Introduction

Patients with minimal change chronic pancreatitis (MCCP) may experience debilitating abdominal pain, recurrent acute pancreatitis, and a declining quality of life. Furthermore, the pancreas of a patient with MCCP lacks the traditional morphologic changes characteristic of large duct calcific disease.1 , 2 Consequently, management of these patients is often challenging as they are usually not candidates for traditional surgical procedures or endoscopic management; most patients are managed medically with varying degrees of success.3 The majority of these patients eventually experience a progressive decline in functional status, and while numerous operative strategies have been proposed to address this disease process, each of these series has been limited to small patient populations at single-centers.4 – 10

Our center has been performing total pancreatectomy and islet cell autotransplantation (TPIAT) as a means of treating patients with severe chronic pancreatitis that have failed previous surgical management, and we have also utilized this strategy in patients with genetic and hereditary pancreatitis.11 Many of these latter patients often present with minimal change pancreatitis (MCP), and our experience with these patients has demonstrated a great benefit of TPIAT. Based on this experience, we started applying a similar treatment algorithm to patients that presented with MCCP secondary to other etiologies. TPIAT adequately treats the disease process through complete gland excision while abrogating Type 3c diabetes via relocation of the patient’s own islet cells into the liver. It is also likely that many patients with MCCP have an undetected underlying genetic mutation and may be at higher risk of developing cancer. Thus, the strategy of TPIAT eliminates this risk. While the indications for TPIAT continue to evolve in patients with refractory MCCP, TPIAT is considered as a first-line surgical therapy at a number of institutions, including ours.

Despite the early success, TPIAT remains a substantial undertaking for patients that must carefully consider the associated cost and morbidity, balancing the prospect of long-term insulin and narcotic independence against short-term surgical risks and costs. The aim of this study was to evaluate the cost-effectiveness of TPIAT as a first-line surgical intervention in refractory MCCP using each patient’s preoperative course as the medical management control group. We hypothesize that TPIAT decreases long-term resource utilization and improves quality of life, justifying the initial costs and risks of the procedure.

Materials and Methods

Study Cohort

The University of Cincinnati Pancreatic Disease Center TPIAT Clinical Database was queried to identify all patients undergoing TPIAT for MCCP from 1999 to 2013 (n = 84). Pertinent demographic, diagnostic, and perioperative data were obtained for each patient. Additional retrospective chart review was performed to identify the patients with complete documentation that included both the 12 months prior to and post-TPIAT (n = 46). Patients without adequate documentation over the study periods were excluded from the analysis (n = 38). Clinical data were collected for each patient, which included hospital admissions, endoscopic procedures, narcotic and insulin requirements, and radiologic imaging over the entire study period.

Each MCCP patient included in this study underwent TPIAT as the initial surgical intervention for either (1) debilitating abdominal pain refractory to medical management or (2) recurrent, acute pancreatitis with documented enzyme elevations. MCCP was diagnosed based on imaging, functional testing, and patient symptomatology. Imaging included CT, endoscopic ultrasound (EUS), and endoscopic retrograde cholangiopancreatography (ERCP). Functional testing included measurement of fecal fat and elastase levels in conjunction with pancreatic bicarbonate levels. In most patients, diagnosis was made by EUS, ERCP, and CT findings. Confirmation with EUS required the presence of at least four or more of the nine conventional criteria: parenchymal findings included hyperechoic foci, hyperechoic stranding, hyperechoic lobularity, and cysts, while ductal findings included irregular duct contour, hyperechoic ductal margin, visible side branches, ductal calculi, and dilated main duct.12 If a patient had less than four Rosemont findings but a concomitant major finding (stones, honeycomb lobularity), then the patient was also diagnosed with MCP. Diagnosis confirmation by ERCP was based on the Cambridge classification.13 Patients with Cambridge II and III were designated as having MCCP. This study was approved by and conducted under the guidelines of the University of Cincinnati Institutional Review Board.

