Efficacy of pancreatic enzyme replacement therapy in chronic pancreatitis: systematic review and meta-analysis

Objective The benefits of pancreatic enzyme replacement therapy (PERT) in chronic pancreatitis (CP) are inadequately defined. We have undertaken a systematic review and meta-analysis of randomised controlled trials of PERT to determine the efficacy of PERT in exocrine pancreatic insufficiency (EPI) from CP. Design Major databases were searched from 1966 to 2015 inclusive. The primary outcome was coefficient of fat absorption (CFA). Effects of PERT versus baseline and versus placebo, and of different doses, formulations and schedules were determined. Results A total of 17 studies (511 patients with CP) were included and assessed qualitatively (Jadad score). Quantitative data were synthesised from 14 studies. PERT improved CFA compared with baseline (83.7±6.0 vs 63.1±15.0, p<0.00001; I2=89%) and placebo (83.2±5.5 vs 67.4±7.0, p=0.0001; I2=86%). PERT improved coefficient of nitrogen absorption, reduced faecal fat excretion, faecal nitrogen excretion, faecal weight and abdominal pain, without significant adverse events. Follow-up studies demonstrated that PERT increased serum nutritional parameters, improved GI symptoms and quality of life without significant adverse events. High-dose or enteric-coated enzymes showed a trend to greater effectiveness than low-dose or non-coated comparisons, respectively. Subgroup, sensitive and meta-regression analyses revealed that sample size, CP diagnostic criteria, study design and enzyme dose contributed to heterogeneity; data on health inequalities were lacking. Conclusions PERT is indicated to correct EPI and malnutrition in CP and may be improved by higher doses, enteric coating, administration during food and acid suppression. Further studies are required to determine optimal regimens, the impact of health inequalities and long-term effects on nutrition.


INTRODUCTION
Chronic pancreatitis (CP) is a progressive fibro-inflammatory disorder with sustained destruction of acinar, ductal and islet cells. 1 CP has an annual incidence of 4-12 per 100 000, 2 increasing worldwide. 3 Aetiologies include environmental toxins ( prolonged, heavy alcohol exposure and cigarette smoking), hyperlipidaemia, single and/or multiple genetic mutations (PRSS1, SPINK1, CTRC, CASR, CFTR, CLDN2 and CPA1) and autoimmune disease. 3 4 CP is generally progressive, with marked variation in abdominal pain and GI symptoms. These greatly impair quality of life (QoL), 5 while exocrine pancreatic insufficiency (EPI) and diabetes mellitus (DM) contribute to the low median

