Appendix 1. Glossary of terms

Acute: sudden.

Analogues: a substance that is similar to another substance.

Antioxidants: substances that inhibit oxidation.

Autodigestion: Breakdown of the same organ that secretes the substance.

Bacterial colonisation: growth and multiplication of bacteria.

Cholangiopancreatography: fully known as endoscopic retrograde cholangiopancreatography (ERCP); a procedure carried out on the pancreatic and bile ducts using an endoscope and x‐rays.

Colonisation: presence of bacteria without causing illness (in this context).

Endoscopic sphincterotomy: endoscopic operation to cut the muscle surrounding the common bile duct and the pancreatic duct.

Endoscopic: with the help of an endoscope, a tube inserted into body (in this context, through the mouth and into the stomach and upper part of the small intestine).

Enzyme: substances that enable and speed up chemical reactions that are necessary for the normal functioning of the body.

Epigastric: upper central abdomen.

Epigastric pain: upper central abdominal pain.

Heterogeneity: variability.

Insulin: substance which helps regulate blood sugar.

Interstitial: space in between.

Morbidity: illness (in this context, it means complications).

Mortality: death.

Necrosectomy: removal of dead tissue.

Necrosis: death and decomposition of living tissue usually caused by lack of blood supply but can be caused by other pathological insult.

Necrotising : causing necrosis.

Oedematous: excessive accumulation of serous fluid in the intercellular spaces of tissues.

Pancreatic pseudocysts: fluid collections in the pancreas or the tissues surrounding the pancreas, surrounded by a well defined wall and contain only fluid with little or no solid material.

Pancreatitis: inflammation of the pancreas.

Pathologic insult: substance or mechanism that causes the condition.

Percutaneous: through the skin.

Peripancreatic tissues: tissues surrounding the pancreas.

Pharmacological: medicinal drugs.

Platelet activating factor: substance that causes platelets (cells responsible for clotting of blood) to clump together and is an intermediary substance in the inflammatory pathway.

Probiotics: microorganisms that are believed to provide health benefits when consumed.

Prognostic: to predict the likely outcome.

Protease inhibitors: substances that inhibit proteases.

Protease: an enzyme that digests protein.

Pseudocyst: a fluid‐filled cavity that resembles a cyst but lacks a wall or lining.

Radiology guided percutaneous treatments: treatments carried out by insertion of needle from the external surface of the body which are guided by a scan (usually an ultrasound or CT (computed tomography) scan).

Randomisation: using chance methods to assign people to treatments.

Retrograde: moving backwards.

Sepsis: life‐threatening illness due to blood infection with bacteria, fungus, or virus.

Serum: clear fluid that separates out when blood clots.

Sphincterotomy: a surgical procedure of the internal anal sphincter muscle.

Transabdominal: through the abdomen.

Transient: temporary.

Tumour necrosis factor‐alpha antibody: antibody to tumour necrosis factor‐alpha, an intermediary substance in the inflammatory pathway.

Appendix 2. CENTRAL search strategy

#1 MeSH descriptor: [Pancreatitis, Acute Necrotizing] this term only

#2 MeSH descriptor: [Pancreatitis] this term only and with qualifier(s): [Etiology ‐ ET]

#3 MeSH descriptor: [Pancreas] this term only and with qualifier(s): [Abnormalities ‐ AB, Pathology ‐ PA, Physiopathology ‐ PP]

#4 (acute near/3 pancrea*)

#5 (necro* near/3 pancrea*)

#6 (inflam* near/3 pancrea*)

#7 ((interstitial or edema* or oedema*) near/2 pancrea*)

#8 #1 or #2 or #3 or #4 or #5 or #6 or #7

Appendix 3. MEDLINE search strategy

1. Pancreatitis, Acute Necrotizing/

2. Pancreatitis/et

3. Pancreas/ab, pa, pp

4. (acute adj3 pancrea*).mp.

5. (necro* adj3 pancrea*).mp.

6. (inflam* adj3 pancrea$).mp.

7. ((interstitial or edema* or oedema*) adj2 pancrea*).mp.

8. 1 or 2 or 3 or 4 or 5 or 6 or 7

9. randomized controlled trial.pt.

