The search was conducted up to 11 September 2025. We searched the PubMed, Embase, Web of Science, and Cochrane databases for RCTs from the inception of their databases. The search terms included opioid-induced constipation, opioid-receptor antagonists, naldemedine, methylnaltrexone, naloxone, naloxegol, bevenopran, alvimopan, and randomized controlled trial. The search strings are detailed in Supplementary Table S2. Additionally, we manually searched the reference lists of the included literature to identify potentially relevant studies.

The selection criteria for this study were based on the PICOS framework. 1) Population: adults (age ≥18 years) who received opioid or opiate drug treatment and were diagnosed with OIC or opioid-induced bowel dysfunction with constipation; 2) Intervention: the treatment group received opioid-receptor antagonists therapy; 3) Comparison: the control group received placebo treatment; 4) Outcome: the primary outcome measures were the change from baseline of spontaneous bowel movements (SBM), the proportion of responders, the incidence of serious adverse events (SAE), and other adverse events (OAE) during the treatment period; secondary outcome measures included changes in Patient Assessment of Constipation Quality of Life (PAC-QOL), changes in the patient assessment of constipation symptoms (PAC-SYM), and changes in the PAC-QOL satisfaction domain score; studies reported at least one related outcome; 5) Study design: the study design was RCT, and both patients and researchers were blinded.

Studies that met the following criteria were excluded: 1) Conference reports, abstracts, animal studies, and duplicate studies; 2) Non-randomized controlled trials; 3) Missing or unobtainable objective data, and inability to obtain the full text after contacting the corresponding author; 4) Studies involving healthy volunteers as subjects.

EndNote software was used to manage and review the literature. Two researchers (Y L and Y T) independently screened titles/abstracts, reviewed full texts, and selected studies for inclusion. Discrepancies were resolved by a third researcher (ZZ).

We considered SBM (defined as a bowel movement without a rescue laxative taken within the past 24 h) (U.S. National Library of Medicine, 2009a) to be the same concept as rescue-free bowel movements (defined as a bowel movement where no laxatives were used during the prior 24 h) (U.S. National Library of Medicine, 2011b). For the PAC-QOL, a lower score was interpreted as a better quality of life. For the PAC-SYM, a decrease in score corresponded to an improvement in constipation symptoms. Responders were defined as those who had at least 3 weeks of SBM at least 3 times per week with at least one additional 1 SBM per week compared to baseline (Webster et al., 2017). If there were multiple time point measurements, we chose the one with the most extended duration to eliminate arbitrariness or duplicate calculations (Nishie et al., 2019).

All available data from the included studies were extracted into Microsoft Excel. The main extracted content was as follows: general information, including study title, first author’s name, year of publication, and gender ratio of the study population; intervention details, including drug name, dosage, frequency of administration, and treatment duration; outcome-related indicators and data. For literature that reported data only in graphical form, we extracted data from the images using GetData Graph Digitizer. When the standard deviation could not be obtained, we performed data conversion according to the values of the baseline and endpoint. If a study included multiple treatment groups compared to a single placebo group, we combined the data of the treatment groups.

Two researchers (YL and YT) independently assessed the risk of bias for each study using the Cochrane tool, with discrepancies resolved by a third researcher (ZZ). The tool’s evaluation criteria include random sequence generation, allocation concealment, blinding of participants and staff, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases. Ultimately, the risk of bias for the studies was categorized as low, high, or unclear.

The GRADE approach was used to assess the certainty of evidence, which was classified as “very low,” “low,” “moderate,” or “high” (Balshem et al., 2011). RCTs were rated as high-certainty evidence. However, the quality of the studies may be downgraded due to limitations, inconsistency, indirectness, imprecision, and publication bias (Guyatt et al., 2011).

Subgroup analysis was conducted based on the type of medication. The statistical significance threshold was set at p < 0.05, with data integration software being RevMan 5.4 and Stata 15.0. Continuous variables were analyzed using the weighted mean difference (WMD) and 95% confidence interval (CI) as effect sizes. For binary variables, the combined risk ratio (RR) and 95% CI were used.

The chi-square test and I² statistic were used to evaluate the statistical heterogeneity of the included studies. In case of P > 0.05 and I² < 50%, we used a fixed-effect model. Instead, a random-effects model was used (P < 0.05 and I² > 50%). A sensitivity analysis was conducted to determine the stability of the study results. In this study, we systematically assessed the robustness of the pooled effect estimate by switching statistical models and by excluding studies one by one.