Significance of this study
What is already known on this subject? ▸ Chronic pancreatitis (CP) is a major health problem associated with health inequalities, causing intractable abdominal pain, peripancreatic complications, exocrine pancreatic insufficiency (EPI), long-term malnutrition and type 3c diabetes mellitus. ▸ Pancreatic enzyme replacement therapy (PERT) is administered by some for EPI in CP but the benefits remain to be confirmed. ▸ No meta-analysis has been conducted previously of randomised clinical trials assessing PERT for EPI in CP, while a previous Cochrane review of PERT in CP was inconclusive.
What are the new findings?
▸ This meta-analysis shows that PERT improves the coefficients of fat and nitrogen absorption versus baseline and versus placebo, reducing faecal fat excretion, faecal nitrogen excretion, faecal weight and abdominal pain without significant adverse events. Follow-up studies have found that PERT increases serum nutritional parameters, improves GI symptoms and quality of life without significant adverse events. ▸ Although there was significant heterogeneity between trials, subgroup analyses did not alter the findings, and exclusion of trials with small sample sizes (<40) or without imaging and/or histology to establish CP significantly reduced the heterogeneity. Data on health inequalities were sparse but for those on CP aetiology attributed to alcohol excess.
survival of 15-20 years from diagnosis. 3 6 The prevalence of CP has been estimated at circa 50 per 100 000 but this is a significant underestimate in view of incidence and median survival, which suggests 100-200 per 100 000; 3 similarly, EPI is probably underestimated. 3 Not surprisingly the management of CP and its complications is resource intensive; 7 were the management of EPI and DM from CP to be improved and complications reduced, these resources would be better spent. EPI, characterised by inadequate pancreatic secretion of digestive enzymes and bicarbonate, is one of the most significant complications of CP affecting >50% of diagnosed patients, 8 resulting in compromised digestion, absorption and metabolism of nutrients. Severe EPI tends to develop between 5 and 10 years following an initial diagnosis of CP 1 and can complicate acute necrotising pancreatitis, cystic fibrosis, DM, pancreatic cancer and following surgery to bypass or resect the duodenum and/or pancreas. 9 EPI from CP reduces absorption of fat including essential fatty acids, fat-soluble vitamins A, D, E and K, calcium, magnesium, zinc, thiamine and folic acid. 10 EPI from CP is frequently experienced as diarrhoea, abdominal discomfort and/or pain, weight loss and grossly as steatorrhoea (strictly, >7 g fat in stool/24 hours). 1 Intermediate and long-term malnutrition from EPI increases the incidence of osteopenia/osteoporosis, 11 lowtrauma fractures, 12 cardiovascular diseases 13 and infections. 6 These complications develop on a background of health inequalities, differences between people's health as a result of social, geographical or other factors, which are associated with less favourable outcomes from CP; foremost is malnutrition. 14 Rates of alcoholism and smoking also have significant socioeconomic gradients, as do osteopenia/osteoporosis, DM and cardiovascular diseases, making health inequality an independent risk factor for the initiation and progression of CP. 4 Pancreatic lipase is the principal enzyme in fat digestion accounting for 90% of total lipase activity. 9 Typically steatorrhoea is unlikely until pancreatic lipase output falls to <10% of normal, 15 although subclinical EPI exists in many individuals with CP. 16 Large-scale Northern European studies suggest that the majority of patients with EPI secondary to CP 17 or pancreatic surgery 18 are undertreated, likely because of insufficient use of pancreatic enzyme replacement therapy (PERT), suboptimal scheduling in relation to meals or insufficient control of gastric acid output. The European Society for Parenteral and Enteral Nutrition 2006 Guidelines 19 suggest improvement of steatorrhoea and maintenance of bodyweight as primary markers of treatment success. Such end points, however, do not detect subtler signs of malnutrition; nor do these guidelines suggest formal measures to assess improvement of steatorrhoea. Furthermore, these end points are unsuitable as primary outcome measures for studies of PERT for EPI, particularly short-term studies. For these, accurate measures that detect alterations and/or improvements in absorption are necessary. 1 3 9 16 In practice, licensed enzyme replacement therapies differ greatly: granules, tablets, microspheres, minimicrospheres or capsules with enteric coating make comparisons challenging. While the European Medicines Agency has delegated regulation of PERT to national authorities, the US Food and Drug Administration (FDA) addressed these differences in 2004 (updated in 2006), 20 issuing requirements for new drug applications, specifying amount, stability and efficacy. To date, six enzymes have FDA approval: Creon, Pancreaze, Pertzye, Ultrase, Viokase and Zenpep, all of porcine origin.
Questions remain as to the indications for and efficacy of PERT, including methods of assessment, formulation, dose, administration and use of acid suppression. Shafiq et al 21 conducted a meta-analysis of PERT for CP regardless of the presence or absence of EPI. Data for weight loss and faecal fat were combined from only two studies; the authors inferred that the evidence for any beneficial role of PERT in CP was inconclusive. Waljee et al 22 and Taylor et al 23 combined trials in cystic fibrosis and CP, concluding that PERT improves but does not normalise the coefficient of fat absorption (CFA) in EPI; these two systematic reviews included a total of only three placebo-controlled trials of PERT in CP, and no quantitative data were combined to determine outcomes. Three further placebo-controlled trials of PERT in CP have since been conducted, two of which were followed by open-label extensions. We have therefore conducted a systematic review of 17 randomised controlled trials (RCTs) with meta-analysis of 14 to determine whether PERT is indicated for EPI in CP, the impact of this treatment and factors contributing to optimisation of PERT. Our study is strengthened by meta-analysis of PERT versus baseline, PERT versus placebo and PERT versus PERT to provide a more rigorous evaluation, comparing CFA, coefficient of nitrogen absorption (CNA), faecal fat excretion (FFE), faecal nitrogen excretion (FNE), faecal weight, abdominal pain and GI symptoms, bodyweight, QoL and adverse events.
Studies meeting all the following criteria were included: (1) in English peer-reviewed journals; (2) prospective, randomised design, investigating efficacy and safety of PERT in EPI from CP in adults (age ≥18 years; including patients who had pancreatic resection for CP but not other indications); (3) reporting clinical outcomes of interest; and (4) only the most recent study of multiple overlapping patient populations from the same institution Significance of this study How might it impact on clinical practice in the foreseeable future? ▸ This systematic review and meta-analysis of 17 randomised trials of PERT for EPI in CP demonstrates the efficacy of PERT for correcting malnutrition in CP. PERT may be optimised by higher doses, enteric coating, ingestion during food and acid suppression. ▸ Further studies are needed to determine optimal methods to address the impact of health inequalities on PERT for EPI in CP.
unless a prior study had higher quality. Abstracts, case reports, letters, expert opinions, editorials, reviews and non-RCTs were excluded. Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) criteria 24 were followed. Two authors (DdlI-G and WH) independently scrutinised all identified studies and reached consensus for final inclusion, with adjudication (PS) when there was disagreement.