10. controlled clinical trial.pt.

11. randomized.ab.

12. placebo.ab.

13. drug therapy.fs.

14. randomly.ab.

15. trial.ab.

16. groups.ab.

17. 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16

18. exp animals/ not humans.sh.

19. 17 not 18

20. 8 and 19

Appendix 4. Embase search strategy

1. acute hemorrhagic pancreatitis/

2. Pancreatitis/et

3. acute pancreatitis/

4. (acute adj3 pancrea*).mp.

5. (necro* adj3 pancrea*).mp.

6. (inflam* adj3 pancrea*).mp.

7. ((interstitial or edema* or oedema*) adj2 pancrea*).mp.

8. 1 or 2 or 3 or 4 or 5 or 6 or 7

9. Clinical trial/

10. Randomized controlled trial/

11. Randomization/

12. Single‐Blind Method/

13. Double‐Blind Method/

14. Cross‐Over Studies/

15. Random Allocation/

16. Placebo/

17. Randomi?ed controlled trial*.tw.

18. Rct.tw.

19. Random allocation.tw.

20. Randomly allocated.tw.

21. Allocated randomly.tw.

22. (allocated adj2 random).tw.

23. Single blind*.tw.

24. Double blind*.tw.

25. ((treble or triple) adj blind*).tw.

26. Placebo*.tw.

27. Prospective study/

28. or/9‐27

29. Case study/

30. Case report.tw.

31. Abstract report/ or letter/

32. or/29‐31

33. 28 not 32

34. 8 and 33

Appendix 5. Science Citation Index search strategy

# 1 TS=((acute or necro* or inflam* or interstitial or edema* or oedema*) near/3 pancrea*)

# 2 TS=(random* OR rct* OR crossover OR masked OR blind* OR placebo* OR meta‐analysis OR systematic review* OR meta‐analys*)

# 3 #2 AND #1

Appendix 6. ClinicalTrials.gov search strategy

"Interventional" [STUDY‐TYPES] AND acute pancreatitis [DISEASE] AND ( "Phase 2" OR "Phase 3" OR "Phase 4" ) [PHASE]

Appendix 7. Planned methods

We planned to conduct network meta‐analyses to compare multiple interventions simultaneously for each of the primary and secondary outcomes when there was direct and indirect evidence for at least one comparison. Network meta‐analysis combines direct evidence within trials and indirect evidence across trials (Mills 2012).

We planned to obtain a network plot (Figure 9) to ensure that the trials were connected by treatments using Stata/IC 11 (StataCorp LP) (see Appendix 9 for the Stata commands used). We planned to apply network meta‐analysis to each connected network. We planned to conduct a Bayesian network meta‐analysis using the Markov chain Monte Carlo method in WinBUGS 1.4. We planned to model the treatment contrast (e.g. log OR for binary outcomes, MD or SMD for continuous outcomes, rate ratio for count outcomes, HR for time‐to‐event outcomes) for any two interventions ('functional parameters') as a function of comparisons between each individual intervention and an arbitrarily selected reference group ('basic parameters') (Lu 2004). We planned to use inactive control (combination of placebo and no‐intervention) as the reference group. We planned to perform the network analysis as per the guidance from the NICE DSU documents (Dias 2013). We planned to perform the network meta‐analysis using arm level data. Further details of the codes we planned to use and the technical details of how we planned to perform the analysis are shown in Appendix 10 and Appendix 11. In short, we planned to use three chains and a burn in of 10,000 simulations to ensure convergence, and to obtain the posterior estimates after a further 20,000 simulations. We planned to run the fixed‐effect and random‐effects models (assuming homogeneous between‐trial variance across comparisons) for each outcome. We planned to choose the fixed‐effect model if it resulted in an equivalent or better fit (assessed by residual deviances, number of effective parameters, and deviance information criterion (DIC)) than the random‐effects model. A lower DIC indicates a better model fit. We planned to use the random‐effects model if it resulted in a better model fit as indicated by a DIC lower than that of the fixed‐effect model by at least three. In addition, we planned to perform a random‐effects inconsistency model suggested by NICE DSU (Dias 2012b). We planned to consider the inconsistency model to be better than the random‐effects consistency model (standard random‐effects network meta‐analysis model) if the model fit of the inconsistency model (as indicated by DIC) was at least three lower than the random‐effects consistency model.

