Analgesia for the treatment of acute pancreatitis: a protocol for a systematic review and network meta-analysis

We will follow the standard guidelines of Cochrane Collaboration and report the findings of the review following Preferred Reporting Items for Systematic Reviews and Meta-Analyses. The protocol is registered at the International Prospective Register of Systematic Reviews website (ID: CRD42023477878). The study will start on 5 April 2024, and it will run through 30 July 2024.

Study type

We will include individual, and cluster randomised controlled trials. We will exclude observational studies such as cohort studies, case-control studies, case reports, case series, controlled before-and-after studies, interrupted time series and commentaries.

Population

We will consider studies that included adult and paediatric patients with non-severe, acute pancreatitis that require hospital admission. Acute pancreatitis will be defined based on the presence of at least two out of three of the following findings: epigastric abdominal pain, elevated levels of serum pancreatic enzymes and/or imaging evidence of pancreatic inflammation. Non-severe forms of acute pancreatitis will be defined as having the absence or transient (less than 48 hours) presence of organ failure, local pancreatic, or systemic complications.3 6 In sum, our definition embraces mild and moderate forms of acute pancreatitis. We will exclude studies that include patients diagnosed with chronic pancreatitis being managed in outpatient settings; however, we will include studies that address acute pancreatitis in patients with recurrent pancreatitis. We will exclude studies restricted to patients with severe pancreatitis because pain management is more complex in these patients and should be addressed in a separate review. We will, however, consider studies where both severe and non-severe cases of pancreatitis were included, and disaggregated data are available for non-severe pancreatitis patients. If the disaggregated data are not available, we will include the study if more than 50% of the patients had non-severe pancreatitis. We will do a sensitivity analysis to assess if the inclusion of such studies affects the results. We will exclude studies where patients had pancreatic cancer.

Intervention

We will assess all analgesic medications that aim to treat pain in patients with non-severe, acute pancreatitis, including opioid and non-opioid analgesics. We will only consider medications administered in the hospital setting. We will include medications irrespective of route of administration, dose, frequency, and duration. We will exclude studies that addressed non-pharmacological interventions such as acupuncture, massage and so on. Table 1 gives a list of commonly used medications for pain management in acute pancreatitis.

Table 1

Medications used to treat pain in patients with acute pancreatitis

Comparison

The comparison group may include a placebo or other active, pharmacological or non-pharmacological interventions to treat pain in acute pancreatitis.

Outcomes

Primary outcomes

  1. Pain score (continuous outcome).

  2. Need for supplementary analgesia (dichotomous outcome).

Secondary outcomes

  1. Number of patients with improved pain as defined by author (dichotomous outcome).

  2. Number of patients with controlled pain as defined by author (dichotomous outcome).

  3. Need for analgesia for breakthrough pain (number of breakthrough analgesia deliveries) (continuous outcome).

  4. Need for analgesia for breakthrough pain (amount of breakthrough analgesia delivered) (continuous outcome).

  5. Serious adverse events (dichotomous outcome).

  6. Non-serious, adverse events: nausea and vomiting (dichotomous outcome).

  7. Non-serious, adverse events: ileus (dichotomous outcome).

  8. Non-serious, adverse events: any cause (dichotomous outcome).

  9. Time to Enteral diet initiation (hours, continuous outcome).

  10. Local complications: peripancreatic fluid collection, peripancreatic/pancreatic necrosis, pancreatic pseudocyst (dichotomous outcome).

  11. Progression to severe pancreatitis (organ or organ system failure lasting >48 hours) (dichotomous outcome).

  12. Length (days) of hospitalisation (continuous outcome).

  13. 30-day all-cause mortality (dichotomous outcome).

  14. Quality-of-life scores (continuous outcome).

Need for supplementary analgesia will be defined as instances when the trial drug fails to relieve pain following the trial protocol. The need for analgesia for breakthrough pain will be defined as occurrences when the primary trial drug continued, and supplementary analgesia was used for breakthrough pain only. The serious adverse events will be defined based on the definitions of the US Food & Drug Administration. We will consider the adverse events per person and not the total number of adverse events per group.

Time of follow-up

For the primary outcomes, we will consider the time of follow-up at 48–96 hours of admission. For the secondary outcomes, we will use the information from the longest available follow-up.