Cost-Effectiveness Modeling and Assumptions

A Markov health-state transition model was constructed to compare two strategies: (1) continued medical (nonsurgical) management of MCCP and (2) surgical intervention, comprised of TPIAT (Fig. 1a, b). The health states for each study cohort included the following four permutations of insulin and narcotic use: (1) insulin independent/narcotic independent (II/NI), (2) insulin independent/narcotic dependent (II/ND), (3) insulin dependent/narcotic independent (ID/NI), and (4) insulin dependent/narcotic dependent (ND/ID) and also death. In the surgical cohort, one additional health state was created to represent surgical intervention. The Markov model employed a 1-month cycle length, beginning at age 36 years (mean age of study cohort) and a lifetime horizon (defined as ending at age 100 years).

Fig. 1
figure 1

Schematics of Markov model. a Surgical intervention with representative postoperative modeling and b medical management model; NI narcotic independent, ND narcotic dependent, II insulin independent, ID insulin dependent

Full size image

Patient intervention and outcome probabilities and rates were based on the University of Cincinnati MCCP patient data. Specifically in the surgical cohort, pertinent perioperative outcomes included perioperative mortality, complication rate, and readmission rate. In both cohorts, frequency and volume of three-dimensional imaging (CT), endoscopy (EGD, ERCP, and/or EUS), insulin use, and narcotic use were incorporated into the model. Means were utilized throughout the analysis to ensure all outcomes were incorporated, though not all data were precisely normally distributed. The model assumed that patients who did not undergo TPIAT would continue their rate of resource utilization (e.g., hospital admission, endoscopy use, etc.) indefinitely. In the surgical cohort, we also assumed that while health states and readmission rates might change within one postoperative year (as dictated by our data), neither health states nor readmission rates changed significantly 1 year after surgery.

Costs

Costs were estimated using 2013 Medicare payment in US$ (Table 1). Direct medical costs included surgeon and islet preparation fees, hospital admissions, and readmissions (DRG 439 for medical admissions and DRGs 405-407 for surgical admissions), narcotic use (in morphine equivalents, MEQs), insulin use, CT use, and endoscopic procedures (EGD, ERCP, and/or EUS). Hospital payment was based on no complications/comorbidities (without CC designation, DRG 407), minor complications (with CC designation; Clavien-Dindo 1-2, DRG 406), and major complications/comorbidities (with MCC designation; Clavien-Dindo 3-5, DRG 405). Costs of endoscopic procedures were calculated with a weighted average of the frequency of the individual components of all endoscopy procedures. Costs were discounted at 3 % per year, a standard practice to reflect the lessening value of the dollar over time.14

Table 1 Costs, in 2013 Medicare US$, of model variables
Full size table

Survival and Quality of Life

Survival was calculated utilizing average annual mortality based on the US age-specific death rates as reported by the National Center for Health Statistics15 and excess mortality from the largest published series’ available for both medical and surgical management of chronic pancreatitis.16 – 18 From these rates, monthly transition probabilities of mortality were calculated for the Markov simulation. Final survival was reported in quality-adjusted life years (QALYs) after appropriate adjustment for quality of life (QOL). In the medical cohort, a QOL coefficient of 0.66 was used in the base case for patients who were either insulin or narcotic dependent, per published literature.19 Patients receiving narcotics and insulin were assumed to have a lower QOL than patients who were both narcotic and insulin independent (QOL coefficient 0.85 based on literature and expert opinion).20 , 21 In the surgical cohort, with similar differences between health states as noted in the medical cohort, the base case QOL coefficient reflected a 55 % improvement in QOL, based on literature examining long-term QOL outcomes in a surgically treated population.22 Based on the published literature, QOL was never assumed to be worse than ongoing medical management. Life expectancy also was discounted at 3 % per year.

Sensitivity Analysis

Sensitivity analyses were performed in a deterministic fashion. One-way sensitivity analysis was based on uncertainty or variability in the published literature that could potentially change the final cost and effectiveness profiles and conclusions. The parameters varied were QOL adjustment in the medical treatment strategy (QOL coefficient range 0.5–0.8), QOL adjustment in the surgical strategy (range 30–80 % improvement in QOL in comparison to QOL in medically treated patients), costs including surgical admission and complications (CC payment for no postoperative complications and MCC payment for all postoperative complications) and islet cell preparation ($10,000–$30,000; 50–200 % of our institutional cost), perioperative mortality (0–5 %), and long-term readmission rates in the medical treatment strategy (50 % below base case up to base case readmission rates).

