%)) among those with colon in continuity ( Figure 6B). No further analyses co be conducted to differentiate the outcomes according to a finer classification of intesti anatomy due to the paucity of available data; in particular, no sufficient data were av able to compare these outcomes between patients with different types of stomas.

Relationship between Teduglutide Efficacy and Intestinal Anatomy
Three studies [28,43,44] reported sufficient data to compare the response rate among patients with or without colon in continuity; two of them evaluated this outcome at 6 months [28,43], and one at 1 year [44]; the latter [44], however, had to be excluded from the analysis due to partial patient overlap with one of the previous two [43]. The response rate was significantly lower (-17%, 95%CI: (-31%, -3%)) among patients with colon in continuity ( Figure 6A). Six studies [43,[45][46][47]49,50] reported sufficient data to compare the weaning rate among those with or without colon in continuity; one of them evaluated this outcome at 6 months [43], one at 1 year [45], one at ≥2 years [49], and three at the last available follow-up [46,47,50]. The weaning rate was significantly higher (+16%, 95%CI: (+6%, +25%)) among those with colon in continuity ( Figure 6B). No further analyses could be conducted to differentiate the outcomes according to a finer classification of intestinal anatomy due to the paucity of available data; in particular, no sufficient data were available to compare these outcomes between patients with different types of stomas. Nutrients 2022, 14, x FOR PEER REVIEW 10 of 18

Relationship between Teduglutide Efficacy and SBS Etiology
Three studies [28,43,44] reported sufficient data to compare the response rate among patients with SBS due to Crohn's disease or with SBS due to a different etiology; two of them evaluated this outcome at 6 months [28,43], and one at 1 year [44]; the latter [44], however, had to be excluded from the analysis due to partial patient overlap with one of the previous two [43]. The response rate did not significantly differ between the two groups, although a nonsignificant trend toward a slightly better outcome among patients with SBS due to Crohn's disease could be observed (+9%, 95%CI: (-8%, +27%)) ( Figure  7A). Five studies [43,[45][46][47]49] reported sufficient data to compare the weaning rate among patients with SBS due to Crohn's disease or with SBS due to a different etiology; one of them evaluated this outcome at 6 months [43], one at 1 year [45], one at ≥2 years [49], and two at the last available follow-up [46,47]. The weaning rate did not significantly differ between the two groups, although a nonsignificant trend toward a slightly better outcome among patients with SBS due to Crohn's disease could be observed (+7%, 95%CI: (-14%, +28%)) ( Figure 7B). No further analyses could be conducted to differentiate the outcomes according to a finer classification of SBS etiology due to the paucity of available data.

Relationship between Teduglutide Efficacy and SBS Etiology
Three studies [28,43,44] reported sufficient data to compare the response rate among patients with SBS due to Crohn's disease or with SBS due to a different etiology; two of them evaluated this outcome at 6 months [28,43], and one at 1 year [44]; the latter [44], however, had to be excluded from the analysis due to partial patient overlap with one of the previous two [43]. The response rate did not significantly differ between the two groups, although a nonsignificant trend toward a slightly better outcome among patients with SBS due to Crohn's disease could be observed (+9%, 95%CI: (−8%, +27%)) ( Figure 7A). Five studies [43,[45][46][47]49] reported sufficient data to compare the weaning rate among patients with SBS due to Crohn's disease or with SBS due to a different etiology; one of them evaluated this outcome at 6 months [43], one at 1 year [45], one at ≥2 years [49], and two at the last available follow-up [46,47]. The weaning rate did not significantly differ between the two groups, although a nonsignificant trend toward a slightly better outcome among patients with SBS due to Crohn's disease could be observed (+7%, 95%CI: (−14%, +28%)) ( Figure 7B). No further analyses could be conducted to differentiate the outcomes according to a finer classification of SBS etiology due to the paucity of available data.

Quality Assessment and Publication Bias
The results of the quality assessment of the studies are reported in Tables S1 and S2. Altogether, the risk of bias was moderate-to-low in all studies, with the most relevant concerns being related to possible confounding and selection bias in some of the included observational studies. No significant publication bias was found with Begg's test with respect to the primary outcome measure at any of the considered time points (p = 0.70 for the studies at 6 months, p = 0.80 for the studies at 1 year, p = 0.12 for the studies at ≥2 years).