Literature search

In conjunction with the review team, a medical librarian at our institution, the University of Iowa, will develop a comprehensive search strategy. The systematic literature search will be conducted in several databases, including PubMed, Embase, Cochrane Library, Cochrane Central Register of Controlled Trials, Scopus, Web of Science collections, Cumulative Index to Nursing and Allied Health and Global Index Medicus. The search strategies for PubMed and other databases are available in the online supplemental document. No studies will be excluded based on language, publication year, publication status (excepting retraction) or outcome. Additionally, the National Library of Medicine’s Clinical Trials database (ClinicalTrials.gov) will be used to identify active or unpublished studies. We will use the clinical trial registration number to find all studies published from a particular trial. Additional studies will be found through citation tracking of all included studies. The references of previously published, network studies will be reviewed for additional inclusion. We will also search for data presented at major gastrointestinal conferences (North American Society of Pediatric Gastroenterolgy, Hepatology and Nutrition (NASPGHAN), Euorpean Society of Pediatric Gastroenterolgy, Hepatology and Nutrition (ESPGHAN), Digestive Disease Week, PancreasFest, World Congress on Gastroenterology, American Pancreatic Association (APA)) and use grey literature by way of websites with content on pertinent projects (Investigator of the INternational Study Group of Pediatric Pancreatitis: In search for a cuRE (INSPPIRE) Pediatric Pancreatitis Research Project, Swedish Pancreatitis Cohort, Dutch Pancreatitis Study Group). Finally, we will reach out to the experts in the field to ask for any known studies on analgesia control in acute pancreatitis.

Supplemental material

Data extraction and synthesis

Selection of studies

The studies from all the databases will be compiled in EndNote, and duplicates will be removed. Two authors will apply a three-phase approach to screen studies to select those to be used for end-phase data extraction; screening and end-phase data extraction will be done using the systematic review software, COVIDENCE V.2.0.22 The initial phase will involve screening titles and abstracts to identify eligible studies. Studies identified in the first phase will move to phase two for full-text review, which will lead to data extraction in the third and final phase. If studies are presented only in truncated format, we will reach out to authors to obtain complete study details; based on the amount of detail provided, authors will decide about inclusion based on available data. If studies are available in a language other than English, we will work to obtain an English text or use local translation resources. If multiple reports of the same study exist, they will be consolidated into a single study, with appropriate extraction from all reports.

Data extraction

Two authors will independently extract data from phase three studies and subsequently compare their findings. Any conflicts will be addressed with assistance from a senior author (AU). The following information will be extracted from each study: study authors, publication journal, study type, study site, study year, population characteristics (age, diagnosis, inclusion criteria, exclusion criteria), intervention characteristics (medication, dose, route, frequency, duration), control characteristics (placebo or medication dose, route, frequency, duration), outcomes and risk of bias. A draft of the data extraction sheet is available as a online supplemental document. Preference will be placed on data that require the least processing or conversion by authors and the least inference by reviewers, and we will aim to extract raw values (means, SD) as opposed to effect sizes. To avoid reviewer bias, we will have a predetermined order of preference for extracting outcomes when data are available in multiple formats. Preference will be placed on data that require the least manipulation by authors or inference by reviewers, and we will aim to extract raw values (means, SD) as opposed to effect sizes. For mortality data, we will give preference to denominators in the following order: number with definite outcome known, number randomised and time of follow-up.

Studies with missing data

Any missing data will be documented. If the authors report the imputed missing data, we will use the same. If no calculations are available for missing data, we will contact the study authors to request missing data. For continuous outcomes, if a study does not report a SD for a mean and cannot be calculated from the reported data, we will use an SD reported from a similar study for the same outcome. When possible, we will use the final values of a continuous outcome in our analysis. If final values are not provided, we will use a clinically appropriate difference or rate of change to calculate the final values. If the final values are not available for the difference between the start and end of the study, we will request this information from the authors. We will use the data based on intention-to-treat analysis (ITT) from randomised control trials. If ITT data are not provided, we will create an appropriate ITT analysis based on known outcomes and conduct a sensitivity analysis to assess the impact of the calculation.