Results

Patient Outcomes

Patient demographic information is listed in Table 2. The majority of patients were female (63.0 %, n = 29) and had either idiopathic (65.2 %, n = 30) or genetically linked (26.1 %, n = 12) chronic pancreatitis (CP). None of the patients had a previous pancreatic resection. Patients required daily narcotic pain medications at a mean of 138 MEQ/day. Only two patients required insulin prior to TPIAT.

Table 2 Patient demographics
Full size table

All patients underwent a successful TPIAT, and there were no perioperative deaths. Operative details are listed in Table 3. The mean operative time of 549.5 min included tissue harvesting, total pancreatectomy, islet isolation and transplantation, and reconstruction. Mean hospital length of stay was 17 days. In this cohort, major complication rate after TPIAT was 30.4 % (n = 14). The initial postoperative health state distribution was as follows: insulin and narcotic independent (II/NI, n = 10), insulin independent and narcotic dependent (II/ND, n = 4), insulin dependent and narcotic independent (ID/NI, n = 21), and both insulin and narcotic dependent (ID/ND, n = 11).

Table 3 Perioperative outcomes
Full size table

Postoperative outcomes regarding pain management and glucose homeostasis are listed in Table 4. At 1 year after TPIAT, narcotic requirements decreased to 37.0 MEQ/day (p = 0.012), and 69.6 % of patients (n = 32) were narcotic independent. All patients achieved stable glycemic control after TPIAT, and 32.6 % of patients (n = 15) were insulin independent at 1 year.

Table 4 Patient outcomes
Full size table

Hospital admissions, days spent in the hospital, endoscopic procedures, and imaging events are depicted in Fig. 2. In the 12 months prior to TPIAT (medical management cohort), 46 patients accounted for a total of 78 hospital admissions and a total of 502 days in the hospital. Annual mean per patient endoscopic procedures was 1.4 (range = 0–6), and imaging events (CT/MRI) was 1.5 (range = 0–4). For the surgical cohort, the 30-day readmission rate was 37.0 %. There were no postoperative deaths in this cohort, and annual mean per patient admissions, endoscopic procedures, and imaging had decreased to 0.9 (range = 0–4), 0.4 (range = 0–2), and 0.9 (range 0–5), respectively.

Fig. 2
figure 2

Clinical events used to populate the Markov transition health-state model with data from both the medical management cohort (a) and surgical management cohort (b). Clinical variables captured include the number of hospital admissions and the number of days spent in the hospital, number of endoscopic procedures, and number of imaging events

Full size image

Cost-Effectiveness and Sensitivity Analysis

Costs and quality-adjusted life expectancy for the medical management and surgical (TPIAT) management cohorts was $196,042/11.5 QALYs and $153,576/14.9 QALYs, respectively. Because the surgical strategy dominated the medical management strategy (lower costs and greater effectiveness), there was no incremental cost-effectiveness ratio (ICER) to calculate.

Results of sensitivity analyses are shown in Table 5. When QOL adjustment was made to reflect a 20 % increase/decrease in QOL above/below baseline, there was no change in cost (as expected), but quality-adjusted life expectancy ranged from 10.6 to 14.7 QALYs and 13.6 to 15.9 QALYs in the medical and surgical cohorts, respectively. Altering the relative improvement in QOL from 30 to 80 % in the surgically treated patients (in comparison to the continued medical therapy cohort) resulted in a range of 13.6–15.9 QALYs. Changing costs for patients without or with minor complications, payment for islet cell preparation, and perioperative mortality changed the cost profile by less than 10 % with no significant change in quality-adjusted survival. For each of these sensitivity analyses, the resultant cost and effectiveness profiles did not change the final conclusions—surgical therapy was always more effective at lower costs.

Table 5 Sensitivity analyses based on variation in key model inputs
Full size table

The one instance in which varying the model inputs did change the final cost-effectiveness comparison was found in changing the long-term readmission rates in the medical cohort. When rates of long-term readmission were ranged from the base case value to 50 % below baseline, the resultant cost in the medical cohort was $127,496. This scenario results in the surgical treatment strategy being more costly, though still with longer quality-adjusted survival and a resultant incremental cost-effectiveness ratio of $12,420/QALY.