Data extraction
Two authors (DdlI-G and WH) extracted data independently using predefined standardised forms. These captured study design, quality assessment (below), baseline characteristics, health equality indicators (ethnicity, place of residence, socioeconomic background, employment/insurance status, profession, alcohol use and cigarette smoking), diagnostic criteria for CP and EPI, exclusion criteria, trial process, details of PERT and outcomes of interest.

Outcomes of interest
CFA has been used most commonly as the primary end point to assess the efficacy of PERT on EPI due to CP and cystic fibrosis in RCTs, more informative than the presence or absence of steatorrhoea. 22 23 We have therefore used CFA as the primary outcome measure, calculated from fat intake (∼100 g/day of dietary fat) and excretion (from 72 hours faecal collection) using the following equation: CFAð%Þ ¼ ½ð fat intake in g--fat excretion in gÞ= fat intake in g Â 100 CNA has been used to measure the effect of PERT on protein absorption in EPI, ignored in previous systematic reviews; [21][22][23] CNA was calculated as follows: CNAð%Þ ¼ ½ðnitrogen intake in g--nitrogen excretion in gÞ =nitrogen intake in g Â 100 Secondary outcomes included CNA, FFE, FNE, faecal weight, faecal consistency (formed/normal or soft/watery), faecal frequency (stools per day), flatulence (none/mild/moderate/severe), abdominal pain (none/mild/moderate/severe) and adverse events. When available, serum nutritional markers, diarrhoea, weight loss/gain and QoL were included.

Statistical analysis
Means and SDs of continuous variable were used for meta-analysis, estimated 26 when medians were given. If CFA and CNA were not available, these were calculated as described.
If different protocols of PERT were used in any single study, average daily lipase dose was calculated in United States Pharmacopoeia (USP) units. Meta-analyses compared PERT versus baseline, PERT versus placebo or PERT versus PERT. Forest plots were generated using Review Manager V.5.3 software (Cochrane Collaboration, Oxford, UK). Continuous variables were expressed as weighted mean differences (WMD) and categorical variables as ORs with 95% CIs. A random-effects model 27 was employed to ensure conservative estimates in meta-analyses. Heterogeneity was evaluated using χ 2 with p<0.1 considered significant. Statistical heterogeneity was assessed using I 2 values with cut-offs of 25%, 50% and 75% to indicate low, moderate and high heterogeneity, respectively. 28 Subgroup analyses examined studies of high quality; parallel, multicentre design; samples ≥40 and Western populations. Sensitivity analyses were conducted by diagnostic criteria and inclusion of pancreatic surgery. Meta-regression analyses assessed influence of age, gender, study design, study quality, lipase dose and publication year on summary estimates using Stata SE V.13 Software (StataCorp, Texas, USA); p<0.05 was considered significant. Publication bias was assessed using funnel plots, 29 and p values generated from both CFA and FFE as per Begg and Mazumdar 30 and Egger et al; 31 p<0.10 was considered significant.

Patient and public involvement
The research design, methods, results and their interpretation were reviewed by the NIHR Liverpool Pancreas Biomedical Research Unit Patient Advisory Group, and modifications made accordingly. The group is comprised of patients with a history of acute pancreatitis, CP or pancreatic cancer, or their carers, or interested members of the public.

Design and quality assessment of included studies
The PRISMA flow diagram is shown in figure 1; 17 studies 32-48 were included. Study design and quality assessment are shown in table 1. There were four 32 33 40 45 conducted in the USA, nine 34-39 41 43 44 in Europe, two 46 47 in the USA and Europe, one 42 in South Africa and one 48

Inclusion and exclusion criteria
The diagnostic criteria for CP and EPI are shown in table 2. CP was diagnosed by imaging and/or histology in six studies, 33 36 39 46-48 imaging and/or history in seven 35 37 38 41-44 and history, abnormal secretin-pancreatozymin test and/or histology in one. 34 The diagnostic criteria in the remaining three studies 32 40 45 were unclear. EPI was defined by FFE >7-8 g/day in five, 33 35 36 41 43 by FFE >10-15 g/day and/or CFA ≤80% in nine and by faecal elastase ≤100 mg/g in one study. 47 One study 32 did not report criteria for EPI and 1 40 included CP patients with documented symptomatic steatorrhoea. Ten 36 38 39 41-43 45-48 reported detailed exclusion criteria.

