Proposed Thresholds for Pancreatic Tissue-Volume for Safe Intraportal Islet-Autotransplantation after Total-Pancreatectomy

Abstract

The simple question of how much tissue volume is really safe to infuse in TP-IAT for chronic pancreatitis precipitated this analysis. We examined a large cohort of CP patients (n=233) to determine major risk factors for elevated portal pressure during islet infusion, using bivariate and multivariate regression modeling. Rates of bleeding requiring operative intervention and portal venous thrombosis were evaluated. The total tissue volume per kg body weight infused intraportally was the best independent predictor of change in portal pressure (ΔPP) (p<0.0001; R²=0.566). Rates of bleeding and PVT were 7.73% and 3.43%, respectively. Both TV/kg and ΔPP are associated with increased complication rates, although ΔPP appears to be more directly relevant. ROC analysis identified an increased risk of PVT above a suggested cut-point of 26 cmH2O (AUC=0.759), which was also dependent on age. This ΔPP threshold was more likely to be exceeded in cases where the total TV >0.25 cc/kg. Based on this analysis, we have recommended targeting a TV <0.25cc/kg during islet manufacturing and to halt intraportal infusion, at least temporarily, if the ΔPP exceeds 25 cmH2O. These models can be used to guide islet manufacturing and clinical decision-making to minimize risks in TP-IAT recipients.

Introduction

For patients with severe and intractable chronic pancreatitis, total pancreatectomy (TP) and islet autotransplant (IAT) may be undertaken. The goal of TP is to relieve pain while the concurrent IAT ameliorates or minimizes the post-operative diabetes associated with this procedure. Although the most significant risk for complications arises from the pancreatectomy itself (1-3), there may be additional risks when this procedure is combined with IAT into the portal vein. Complications of intraportal infusion of pancreatic tissue may include elevation of portal pressures (usually transient), portal thrombosis, or bleeding. Although numerous publications address the safety of allogeneic islet transplantation, relatively little attention has been given to safety and complications resultant from IAT until recently (4-7).

During transplant of allogeneic islets (from deceased donors), the reported risks are now low (4, 5, 8), due primarily to advances in techniques and care (8-11), including administration of heparin, a minimally invasive surgical approach, and limiting the infusion to a smaller tissue volume (12). (4-7). Presumably, most centers have adopted the techniques and recommendations from the allo islet literature, although these two procedures should be considered distinct. There are major differences in the manufactured islet product obtained from a patient with chronic pancreatitis compared to a deceased donor. There is not the luxury of selecting (controlling for) the donor in IAT cases, and as a result, there is wide variability in the underlying pancreatic damage, patient demographics (including age, BMI, and prior surgical manipulation of the pancreas), size and morphology of the pancreas, and even differences in islet isolation techniques utilized to recover islets. The underlying disease process of chronic pancreatitis and the extent of surgery additionally contribute to risk of complications (3, 13, 14). Perhaps most importantly, the patient with chronic pancreatitis does not have the opportunity for selection of a new donor if the isolation fails or subsequent islet infusions as is possible with allogeneic islet transplants. Therefore, there is a need to maximize islet yield and the amount infused, as the patient only has one opportunity for transplant.

Periodically, a large tissue volume of impure islets is recovered from chronic pancreatitis (CP) pancreases, and in these cases purification is preferred to reduce complication risks such as elevated portal pressure (PP), portal vein thrombosis (PVT), or bleeding. However, traditional islet purification routinely employed in allo islet isolation may result in significant loss of islet mass, particularly when islets are mantled by exocrine tissue or of abnormal density as is frequently the case in CP pancreases and in younger donors (15). Even altered purification techniques such as those employed by our program (16) cannot guarantee 100% recovery of harvested islets. As there is a substantial metabolic benefit to the patient with islet dose transplanted per kg recipient body weight (17, 18), the decision to purify is not trivial and requires weighing short-term complication risks versus anticipated long-term metabolic outcomes. Currently, in allo islet manufacturing, purification and tissue infusion limits are a standard part of protocols, but in TP-IAT there is no standard published approach.

The objective of the current analysis was to determine the risk of complications presented specifically by the infusion of the islet product in these procedures, and to determine which modifiable factors increase the risk of complications, using the high volume of TP-IAT patients at our institution (17). We further examined the factors determining portal pressure changes during islet infusion, and sought to recommend “safe” TV and portal pressure thresholds, below which complications are minimized while preserving, as much as possible, islet mass available for transplant and thereby maximizing the potential metabolic benefit of IAT for the patient.