Overall Efficacy
The results of this systematic review and meta-analysis confirm the efficacy of teduglutide for parenteral support reduction in adult patients with SBS. As a primary outcome measure, we evaluated the proportion of patients achieving a ≥20% reduction in PS volume during teduglutide treatment ("response rate"); this choice was consistent with the available literature, in which this cut-off had been unanimously considered as the one of choice for the distinction between responders and non-responders. As a secondary outcome measure, we evaluated the proportion of patients achieving PS independence ("weaning rate").
The results of our analyses show that the response rate to teduglutide treatment could be estimated as 64% at 6 months, 77% at 1 year, and 82% at ≥2 years; on the other hand, the weaning rate could be estimated as 11% at 6 months, 17% at 1 year, and 21% at ≥2 years. When specifically addressing the question of the course of the response to teduglutide over time, the response rate showed a significant increase between 6 months

Quality Assessment and Publication Bias
The results of the quality assessment of the studies are reported in Tables S1 and S2. Altogether, the risk of bias was moderate-to-low in all studies, with the most relevant concerns being related to possible confounding and selection bias in some of the included observational studies. No significant publication bias was found with Begg's test with respect to the primary outcome measure at any of the considered time points (p = 0.70 for the studies at 6 months, p = 0.80 for the studies at 1 year, p = 0.12 for the studies at ≥2 years).

Overall Efficacy
The results of this systematic review and meta-analysis confirm the efficacy of teduglutide for parenteral support reduction in adult patients with SBS. As a primary outcome measure, we evaluated the proportion of patients achieving a ≥20% reduction in PS volume during teduglutide treatment ("response rate"); this choice was consistent with the available literature, in which this cut-off had been unanimously considered as the one of choice for the distinction between responders and non-responders. As a secondary outcome measure, we evaluated the proportion of patients achieving PS independence ("weaning rate").
The results of our analyses show that the response rate to teduglutide treatment could be estimated as 64% at 6 months, 77% at 1 year, and 82% at ≥2 years; on the other hand, the weaning rate could be estimated as 11% at 6 months, 17% at 1 year, and 21% at ≥2 years. When specifically addressing the question of the course of the response to teduglutide over time, the response rate showed a significant increase between 6 months and 1 year and was at least maintained afterwards. Data about the weaning rate were less clear, even if a trend towards an increase in efficacy over time could still be noted. This result is in line with the notion that, as already supposed by other authors [35,45,49], the benefit of teduglutide treatment for PS volume reduction might not only be sustained, but likely increases over time after treatment initiation.

Efficacy by Intestinal Anatomy
Our analysis shows that the efficacy of teduglutide was significantly influenced by intestinal anatomy; the presence of colon in continuity reduced the response rate by −14%, but, conversely, was associated with an increase in weaning rate of +16%.
With respect to the response rate, our findings reinforce those already obtained by previous studies [33][34][35], in which lower PS volume reductions could be observed in patients with colon in continuity compared to those with jejunostomy/ileostomy. Phenotypically, patients with jejunostomy/ileostomy are frequently characterized by accelerated gastric emptying, poor adaptation following resection, and higher PS volume needs [33,51,52]. In these patients, a reduced endogenous GLP-2 secretion was reported, as a consequence of the resection of the terminal ileum and the colon, where the L-cells are predominantly located [33,53,54]; on the contrary, in patients with colon in continuity, the endogenous GLP-2 secretion is essentially preserved [33,53,54]. Therefore, patients experiencing the most pronounced benefits from the administration of an exogenous GLP-2 analog would most probably be those without colon in continuity.
Regarding the weaning rate, irrespectively to teduglutide treatment, the presence of colon in continuity is known to be associated with lower PS volume needs and a higher probability of progressive spontaneous adaptation with possible weaning from PS after an intestinal resection [2,[55][56][57]. In general, patients with colon in continuity may have a greater innate potential for adaptation because of the colon's role in fluid absorption and energy conservation [2,33]. Thus, a similar result could also be expected among teduglutidetreated patients. However, even though some authors have already pointed out a possible tendency towards a higher weaning rate among teduglutide-treated patients with colon in continuity compared to those with other intestinal anatomies [37,43,49], none of the published studies were able to reach a statistical significance on this outcome in univariate analysis, probably due to the low number of included patients. Therefore, our study was the first to actually prove a statistically significant difference in weaning rates between patients with and without colon in continuity.
In summary, the available evidence suggests that the presence/absence of colon in continuity predicts in a different and opposite way the achievement of response and/or weaning after the initiation of teduglutide treatment. This may be explained by the fact that these two processes likely recognize different pathophysiological mechanisms and, thus, different predictors. The achievement of a ≥20% reduction in PS (commonly adopted as the definition of "response") is more likely to be associated with higher baseline PS volume needs and the absence of a residual endogenous GLP-2 secretion; on the contrary, the possibility of completely weaning off PS is more likely associated with lower baseline PS volume needs and the preservation of a certain amount of resorptive capacity by residual colon.