When ≥10 studies were included, we used funnel plots, Egger’s test, and Begg’s test to assess potential publication bias (indicating publication bias when P < 0.05) (Begg and Mazumdar, 1994). Conversely, we used funnel plots (for continuous variables) or labbe plots (for binary variables) for assessment. If significant publication bias exists, we used the Duval and Tweedie trim-and-fill method to supplement missing studies (Irwig et al., 1998).

A total of 1,305 studies were retrieved from four databases. Additionally, we manually retrieved three studies. After excluding duplicate studies, reviewing titles and abstracts, and reading full texts, 20 studies (22 RCTs) were included in the meta-analysis (7,761patients). Figure 1 shows the screening process and the reasons for excluding studies.

Among 20 studies, four studies used naloxone (Akhgarandouz et al., 2024; Sanders et al., 2015; Meissner et al., 2009; U.S. National Library of Medicine, 2009a), five used naldemedine (Webster et al., 2017; Hale et al., 2017; Camilleri et al., 2021; Katakami et al., 2017a; Katakami et al., 2017b), one used alvimopan (Irving et al., 2011), four used methylnaltrexone (U.S. National Library of Medicine, 2011b; Shah et al., 2023; U.S. National Library of Medicine, 2009b; Michna et al., 2011), three used naloxegol (Webster et al., 2013; U.S. National Library of Medicine, 2011c; Chey et al., 2023), and three used bevenopran (U.S. National Library of Medicine, 2011a; U.S. National Library of Medicine, 2010; U.S. National Library of Medicine, 2012a). Two articles each reported on two trials. COMPOSE1 and COMPOSE2 were reported in one publication (Camilleri et al., 2021), COMPOSE3 and COMPOSE4 were reported in another (U.S. National Library of Medicine, 2010). The remaining 18 articles reported data from only one trial each. All studies were published in English. The detailed characteristics of each RCT are shown in Supplementary Table S3.

The risk of bias for all included studies is shown in Figures 2, 3. All 22 included trials reported random allocation. However, 11 trials did not specify the details of random sequence generation, and 12 did not specify the details of allocation concealment, both of which were rated as unclear. Both patients and researchers in these studies were blinded to the treatment. Three trials did not clearly state whether assessors were blinded and were rated as unclear. There were no incomplete outcome data or selective reporting in any of the included studies. Among other biases, 20 RCTs were rated as low risk, and two RCTs with a small number of patients (40) were rated as high risk.

Most outcome indicators (SBM, responder rates, PAC-QOL, OAE) met high GRADE standards. Satisfaction levels (as most studies did not provide detailed randomization methods) and SAEs (with high heterogeneity) were assigned a moderate level of evidence. Only PAC-SYM received a low GRADE rating (unclear randomization procedures and limited robustness of the pooled effect). The GRADE rating results are presented in Supplementary Table S4.

In the analysis of SBM changes, we found that naloxegol showed greater improvement at higher doses in the 5–50 mg range (Webster et al., 2013). Similarly, this effect was seen with methylnaltrexone (150–450 mg) (U.S. National Library of Medicine, 2011b), naloxone (2.5–20 mg) (U.S. National Library of Medicine, 2009a; Sanders et al., 2015), and bevenopran (0.1–0.25 mg) (U.S. National Library of Medicine, 2010). However, this effect was not found in the alvimopan group (Irving et al., 2011). Kistemaker found no significant difference in short-term and medium-term responses between low doses (0.15 mg/kg) and high doses (0.30 mg/kg) of methylnaltrexone, despite the low quality of evidence (Kistemaker et al., 2024). In contrast, Rauck’s study demonstrated that 450 mg of methylnaltrexone was superior to 300 mg in terms of efficacy, and this effect was maintained throughout the treatment period (Rauck et al., 2017). Therefore, we are currently unable to establish a clear dose-efficacy correlation, and further high-quality RCTs are needed for validation.

In terms of the proportion of responders, there appears to be a difference in efficacy among the different drugs. The meta-analysis suggests that alvimopan may be inferior to other drugs in improving outcomes. Only the 0.5 mg (BID) dose showed a non-statistically significant trend towards improvement. However, more evidence is required in the future. This finding contrasts with the substantial efficacy reported in another vital study (Jansen et al., 2011). Irving explained that the unusually high placebo response rate and the differences in baseline severity of constipation among patients are the main reasons (Irving et al., 2011). Two factors may have collectively led to the “dilution” of the treatment effect of alvimopan. It makes alvimopan’s efficacy difficult to manifest (Irving et al., 2011).