Methods

Patients

We analyzed isolation and IAT infusion outcomes from 233 patients undergoing TPIAT between May 2007 and March 2012, and only cases where the entire islet infusion was intraportal were considered. Selection of TP-IAT candidates and the surgical resection procedure have been described in detail elsewhere (17, 19), and have evolved over the years. For the cohort studied in this report, these methods can be considered to have been relatively stable.

Data was collected under an IRB approved study protocol with informed consent obtained from all participants (Protocol#0609M91887).

Patient demographics, islet manufacturing data, portal pressures during islet infusion, and early bleeding or thrombosis complications (detected during the first several weeks post-operatively) were recorded. Liver volume was calculated using Mosteller’s formula.

Islet Isolation & Intraportal Infusion

The method of islet isolation has been described previously (20-22). For isolation of islets from CP pancreases, a modified approach to enzyme dosing and purification has been used at our institution since 2008. A collagenase with intact C1 is used in combination with neutral protease from Clostridium Histolyticum and dosed according to pancreas weight (20). Modifications to the standard enzyme dosing are described elsewhere (manuscript in preparation), but in brief the collagenase:protease ratio and/or the unit dose may be adjusted based on individual pancreas factors, principally donor age and the severity of fibrosis. Purification was performed on a COBE 2991 processor on an individual basis, when tissue volume was deemed to be excessively large for intraportal infusion. Additionally, in 54 cases since 2008, we have utilized a modified technique for islet purification, in order to preserve endocrine mass available for transplant while also attempting to reduce the overall tissue volume. Rather than the traditional density gradients used in allogeneic transplants (1.060 and 1.100 gm/ml for light and heavy layers), heavy gradients are used by our institution (1.060 – 1.070 and >1.110) to prevent all but the most impure tissue from going to the COBE bag and accomplish a tissue volume reduction rather than a true purification. To minimize tissue aggregation and ischemia prior to infusion, the final islet product was loaded in transfer bags with a maximum of 10cc tissue allowed per bag.

Upon arrival of the islet product at the OR, the patient was given 70-100 u/kg heparin bolus which is allowed to circulate at least 3-5 minutes. The splenic vein stump or the middle colic vein were cannulated and attached to pressure tubing with an inline manometer which was typically positioned over the patient’s pubis. The islets were infused by gravity into the portal venous system. Infusion was done over 15-60 minutes, depending on tissue volume and changes observed in portal pressure. Portal pressure was measured before infusion (baseline), and at minimum, after each bag. About 15 minutes after completion, the portal pressure was rechecked (final). The highest recorded pressure was labeled the “peak” pressure. The ΔPP was defined as the difference between the peak and baseline PP. The patient was placed on a heparin drip and/or transitioned to enoxaparin at the discretion of the individual surgeon. In the earliest cases in this report, anticoagulation may have occurred only during the infusion. Duplex ultrasound or CT of the liver was often performed post-operatively (typically at day 5-7), and if normal the enoxaparin was discontinued. However, in earlier cases, imaging was often performed selectively—in high risk cases where portal pressure or tissue volume were high or if there were clinical concerns for complications. In the event of high platelet counts, aspirin was administered at the surgeon’s discretion.

Statistical Analysis

The effect of patient demographic, islet isolation, or infusion characteristics on complication rates were reviewed and analyzed using JMP 9.0.0 and SAS 9.3 software (SAS Institute, Cary, NC). Summary statistics are expressed as mean ± standard deviation. A two-sample t-test was used for comparison of numerical means, and Pearson’s chi-square for differences in categorical variable frequencies. All comparisons were two-tailed. After least squares linear regression to identify predictors of ΔPP, multivariate linear regression with stepwise elimination was performed using only factors with a p-value ≤ 0.1. After one-factor logistic regression and ROC analysis, forward step-wise regression was employed to create a parsimonious model that included (with increased PP) additional covariables independently associated with complications. Standard ROC curve analysis and cutoff plots were used to identify the thresholds for increased portal pressure and age that maximized sensitivity and specificity, and the area under the curve used to measure discrimination of the final models; Hosmer-Lemeshow was used to ascertain model calibration. A P value <0.05 was considered to be statistically significant for all analyses.