Details of trial process
The trial process and adverse events are shown in table 2. RCTs of PERT typically had a run-in phase to scrutinise and stabilise eligible participants before treatment initiation. During run-in or washout, PERT was stopped. Six studies 34-37 43 44 reported a 3-15-day no-treatment run-in, while five 38 40 45-47 had a 5-14-day period for placebo administration before commencing the trial. Three studies 41 42 48 reported a 14-day run-in with the last seven days before randomisation treated by PERT. The remaining three studies 32 33 39 did not report a run-in phase. During treatment, PERT or placebo was used with or without a washout in between switching treatments. Daily fat intake was recorded in the majority (normally 100 g/day), but not in three. 36 37 44 Only three 37 46 47 reported use of a faecal dye. The equilibration time was normally 2-5 days before a 3-day period of faecal fat collection. Adverse events were reported in six, 36 41 45-48 with an incidence of 11.4-42.7%.
RCT study duration ranged from several days to 2 months; none assessed long-term effects of PERT, although two openlabel extension studies examined the nutritional impact of PERT over 6 months 49  Delhaye's study 39 found no significant differences in the efficacy of Pancrease HL and Creon 3 (both enteric-coated microspheres). Halm's study 41 showed the primary outcome CFA and adverse events to be similar for Creon 10000 microspheres and minimicrospheres. Vecht's study 44 found low-dose (lipase 10 000 USP units tds) improved fat absorption and reduced symptoms compared with normal dose (lipase 20 000 USP units tds) Pancrease when combined with strong acid inhibition (omeprazole 60 mg). Toskes's study 47 showed low-dose (lipase 7×5000 USP units/day) and high-dose (lipase 7×20 000 USP   43 suggested that PERT administration during or after meals may be more appropriate than before meals. In all other studies, PERT was taken before or with meals or snacks at various different doses. The use of proton pump or H 2 inhibitors was reported or allowed in five 34 39 44 46 48 with only one 39 comparing PERT with versus PERT without omeprazole; in this last study, omeprazole improved fat digestion but compromised protein digestion.

Meta-analysis results
In total, 14 of 17 included RCTs 32-38 40 42 44-48 had data on predefined clinical outcomes of interest suitable for quantitative comparison of PERT versus baseline, versus placebo and versus PERT.

PERT versus baseline
The clinical outcomes of PERT versus baseline are presented in figure 3 and summarised in  46 48 without significant effects on stool frequency. 37 48

PERT versus placebo
The clinical outcomes of PERT versus placebo are displayed in figure 4 and summarised in table 3. Pooled results from seven studies 35 35 38 40 42 48 were consistently attenuated by PERT compared with placebo (all p≤0.005), as was abdominal pain ( p=0.01), 46 48 with a tendency to improved faecal consistency but not stool frequency and flatulence. 45 46 48 Pooled data from four studies 38 45 46 48 found adverse event profiles similar with PERT as placebo ( p=0.9).

Subgroup analysis
Subgroup analyses of PERT versus baseline for studies of highquality, parallel multicentre, sample size ≥40 and in Western populations did not change primary meta-analysis results or statistical heterogeneity for CFA and FFE (table 4). Subgroup analyses of PERT versus placebo were similar, but heterogeneity was significantly reduced when sample size ≥40 for CFA (I 2 =86% to 74%) and FFE (I 2 =85% to 67%).

Sensitivity analysis
Sensitivity analyses were performed as described in online supplementary table S5. For PERT versus baseline, the results and heterogeneity of CFA and FFE were not affected by diagnostic criteria for CP or EPI or by pancreatic surgery or by only including studies with a proper run-in phase. Heterogeneity in PERT versus placebo disappeared however for both CFA (I 2 =0%; p=0.70) and FFE (I 2 =0%; p=0.92) when only including CP defined by imaging and/or histology.

Meta-regression analysis
Covariates of study design ( p=0.04) and daily lipase dose ( p=0.07) appeared to contribute to heterogeneity for FFE (see  online supplementary table S6); age, gender, study quality and year of publication did not; limited data prevented assessment of alcohol and DM.