Figure 9

---

---

Network plot showing the treatment comparisons that included short‐term mortality. The circles represent treatments while the lines represent the comparisons between the treatments.

For multi‐arm trials, one can enter the data from all the arms in a trial as: the number of people with events and the number of people exposed to the event, using the binomial likelihood and logit link for binary outcomes; the mean and standard error using the normal likelihood and identity link for continuous outcomes requiring calculation of the mean difference; the mean and standard error of the treatment differences using the normal likelihood and identity link for continuous outcomes requiring calculation of the standardised mean difference; the number of events and the number of people exposed to the event using the Poisson likelihood and log link for count outcomes; the follow‐up time in the study, number of people with the event and the number of people exposed to the event using the binomial likelihood and cloglog link for time‐to‐event outcomes. We planned to report the treatment contrasts (e.g. log ORs for binary outcomes, MDs for continuous outcomes, and so on) of the different treatments in relation to the reference treatment (inactive intervention i.e. combined placebo and no‐intervention), the residual deviances, number of effective parameters, and DIC for the fixed‐effect model and the random‐effects model for each outcome. We also planned to report the parameters used to assess the model fit (i.e. residual deviances, number of effective parameters, and DIC) for the inconsistency model for all the outcomes and the between‐trial variance for the random‐effects model (Dias 2012a; Dias 2012b). If the inconsistency model resulted in a better model fit than consistency models, the transitivity assumption is likely to be untrue and the effect estimates obtained may not be reliable. We planned to highlight such outcomes where the inconsistency model results in a better model fit than consistency models.

We found significant clinical heterogeneity in the type of participants included under the different comparisons. To overcome the heterogeneity in the type of people included in different comparisons (See 'Included studies') we planned to perform a separate network meta‐analysis for interventions for mild pancreatitis separately from moderately severe or severe pancreatitis. This is because mild pancreatitis has no local or systemic complications and combining participants with mild and severe acute pancreatitis in the same network meta‐analysis may violate the transitivity assumption (the assumption that the participants included in the different studies with different treatments can be considered to be a part of a multi‐arm randomised controlled trial ‐ i.e. they should be reasonably similar in characteristics). We then planned to assess inconsistency again. However, this was not appropriate in the subgroup of severe acute pancreatitis because of the absence of any comparison in which direct and indirect comparison was available. If there was no evidence of inconsistency in the revised analysis, we planned to present the results of the analysis for mild and moderate or severe acute pancreatitis separately. If there was persistent evidence of inconsistency, we planned to present the results from the direct comparison in the 'Summary of findings' table.

We planned to calculate the 95% CrIs of treatment effects (e.g. ORs for binary outcomes, MDs for continuous outcomes, and so on) in the Bayesian meta‐analysis, which is similar in use to the 95% confidence intervals in the frequentist meta‐analysis. These are the 2.5^th percentile and 97.5^th percentiles of the simulations. We planned to report the mean effect estimate and the 95% CrI for each pair‐wise comparison in a table. We also planned to estimate the probability that each intervention ranks at one of the possible positions, and have presented this information in graphs. It should be noted that a less than 90% probability that the treatment is the best treatment is unreliable (i.e. one should not conclude that the treatment is the best treatment for that outcome if the probability of it being the best treatment is less than 90%) (Dias 2012a). We also planned to present the cumulative probability of the treatment ranks (i.e. the probability that the treatment is within the top two, the probability that the treatment is within the top three, etc.) in graphs. We also planned to plot the probability that each treatment is best for each of the different outcomes (rankograms) which are generally considered more informative (Dias 2012a; Salanti 2011). We planned to perform direct comparisons using the same codes. This would have allowed us to assess the heterogeneity in the comparisons and provide additional information in the 'Summary of findings' table. We also planned to use the Tau² statistic to measure heterogeneity among the trials in each analysis. The Tau² statistic provides a measure of the variability of the effect estimate across studies in a random‐effects model (Higgins 2011). If we identified substantial heterogeneity, we planned to explore it by meta‐regression. We also planned to assess the differences in the effect estimates between the subgroups using meta‐regression for each source of heterogeneity (i.e. one analysis for each source of heterogeneity) with the help of the code shown in Appendix 12. We planned to perform the following subgroup analyses regardless of heterogeneity. We planned to calculate the interaction term (Dias 2012c). If the 95% CrI of the regression coefficient of the interaction term does not overlap zero, we considered this statistically significant.