Discussion

In the current study, we have examined the cost-effectiveness and resource utilization of TPIAT as compared to continued medical management in a population with MCCP. This analysis demonstrated that MCCP patients have a high burden of hospital admissions, days spent in the hospital, and diagnostic/therapeutic endoscopic procedures in addition to recurrent imaging events. Perhaps, the most profound finding was that in a medially managed cohort, a total of 502 days were spent in the hospital, with 78 hospital admissions over a 12-month time period. In the surgical cohort, there were a high number of readmissions in the immediate months following TPIAT; however, this burden was significantly reduced 6 months after TPIAT. Populating a Markov transition health-state model with data from both the medical management and surgical management cohorts, we were able to demonstrate that TPIAT is associated with both improved quality-adjusted survival but also reduced lifelong medical costs.

This study represents the first cost-effectiveness analysis comparing traditional medical management of MCCP to primary surgical intervention with TPIAT. Traditionally, medical management has been based on the notion that MCCP can only be treated medically, with no meaningful surgical intervention available for this patient population. Because these patients are not amenable to drainage procedures typically utilized for large-duct disease, therapeutic options have been quite limited. However, more recent publications have demonstrated persistent pain despite medical therapy over 10-year follow-up, and clearly, more effective treatment options are needed.4 , 23 In fact, a study by Nealon and Thompson demonstrated that in a patient cohort of 56 patients with nondilated duct CP, only one patient experienced spontaneous pain relief.4 Since that time, numerous case series have reported improved symptoms after surgical intervention, and only one treatment—total pancreatectomy—completely removes the inciting organ and source of the symptoms. With advances in islet cell autotransplantation, TPIAT has become a viable option with improved pain relief, stable glycemic control, and improved QOL.16 , 22 , 24 , 25 Only, one previous study from Leicester26 has examined the cost of TPIAT. They showed equivocal results when comparing TPIAT and medical management however did not account for quality-adjusted survival in this analysis.26 Findings of our current study add to the evidence favoring TPIAT with improved health care costs and prolonged quality-adjusted survival.

The limitations of this study must be recognized. First, the patient utilities used for QOL adjustment are not uniformly reported and consistently used across study populations. As such, QOL adjustments for both medical and surgical cohorts were independently populated, based on the largest, highest quality literature available. It is unfortunate that such QOL measurements are not more standardly gathered and reported, as consistency within a disease state would make all subsequent comparative studies more interpretable and of higher integrity. However, based on both the significant cost savings and the sensitivity analyses incorporating a wide range of QOL adjustments in the surgical cohort (including the extremely unlikely notion of no improvement in QOL after TPIAT), the conclusions in this study would not change; TPIAT is more cost-effective than medical management for MCCP. Secondly, we do not have lifelong data for these patients, and the assumptions about resource utilization and QOL may not hold true years after the index operation. Our group is currently accumulating patient-specific outcomes for patients 5 years after surgery, and beyond, which will provide increasingly accurate data for future modeling.

Lastly, our cost estimates are based on Medicare reimbursement, and this facet of our study deserves special attention. While publicly available and verifiable, these payments do not reflect reimbursements of all age groups and insurance types represented in our patient sample, and the absolute dollar value of costs reported here is likely to be low. However, the costs ascribed to patients in this study are consistent between study cohorts, and again, the sensitivity analysis demonstrates that our conclusions would not change, even with wide variability in payment for individual components of our model. Furthermore, Medicare payments are almost certainly significantly underestimated relative to payments made by private insurers. If payment amounts of private insurers were available for this study, it is quite likely that the absolute cost savings of TPIAT would be more pronounced. For example, if hospital costs for an acute exacerbation of chronic pancreatitis were reimbursed at a rate of $15,000 per admission instead of approximately $5,000 (Medicare payment), the compounding excess costs of readmission associated with medical management would make TPIAT even more desirable from a cost-effectiveness perspective.

Conclusion

In summary, this study further validates the early use of TPIAT as a first-line surgical therapy in patients with refractory, minimal change CP. There is a high burden of medical care devoted to patients with MCP managed with traditional medical therapies. TPIAT is associated with an initial upfront cost and early postoperative admission burden, but when compared to medical management over a patients’ expected lifetime, TPIAT is associated with decreased cost and increased quality-adjusted survival. These findings suggest that tertiary care centers should look to expand the use of TPIAT in this patient population and that the initial costs and risks of TPIAT are well justified.