During the follow-up period without treatment, the improvement in SBM decreased. In PAC-QOL, improvement relative to baseline persisted but also declined. Similarly, another study indicated that the responder rate significantly increased in OIC patients after receiving drug treatment, but then returned to near baseline levels after discontinuation (Webster et al., 2008). These suggest that patients require ongoing treatment to achieve sustained improvement (Sanders et al., 2015). Multiple studies have reported that the efficacy of ORA in treating OIC is long-lasting and well-tolerated, with patients not developing tolerance, thereby maintaining efficacy (Camilleri et al., 2021; Webster et al., 2008). It means that ORAs may be a viable option for long-term pharmacotherapy in OIC patients.

The most adverse events occurring during treatment were gastrointestinal reactions, which occurred at a higher rate than in the placebo group. Nevertheless, the severity was mostly mild to moderate. Gastrointestinal adverse reactions mainly manifested as diarrhea and abdominal pain (Meissner et al., 2009; Sanders et al., 2015; Hale et al., 2017; Katakami et al., 2017a; Michna et al., 2011; Chey et al., 2023; Attal et al., 1990; Paulson et al., 2005). The incidence of diarrhea increases with increasing drug dosage (Webster et al., 2017; Meissner et al., 2009; Katakami et al., 2017a). However, for most patients, diarrhea is transient and is often regarded as a “transient hypermotility”. Some researchers also believe that the gastrointestinal events experienced by patients may primarily stem from the impact on the intestines (Sanders et al., 2015; Katakami et al., 2017a). For example, a pooled analysis of Phase III study data revealed that 10.3% of patients discontinued the drug due to abdominal pain or diarrhea, compared to 5% in the placebo group (Sykes, 1996). In a pooled analysis of naldemedine, the incidence of gastrointestinal adverse events in the drug treatment group was higher than that in the placebo group (Hale et al., 2017). These findings suggest that the high incidence of abdominal pain and diarrhea may be due to the drug’s peripheral antagonistic effects on the gastrointestinal tract (Sanders et al., 2015; Katakami et al., 2017a).

Gastrointestinal perforation associated with PAMORAs is reported to be a rare occurrence (Mehta et al., 2025). ORA does not inherently increase the risk of perforation. Instead, the risk may be associated with pre-existing gastrointestinal vulnerability or concomitant medications in susceptible patients (Yokota et al., 2024). For instance, chronic opioid use can induce intestinal inflammation and alter permeability (Lacy and Cangemi, 2024). Furthermore, retained feces may lead to ischemic necrosis of the bowel wall (Farmer et al., 2019). In patients with a compromised intestinal wall, the increased motility and intraluminal pressure following PAMORA administration could potentially contribute to severe distension or perforation (Yang and Ni, 2008). Concomitant use of medications such as anti-VEGF therapies (e.g., bevacizumab) or anti-inflammatory agents may also increase the risk of gastrointestinal perforation. The annual reporting rates of fatal or life-threatening gastrointestinal perforation events for various PAMORAs in the FDA Adverse Event Reporting System were relatively similar (Mehta et al., 2025). Therefore, Italian experts recommend that close monitoring of the condition and adverse reactions is essential during ORA treatment (Varrassi et al., 2024). To avoid serious complications, three drugs (including naldemedine, methylnaltrexone, and naloxegol) are contraindicated in patients with known or suspected gastrointestinal obstruction or those at risk of recurrent obstruction (Food and Drug Administration, 2018; Food and Drug Administration, 2020; Food and Drug Administration, 2025).

Some drugs had a dose-dependent effect, with the incidence of AEs increasing with higher doses. For example, the incidence of AEs with 50 mg naloxegol was higher than with placebo, whereas the incidence with 5 or 25 mg was comparable. Naloxone and naldemedine also exhibited a dose-dependent effect.