Results

The 233 IAT recipients included in the analysis had a mean age of 34.8 +/− 13.9 years, weight 68.0 +/− 20.1 kg, and BMI of 24.7 +/− 5.75 kg/m² (Table 1). Patients received a mean islet dose of 3716 +/− 2117 IEQ per kg body weight, corresponding to a tissue volume of 8.92 +/− 6.0 cc (range 0.3 to 32), an elevation in portal pressure from baseline of 12.6 +/− 10.2 cmH2O (−11 to 46), with peak portal pressure of 15.1 +/− 10.8 cmH20 (−4 to 48).

Table 1.

Basic demographics for patient cohort and transplanted islet products (n=233)

	Mean +/− SD	Range
Height (m)	1.65 +/−0.12	1.15-1.93
Weight (kg)	68.0 +/− 20.1	21.9-118.9
BMI (kg/m2)	24.7 +/− 5.75	13.7-40.8
Age (yr)	34.8 +/− 13.9	5-72
Liver Volume (cm3)	2834 +/− 601	945-4300
Severity of Fibrosis (0-10)	5.60 +/− 2.27	1-10
IEQ Tx	244133 +/− 142703	3207-759000
per kg	3716 +/−2117	36-9561
IPN Tx	276389 +/− 154037	5400-959900
per kg	4305 +/− 2491	60-12235
Tissue Volume Tx (cc)	8.92 +/− 6.0	0.3-32
per kg	0.138 +/−0.09	0.003-0.440

Overall 11.1% of patients had a complication possibly related to IAT, with 19 (7.7%) experiencing bleeding necessitating invasive intervention and 8 (3.4%) with partial PVT. Post-infusion doppler ultrasound or CT were performed in 52.8% of the cases, generally if the tissue volume or ΔPP for infusion was high but at the individual surgeon’s discretion.

As portal hypertension is the primary control/monitoring factor during the islet infusion process, factors contributing to ΔPP were screened using least squares regression (Table 2), which identified a number of factors as significantly associated with the change in portal pressure from baseline (ΔPP). A greater increase in ΔPP was associated with female gender, and increasing IPN, IEQ and TV transplanted. In contrast, 7 factors were associated with a decrease in ΔPP, including size index (IEQ/IPN), height, weight, body surface area, liver volume, severity of pancreas fibrosis, and islet concentration (IEQ/cc tissue). Correlation with ΔPP was improved when TV was normalized to body weight, BMI, or BSA in all cases. Overall, the best predictor of ΔPP was the amount of tissue transplanted per kg recipient body weight (TV/kg) (p<0.0001; R² = 0.547), with a quadratic fit slightly improving the model (p<0.0001; R² = 0.566), and with both the linear (p<0.0001) and quadratic terms (p=0.0019) statistically significant. The single-factor quadratic fit model with predictive equation is shown in Figure 1.

Table 2.

Bivariate patient and islet product risk factors for portal pressure change

Factor	B (SE)	Intercept (SE)	p-value	R2
Age (yrs)	−0.046 (0.049)	14.2 (1.8)	0.343	0.004
BMI (kg/m2)	0.196 (0.118)	17.4 (3.0)	0.099	0.012
Surgeon	NA	NA	0.366	0.018
Index (IEQ/IPN)	−4.606 (2.067)	16.7 (2.0)	0.027	0.021
Weight (kg)	−0.081 (0.033)	18.0 (2.4)	0.017	0.025
Height (m)	−14.206 (5.572)	36.0 (9.2)	0.011	0.028
BSA (m2)	−3.696 (1.409)	23.1 (4.1)	0.009	0.029
Liver Volume (cc)	−0.003 (0.001)	20.8 (3.2)	0.009	0.029
Female Gender	NA	NA	0.008	0.038
IEQ Tx	2E-05 (4E-06)	8.3 (1.3)	0.0002	0.059
Severity of Fibrosis (0-10)	−1.247 (0.285)	19.5 (1.7)	<0.0001	0.077
Islet Conc (IE/cc)	−5E-05 (1E-05)	14.7 (0.8)	<0.0001	0.087
IEQ Tx per Kg	0.0017 (3E-04)	6.4 (1.3)	<0.0001	0.117
IPN Tx	3E-05 (4E-06)	5.1 (1.3)	<0.0001	0.165
IPN Tx per Kg	0.002 (2E-04)	3.8 (1.2)	<0.0001	0.240
Tissue Tx (cc)	1.112 (0.086)	2.7 (0.9)	<0.0001	0.425
Tissue Tx per Kg	82.4 (5.0)	1.2 (0.8)	<0.0001	0.547

Figure 1. Current predictive model of portal hypertension based on transplanted tissue normalized to patient body weight.