In the presence of adequate data where authors report the outcomes of participants at multiple follow‐up time points, we planned to follow the methods suggested by Lu 2007 to perform the meta‐analysis.

We planned to use methods and recommendations described for grading network meta‐analysis (Puhan 2014). This includes grading the quality for direct comparison, indirect comparison, and network meta‐analysis and presenting the information in tabular format.

Appendix 8. WHO ICTRP search strategy

Acute pancreatitis

Appendix 9. Stata code for network plot

networkplot t1 t2, labels(T1 T2 T3 ..)

Appendix 10. Winbugs code

Appendix 11. Technical details of network meta‐analysis

The posterior probabilities (effect estimates or values) of the treatment contrast (i.e. log odds ratio, mean difference, standardised mean difference, rate ratio, or hazard ratio) may vary depending on the initial values to start the simulations. In order to control the random error due to the choice of initial values, we performed the network analysis for three different initial values (priors) as per the guidance from The National Institute for Health and Care Excellence (NICE) Decision Support Unit (DSU) documents (Dias 2013). If the results from three different priors are similar (convergence), then the results are reliable. It is important to discard the results of the initial simulations as they can be significantly affected by the choice of the priors and only include the results of the simulations obtained after the convergence. The discarding of the initial simulations is called 'burn in'. We ran the models for all outcomes for 10,000 simulations for 'burn in' for three different chains (a set of initial values). We ran the models for another 20,000 simulations to obtain the effect estimates. We obtained the effect estimates from the results of all the three chains (different initial values). We also ensured that the results in the three different chains are similar in order to control for random error due to the choice of initial values. This was done in addition to the visual inspection of convergence obtained after simulations in the burn in.

We ran three different models for each outcome. The fixed‐effect model assumes that the treatment effect is the same across studies. The random‐effects consistency model assumes that the treatment effect is distributed normally across the studies but assumes that the transitivity assumption is satisfied (i.e. the population studied, the definition of outcomes, and the methods used were similar across studies and that there is consistency between the direct comparison and indirect comparison). A random‐effects inconsistency model does not make the transitivity assumption. If the inconsistency model resulted in a better model fit than the consistency model, the results of the network meta‐analysis can be unreliable and so should be interpreted with extreme caution. If there is evidence of inconsistency, we planned to identify areas in the network where substantial inconsistency might be present in terms of clinical and methodological diversities between trials and, when appropriate, limit the network meta‐analysis to a more compatible subset of trials.

The choice of the model between fixed‐effect and random‐effects was based on the model fit as per the guidelines of the NICE TSU (Dias 2013). The model fit will be assessed by deviance residuals and Deviance Information Criteria (DIC) according to NICE TSU guidelines (Dias 2013). A difference of three or five in the DIC is not generally considered important (Dias 2012c). We used the simpler model, i.e. fixed‐effect model if the DIC are similar between the fixed‐effect and the random‐effects models. We used the random‐effects model if it results in a better model fit as indicated by a DIC lower than that of the fixed‐effect model by at least three.

We planned to calculate the effect estimates of the treatment and the 95% credible intervals using the following additional code.
# pairwise ORs and MD for all possible pair‐wise comparisons, if nt>2
for (c in 1:(nt‐1)) {
for (k in (c+1):nt) {
OR[c,k] <‐ exp(d[k] ‐ d[c])
#MD[c,k] <‐ (d[k]‐d[c])
}
}

where c indicates control group, k indicates intervention group, OR indicates odds ratio or other ratios, and MD indicates mean difference or other differences.

Appendix 12. Winbugs code for subgroup analysis