This study showed that treatment duration may be associated with the incidence of OAE, but not with SAE (P = 0.634). For the incidence of OAE, a clear trend of increasing incidence was observed in the larger subgroup studies at 4 and 12 weeks. The incidence of OAE increased by 22% at 4 weeks (7 RCTs) and by 54% at 12 weeks (5 RCTs). In contrast, treatment for 3 weeks (1 RCT) showed a lower incidence of OAE compared to placebo. However, it should not be interpreted as conclusive evidence of protective effects for short-term treatment. Furthermore, a significant confounding factor in the current evidence was the lack of long-term study data for certain drugs. For example, the trial durations for naloxegol and methylnaltrexone were all within 4 weeks. The observed impact on the incidence of OAE might stem from either an interaction between the drug and treatment duration or from the specific properties of individual drugs, necessitating future investigation. Clinical decision-making must weigh benefits against risks. For medium-to long-term treatment (≥4 weeks), patients should be informed of the potential increase in the incidence of OAE, and drug selection should consider existing safety evidence. For patients planning treatment beyond 12 weeks, prioritizing drugs with established long-term safety data (e.g., naldemedine and naloxegol) is a more prudent approach.

It is noteworthy that some studies using sustained-release naloxone solution showed improvement in symptoms among OIC patients, but this was accompanied by a reduction in analgesia and opioid withdrawal symptoms (Attal et al., 1990; Tsuruoka and Willis, 1998). Patients on long-term opioid therapy are more likely to experience withdrawal symptoms (Tsuruoka and Willis, 1998; Berry and Dunn, 2024). Additionally, the confidence interval for the naloxone subgroup in this study was wide. Therefore, the estimate should be considered highly imprecise, as it is primarily driven by a single early study (2009) (Meissner et al., 2009). Additionally, considering the limited number of studies (only two), this signal is more likely to stem from chance or specific study limitations rather than an inherent risk associated with naloxone. Therefore, the results of the naloxone subgroup do not affect the overall conclusion: the incidence of adverse events with ORA showed no significant difference compared to placebo.

In the 2019 American Gastroenterological Association Institute guidelines for the management of OIC, naldemedine is recommended as first-line therapy for patients with an inadequate response to laxatives (strong recommendation, high-quality evidence) (Crockett et al., 2019b). A 2020 network analysis showed that naloxegol had the highest efficacy in terms of responder rates, while naloxone demonstrated better improvement in SBM (Ouyang et al., 2020). In another study, naloxegol produced more SBMs compared to alvimopan and naldemedine (but did not include naloxone) (Rekatsina et al., 2021). Combining the results of this analysis, naloxegol, naloxone, and naldemedine seem to be more efficacious. However, there are currently no head-to-head clinical studies for ORA, making it impossible to directly compare the efficacy of different medications (Ouyang et al., 2020; Pergolizzi et al., 2020). Ouyang R’s research demonstrated that naloxone is very unlikely to cause adverse events (Ouyang et al., 2020). In addition to considering the effectiveness and safety of the drugs, economic cost is also a concern for patients. The clinical application of naldemedine, naloxegol, and methylnaltrexone may be limited by their high economic costs (Crockett et al., 2019b). However, for patients with advanced disease, subcutaneous injection of methylnaltrexone may provide more benefits (Rekatsina et al., 2021; Pergolizzi et al., 2020). Naldemedine was associated with a higher incidence of adverse events compared to placebo in the cancer patient subgroup (Song et al., 2019). Although naldemedine is not recommended in severe hepatic impairment, naldemedine has no restrictions in patients with renal impairment (Rekatsina et al., 2021). Moreover, its safety has been demonstrated in long-term treatment studies lasting up to 52 weeks in this study. In summary, naldemedine is more suitable for non-cancer patients who require long-term oral medication (Rekatsina et al., 2021).

The study has several limitations. Firstly, the funnel plot shows asymmetry between the change in SBM and the incidence of SAE, suggesting publication bias. The existence of publication bias implies a potential overestimation of the treatment effect. The adjusted results continued to support the conclusion that ORA significantly increased SBM, consistent with findings from another study (Nishie et al., 2019). Meanwhile, the incidence of SAE showed no significant difference compared to placebo, raising the possibility of an underrepresentation of smaller studies reporting better safety profiles. During the literature selection process, we discovered that some trials had been completed but had not yet uploaded their data (U.S. National Library of Me dicine, 2003; U.S. National Library of Medicine, 2012b; U.S. National Library of Medicine, 2012c; U.S. National Library of Medicine, 2017; U.S. National Library of Medicine, 2019; U.S. National Library of Medicine, 2024). Secondly, some included studies were rated as unclear regarding random allocation methods and allocation concealment. It suggests that inadequately reported randomization procedures may introduce selection bias, which can bias effect estimates upward. Thirdly, the small number of included studies on satisfaction may indicate limited strength of evidence.