Increase of portal pressure from baseline is best predicted by a second-order relationship with tissue transplanted per kg recipient body weight, according to the equation shown (p<0.0001; R2 = 0.566). Recommended thresholds of 0.25 cc/kg and 25 cmH20 are indicated, as well as 95% confidence and prediction intervals. Incidence of complications are also indicated for bleeding (“B”/red) or portal venous thrombosis (“T”/blue).

When these factors were entered into a multivariate model, tissue transplanted per kg (p<0.0001) and total transplanted islet particle number (IPN; p=0.0417) were independent risk factors, although the latter was not significant when included with the quadratic fit model (p=0.1853). In the final model (single-factor quadratic), a TV/kg >0.25 cc/kg was also predictive of a high risk of elevated ΔPP above 25 cmH20.

When the dataset was stratified by either high TV transplanted (>0.25cc/kg) or high ΔPP during infusion (>25 cmH2O), thresholds currently in use at our institution, a significant difference in the rates of bleeding and thrombosis were seen (Table 3). A high TV was associated with a significantly increased risk of bleeding (p=0.0402; relative risk 2.71 with 95% CI of 1.04 – 7.03), but there was no difference in rate of PVT (p=0.996; relative risk 1.00 with 95% CI of 0.13-7.88). Patients with high ΔPP had a trend towards an increase in rate of bleeding (p=0.0905; relative risk 2.30 with 95% CI of 0.88-6.02), and a highly significant increase in PVT (p<0.0001; relative risk 9.95 with CI 2.50 – 39.7). In a single-factor ROC curve analysis, a cut point of ΔPP of 26 cmH2O had a reasonable area under the curve (0.759) for predicting thrombosis events (but not bleeding), with maximum sensitivity and specificity (89.2 and 62.5%, respectively), PPV and NPV of 0.08 and 0.99, indicating possible utility and justification for the proposed thresholds. After multivariate logistic regression modeling, two factors were independently associated with PVTs: increased portal pressure again (p=0.006) and patient age (p=0.044). The ROC curve for the multivariate model and cutoff plot is shown in Figure 2, with an improved AUC of 0.82 and calibration verified by Hosmer-Lemeshow (p=0.569). The optimal cut-point for the multivariate model was 0.040, suggesting that the portal pressure cutoff varies with patient age, with a greater tolerance for portal pressure in younger patients (Table 4). This model provided a sensitivity and specificity of 75.0 and 76.0%, with PPV and NPV of 0.10 and 0.98. Other cutpoints that increase the PPV (0.15 at logit 0.095 and max of 0.22 at logit 0.120) are shown for comparison, as emphasis on specificity could be considered a stronger method clinically.

Table 3.

Complication rates by current thresholds for high/low tissue volume or portal pressure increase (RR=relative risk; CI=95% confidence interval)

	Bleed Rate (%)	Thrombosis Rate (%)
Overall	7.73% (19/233)	3.43% (8/233)
Low Tissue	6.37% (13/191)	3.43% (7/197)
High Tissue (<0.25 cc per kg)	17.24% (5/24)	3.45% (1/28)
	p=0.0402	p=0.996
	RR = 2.71 (CI:1.04-7.03)	RR = 1.00 (CI:013-7.88)
Low Portal Pressure	6.60% (13/184)	1.52% (3/194)
High Portal Pressure (>Δ25 cmH20)	15.2% (5/28)	15.2% (5/28)
	p=0.0905	p<0.0001
	RR = 2.30 (CI:0.88-6.02)	RR = 9.95 (CI:2.50-39.7)

Figure 2. Receiver operating characteristic curve analysis of portal pressure changes and age for incidence of thrombosis with cut-off plot.

An overall cut-off of 26 cmH20 maximized sensitivity at 89.2% and specificity at 62.5%, PPV and NPV of 0.08 and 0.99, with an area under the curve of 0.759 (not shown). When performed multivariately, the cut-off was also dependent on age with a recommended logit of 0.040 according to cut-off plot analysis (right panel), and with higher portal pressure potentially tolerated by younger donors (see Table 4). Sensitivity and specificity were 75.0 and 76.0%, with PPV and NPV of 0.10 and 0.98. Similar ROC analysis for bleeding or tissue volume did not produce statistically significant models.

Table 4.

Recommended cut-points for maximum change in portal pressure dependent on patient age at sample logits of risk. Maximum sensitivity and specificity are achieved at a logit of 0.040, however maximum positive predictive value at 0.120. As the latter reflects increased specificity, this may be the preferred approach as there will be reduced patients with a predicted “safe” ΔPP who experience a thrombosis.