Assessment of risk bias in included studies

The risk of bias in randomised control trials will be evaluated using the Cochrane Risk of Bias Tool (ROB V.2.0). This tool focuses on the risk-of-bias assessment in different aspects of study design, study execution and study reporting. Specifically, five domains will be addressed: bias during initial randomisation, bias due to deviation from intervention design, bias due to missing outcomes data, bias due to outcome measurement and bias during result selection and reporting. Possible risk-of-bias judgments are (1) low risk of bias, (2) some concerns for bias, or (3) high risk of bias.23 Risk-of-bias assessment for each study will be carried out independently by two authors, and a senior author will address disagreement.

Data synthesis

The findings from the included studies will be reported qualitatively and quantitively. Characteristics of the included studies and methodology will be summarised, and if the included studies are deemed to be homogenous by members of the research team, meta-analyses will be performed.24

We will first conduct an updated traditional pairwise meta-analysis that compares two interventions at a time, including an analysis of opioids versus non-opioids. We will consider network meta-analysis that compares multiple interventions via direct and indirect comparison if the research team members conclude that assumptions of transitivity and coherence hold true. Direct comparison summary estimates can be mathematically summated to transitively produce indirect comparisons between various analgesics across multiple randomised controlled trials if the interventions have a point of connectivity.25 The transitivity will be assessed based on study characteristics and the presence of an effect modifier. We will consider the following effect modifiers: age, gender and route of administration. The coherence assumption will be assessed by calculating a coherence factor. Data from each study will be extracted, with the effect size calculated for a given comparison. The treatment effect of analgesics for pain control shall be calculated using the mean difference or standardised mean difference for the continuous outcomes at a given follow-up. For the primary outcome of pain scores, we will consider the pain scores at the end of the follow-up. Dichotomous outcomes will be pooled to obtain a relative risk. Due to treatment variation across studies, we will use a random effect model for pairwise meta-analysis. Network meta-regression and subgroup analysis will be conducted to evaluate for any effect modifier that may influence the effect sizes of interventions within the network of network meta-analysis.24

We anticipate that all the studies may not report all the outcomes of interest in the desired format. For example, different studies may use different pain scales for pain scores. Even though the Visual Analogue Scale is the most commonly used scale for the assessment of pain, we will consider other scales that are used by primary studies and pool the data via standardised mean differences provided that the assumption of direction of effect is similar across the scales (eg, lower scores indicated better pain control). If all data for a continuous outcome are provided in the same scale, we will use mean difference for analysis; however, if data for a continuous outcome are provided in different scales, we will use standard mean difference for analysis.

The GRADE approach will be used to report the certainty of evidence for each pairwise comparison of interventions in the network meta-analysis.26–28 The GRADE approach is a comprehensive framework to assess the overall quality of the evidence for an outcome based on study design, risk of bias, heterogeneity of studies, directness of evidence, precision of effect estimates and publication bias. The approach to GRADE analysis of the network meta-analysis will be that we will rate the certainty of evidence for each of the pairwise analyses in a network separately, and both direct and indirect evidence (combined effect) will be considered for the same. The key considerations specific to ratings of evidence from a network meta-analysis include a differential approach to the consideration of imprecision in rating, taking into account the certainty of direct and indirect evidence to consider the implication of incoherence in the network, and pairwise analysis. The evidence quality for each pairwise analysis will be characterised as very low (little confidence in the effect estimate), low (limited confidence in the effect assessment), moderate (moderate confidence in the effect assessment), or high (high confidence in the effect assessment).26–28 We will use GRADEpro GDT software that will assist in the development of a summary of the findings table from the GRADE analysis.29 We will rate the quality of evidence for the following outcomes: pain score, need for supplementary analgesia (dichotomous outcome), length (days) of hospitalisation (continuous outcome), serious adverse events and quality-of-life scores.

We are interested in the totality of the evidence for analgesics used to treat acute non-severe pancreatitis. For the interpretation of data for a given analgesic, we will assess the benefits and adverse events. We will consider a minimal clinically important difference of 20% to establish non-inferiority versus superiority of an analgesic versus a control group in a pairwise analysis.

This post was originally published on https://bmjopen.bmj.com