Efficacy by SBS Etiology
Our analysis did not show any significant difference in teduglutide efficacy according to SBS etiology. Nevertheless, a nonsignificant trend toward a slightly better outcome among patients with SBS due to Crohn's disease could be observed, both in terms of response rate and weaning rate.
With respect to the response rate, a possible explanation for this finding might be related to the fact that PS-dependent patients due to Crohn's disease are more likely to Nutrients 2022, 14, 796 13 of 18 have no colon in continuity and higher PS volume needs [33,37,43], with the implications already described within the previous paragraph.
Regarding the weaning rate, the slight tendency towards an increased weaning probability among patients with Crohn's disease held true even though the majority these patients did not have colon in continuity [33,37,43], which we showed to be a condition associated per se with lower weaning rates. Therefore, it might be argued that an SBS etiology due to Crohn's disease might actually be a positive predictive factor for weaning, but that this could not be highlighted in our analysis due to the unavailability of individual patient data, and to the consequent impossibility of controlling the role of intestinal anatomy as a confounder and/or effect modifier. This reasoning, though speculative, would overall be in line with some preliminary findings by Joly et al. [43], which found that Crohn's disease was not a predictor of weaning at univariate analysis (p = 0.91) but almost reached a statistical significance at multivariate regression after adjustment for age, intestinal length, and intestinal anatomy (p = 0.06).

Strengths and Limitations
This was the first meta-analysis that quantitatively assessed the efficacy of teduglutide in patients with SBS, providing pooled estimates of the response rate and the weaning rate among teduglutide-treated patients over time. It confirms and reinforces the findings of previous studies about the role of intestinal anatomy and SBS etiology as predictors of response. Moreover, it better clarifies their role as predictors of weaning, as no clear evidence was available in this regard.
Our analysis had some limitations. First, the robustness of the conclusions was limited by the low number of available studies and patients; this was particularly true for the subanalyses that examined the role of intestinal anatomy and SBS etiology as predictors of the outcome, whose results were weakened by the paucity of available data. Second, the quality was limited by that of the included studies, which were mostly characterized by an observational design; however, the risk of bias was generally moderate-to-low, which confirmed the likely small impact of this issue on the final results. Third, patients' inclusion criteria and PS adjustment protocols were partly different between studies, and this could be responsible for a certain degree of heterogeneity in the considered outcomes; however, heterogeneity is a common limitation of all meta-analyses, and appropriate statistical methods-such as the use of a random-effect model-were adopted to account for it. Fourth, the comparison of response and weaning rates according to intestinal anatomy and SBS etiology were based on crude rate differences, as derived by univariate analyses; thus, the possible interplay between these predictors could not be evaluated.

Conclusions
In conclusion, our meta-analysis confirms the efficacy of teduglutide for PS reduction and/or discontinuation in adult patients with SBS, with a benefit increasing over time up to at least 1 year from treatment initiation. The presence of colon in continuity was a negative predictive factor for response but a positive predictive factor for weaning. An SBS etiology due to Crohn's disease might be a positive predictive factor for both response and weaning, but the available data were insufficient to draw definite conclusions.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nu14040796/s1. Table S1: risk of bias assessment of the RCTs according to RoB 2 tool; Table S2: risk of bias assessment of the observational studies according to ROBINS-I tool; Figure S1: Forest plot for the evaluation of the difference in response rate between the teduglutide and the placebo group in the available RCTs; Figure S2: subgroup analysis for the comparison of the response rate at 6 months in observational studies and RCTs; Figure S3: forest plot for the evaluation of the difference in response rate between the teduglutide and the placebo group in the available RCTs, after the exclusion of the patients treated with a teduglutide dosing schedule of 0.1 mg/kg/day; Figure S4: forest plot for the estimation of response rate at 6 months, 1 year, and ≥2 years, after the exclusion of the patients treated with a teduglutide dosing schedule of 0.1 mg/kg/day; Figure S5: subgroup analysis for the comparison of the response rate at 6 months in observational studies and RCTs, after the exclusion of the patients treated with a teduglutide dosing schedule of 0.1 mg/kg/day; Figure S6: Forest plot for the estimation of the weaning rate at 6 months, 1 year, and ≥2 years, after the exclusion of the patients treated with a teduglutide dosing schedule of 0.1 mg/kg/day.