Logit of Risk  Cut-point	0.040*	0.095	0.120**
Patient Age (yrs)
20	31	39	40
30	24	35	36
35	22	30	33
40	19	27	30
50	13	20	24
60	7	15	18

^*Maximizes sensitivity and specificity

^**Maximizes positive predictive value

In this series, 8 patients had evidence of a thrombotic complication involving the portal venous system, which we examined for risk factors and outcomes (Table 5). Five of eight patients had a ΔPP above the 25 cmH20 threshold, while only one had >0.25 cc/kg. The four earliest patients received bolus heparin only (due to increased risk of bleeding), while the later four also received anticoagulation during follow-up. Thromboses were detected by ultrasound (n=7) or CT (n=1) incidentally and were not associated with liver dysfunction aside from mild and transient elevation of liver enzymes in three patients. Five involved the main portal vein, 4 of which were non-occlusive. In total, the left portal vein was more often involved (6/8), and in 4 of 6 were partially thrombosed. Most of the thrombosis in the series were treated with Coumadin and were resolved on follow-up imaging. No cases of thrombosis were associated with postoperative bleeding. These patients had modest islet yield (mean 4217 IE/kg, range 2440-5587) and 1 was insulin independent, 1 required once-daily Lantus, while 6 required basal-bolus insulin regimen.

Table 5.

Case details of patients with documented portal vein thrombosis after IAT

Case #	Age	Weight	Tissue Volume	Tx tissue per Kg	Tx IEQ per Kg	PP Change	Complications Comments	Follow-up Function
1	48	65.6	15	0.228	3863	29	POD 11, thrombosis of superior mesenteric vein, portal vein, R and L portal vein	Basal only (12 units/day)
2	53	60.3	1.5	0.025	3422	3	Small non-occlusive thrombus in main portal vein on US	Independent
3	44	113.8	12	0.105	2754	10	Non occlusive thrombus of the main portal vein seen on CT scan.	Basal-bolus (~39 units/day)
4	48	57.9	10	0.173	3045	27	Thrombus in left and main portal vein, which is partially occlusive. No heparin at islet tx.	Basal-bolus
5	39	44.0	14	0.318	3418	34	Non-occlusive thrombus in main portal vein and occlusive thrombus in left portal vein; h/o DVT at PICC lines in arms	Basal-bolus, labile blood sugars
6	43	69.4	14	0.202	2441	18	Thrombosis of left portal vein	Mostly basal-bolus, various dosing, h/o nocturnal hypoglycemia
7	30	78.2	18	0.230	3624	30	Left portal vein thrombosis on U/S, clinically asymptomatic; h/o prior clotting episodes including PE	Basal-bolus (2-3 inj/day), 44 units/day
8	38	79.7	16	0.201	5587	27	Thrombus in the left portal vein observed POD1, subsequently in anterior left portal vein alone, and then entirely resolved	Basal-bolus regimen, 31 units/day

Discussion

A most basic question precipitated the analysis in this study. How much dispersed pancreatic tissue is really safe to infuse following total pancreatectomy for IAT? The answer to this has importance for both islet manufacturing and OR/clinical personnel and we sought to provide an evidence-based threshold for both. The single-factor quadratic fit model presented here has been adopted by our TP-IAT team to help guide islet manufacturing (ie. decision to purify) and islet infusion (ie. decision to stop intraportal infusion) since approximately mid-2011, and the equation shown in Figure 1 used to predict ΔPP depending on TV and the recipient’s body weight. Additionally, a TV per kg of 0.25 or a ΔPP of 25 cmH2O have been identified as thresholds to aid in decision-making, as they are convenient to calculate and memorize and are relevant to complication risks as demonstrated in this large and recent cohort of TP-IAT patients from our institution.

Other islet transplant centers have reported and suggested cut-offs ranging from 20-31 cmH2O or 5-20cc of total tissue (4, 6, 8, 10, 12, 23), although these may be limited because they were mainly derived in the allo islet (deceased donor) setting where other factors such as previous islet transplants or method and site of infusion likely play a role (24, 25). These suggestions may not be directly relevant to TP-IAT, but may prove useful for comparison. There are many potential differences between the allo and auto islet transplant environment that may influence maximal tolerated tissue volume, namely alteration of portal flow dynamics after splenectomy, as well as inflammation mediated by the infusion of exocrine tissue (auto) versus the immune response (allo). Perhaps the major factor that theoretically decreases the maximum allowable tissue volume in allo islet transplantation is the pro-inflammatory nature of islets from deceased donors as examined in several studies (26). The Baylor group recently demonstrated that the instant blood-mediated inflammatory response is active in TP-IAT (personal communication), however it may be to a lesser degree in these living donors. To our knowledge this was the first demonstration of IBMIR in IAT and deserves further study.