Publication bias
There was no significant evidence of publication bias for either CFA (see online supplementary figure S1) or FFE (see online supplementary figure S2)

DISCUSSION
We found PERT to improve fat and protein absorption significantly in CP, demonstrated by marked, consistent increases in CFA and CNA compared with baseline or placebo. Significant reductions in FFE, FNE, faecal weight and improvements in GI symptoms were also observed across RCTs, unchanged by subgroup, sensitivity and meta-regression analyses. Unlike in the previous Cochrane review with meta-analysis of only two studies, 21 our meta-analysis of 14 RCTs demonstrates that PERT is clearly indicated in CP for EPI. Even though long-term effects on complications and mortality could not be determined, these findings are similar to PERT for EPI in cystic fibrosis, 22 23 which, if extrapolated, suggest long-term benefit. Although no RCT was conducted over a period longer than 2 months, two open-label extensions of up to 1 year demonstrated significant improvements in serum nutritional parameters, weight, GI symptoms and QoL. Despite PERT not wholly normalising fat absorption, driving the search for better enzymes and/or correction of other factors, 9 the safety profile of PERT was comparable to placebo, also favourable in the extension studies. Healthy nutrient digestion and absorption requires complex coordination of mechanical and enzymatic breakdown of food, a finely tuned process depending on integration of multiple upper GI functions including regulated, plentiful pancreatic exocrine and biliary secretion. 9 Breakdown of fat relies on pancreatic lipase, which is highly susceptible to gastric acid; normally pancreatic bicarbonate secretion ensures intra-duodenal pH of 5-6 for optimal enzymatic activity, also preventing bile acid precipitation. One study found that endogenous lipase activity has to fall to <10% of normal before steatorrhoea occurs 15 and thus PERT should deliver ≥10% of normal lipase activity. The amount of postprandial lipase secreted in a healthy adult is estimated at between 9000 and 18 000 USP units/min for up to 4 hours. 9 Therefore, delivery of at least 100 000 USP units per meal (up to 400 000 USP units/120 000 international units in severe EPI) is required to correct EPI in adult CP; although we found a trend for higher doses to increase CFA, this increase was not statistically significant and did not fully correct malabsorption. Gastric acid inhibition can further enhance the efficacy of PERT; 44 51 52 we found equal efficacy of PERT at higher doses alone compared with lower doses with gastric acid suppression therapy, and an RCT not varying PERT between groups found acid suppression therapy to significantly improve fat absorption. 52 Further improvement may occur if PERT is given during meals, 43 corresponding with normal peak enzyme secretion some 30 min after food, followed by an elevated plateau. 9 Future RCTs are required for definitive conclusions on PERT optimisation.
EPI is frequent but variable in CP and typically progressive over a number of years, 1 3 5 8 contributing to long-term complications from malnutrition. Quantification of CFA and/or FFE is rarely undertaken routinely and clinical assessments of EPI are inexact; EPI may be inferred from patient and imaging characteristics, deficiencies in fat-soluble vitamins and osteoporosis, or identified by endoscopic pancreatic function testing (normal peak pancreatic bicarbonate secretion >80 mEq/L). 2 Lipid-soluble vitamins, retinol-binding protein, albumin and prealbumin may be useful to monitor responses to PERT. 49 50 Nevertheless, once the diagnosis of CP is established and since some degree of EPI is likely, 3 8 16-18 PERT is the treatment of choice to reduce and/or minimise long-term malnutrition, unless pancreatic secretion is demonstrated to be normal. Support for this comes from a prospective, non-randomised, multicentre 1-year cohort study of 206 patients with EPI from CP already on PERT and 88 with newly diagnosed EPI from CP newly prescribed PERT. 53 EPI was identified by maldigestion, diarrhoea/ steatorrhoea, weight loss, meteorism, dyspepsia, recurrent pain, nausea and vomiting. PERT was associated with significant reductions in recurrent abdominal pain, GI symptoms and GI QoL index (all p<0.001) in both cohorts, although uncertainty remains as to how PERT reduces intestinal and/or pancreatic pain. Despite use of the more conservative random-effects model, we found significant heterogeneity between studies. Subgroup analyses, however, did not alter estimates of the effect of PERT versus baseline or placebo on CFA and FFE, nor estimates of heterogeneity, although restriction of analysis to larger studies reduced heterogeneity. Sensitivity analyses did not alter estimates for CFA and FFE, but heterogeneity was abolished for FFE in PERT versus placebo when CP was diagnosed by imaging and/or histology criteria. The heterogeneity identified highlights the need for greater international consensus on the definition and diagnosis of CP. 54 Access to medical expertise, compliance, diet and lifestyle is heavily influenced by health inequalities. 4 Proxy indicators, notably alcohol usage and cigarette smoking, are themselves independently associated with progression of CP and nutrient deprivation. 4 In our study, 76.4% of patients were male and 89.1% of patients had alcohol-associated CP, but none of the RCTs reported measures of smoking, residence, socioeconomic or employment status, diet or comorbidity. As genetic factors and smoking have become increasingly recognised in CP progression, 2 3 4 health inequalities should be addressed in future studies of PERT to increase the applicability of findings to all patients with CP.