Only a few other centers have reported on infusion risks of islet autotransplants, but in those instances only a small number of cases were included and the method of infusion was percutaneous and not intraoperative as done at our institution (4, 6). While the Baylor group also recently demonstrated that high tissue volume is associated with portal pressure rise and therefore complications, our analysis extends this further to provide a strong predictive model of portal hypertension based on tissue volume and the relative risks of high tissue volume and high portal pressure for post-operative bleeding or PVT.

Importantly, correction of the tissue volume to body weight was found to significantly improve the predictive ability of the model. This supports our historical belief that a larger volume is safe to infuse in larger patients (but should be reduced in smaller patients), which is a critical consideration especially in the pediatric setting where patient size varies significantly with age. The presence of curvature in the model may suggest that there is a theoretical maximum change in portal pressure that can occur during islet infusion, although this needs to be considered more carefully. We speculate that the arterial capacity for compensation is finite, and have observed that more rapid infusion of tissue (by gravity) results in steeper spikes in portal pressure which partially resolves after a rest period. Therefore, portal pressure appears to be a function of infusion time, but we do not yet understand this fully. We have recently begun to perform routine dynamic portal flow monitoring (in addition to portal pressure) during the infusion, and these show a close inverse association (unpublished data). Full recovery of portal flow to baseline typically occurs over the course of days to weeks in our experience, although more rigorous examination of these dynamics would be of value. The curvature in the model could also be explained by altered clinical care practices (ie. slower infusion rate) when the portal pressure is elevated that are not precisely documented and could not be analyzed.

Although the model presented adequately predicts the change in portal pressure, some variation still remains which could not be accounted for and which may explain some of the lack of predictive value of patients experiencing a PVT even with sub-threshold ΔPPs and/or tissue volumes (Table 5). Surprisingly, no factors related to size, volume, or purity of the islet products transplanted were able to improve the predictive ability, nor was the calculated BSA or liver volume of the patients. These might be expected to affect PP, as large endocrine or exocrine tissue particles may occlude a greater fraction of the liver vascular volume by lodging further upstream compared to smaller particles. Regardless, this finding may be of particular importance in this patient population, as these islet products are infrequently purified in order to avoid loss of islet tissue (approximately 1 in 6). We speculate that other factors such as the specific anatomy of each liver (vascular volume, branching structure), heterogeneity of tissue distribution during infusion, catheter position, or other measurement variability may have larger effects. Unfortunately, these aspects are particularly challenging to measure. Factors unrelated to mechanical obstruction that might contribute were discussed previously by our group (27), principally related to hemodynamic effects. In light of our current study, it is appealing to speculate that factors like clot, islet thrombus, or aggregate formation may effectively increase the TV of the preparation and therefore portal hypertension and complication rates in a direct or indirect fashion.

Risk for bleeding or PVT are probably inversely correlated and heavily dependent on the management of coagulation in the peri- and post-transplant periods. This may be one significant factor leading to inter-site or inter-surgeon variation in complications rates – aggressiveness of anticoagulation. Rates of bleeding in islet transplantation reported elsewhere cannot be compared to this cohort, as those rates are confounded by the use of the percutaneous infusion approach necessitating ablation of the catheter tract (12). However, incidence of PVT between sites is more relevant, and the rate reported here (3.43%) coincides with the 0 - 3.7% reported in the more TV-limited allo setting (9, 12), and superior to the rates of 5.6% - 14.8% reported elsewhere for IAT (4, 6). Although there is some evidence that the open intraoperative infusion approach may partially explain this (5), the low complication rate is likely due to extensive experience in CP patient management, case volume, and consistent anticoaguation approaches to the islet product and patient with routine follow-up monitoring. By way of comparison of these rates, patients with diagnosed acute or chronic pancreatitis not undergoing pancreas surgery, experience incidence of extrahepatic thrombosis of 10-37%, with numerous pathophysiological mechanisms implicated (13) and some cases of PVT occurring as well. Clearly there are inflammatory processes at work in these patients which may differ significantly from allo islet recipients. Although the islet products transplanted in these autologous cases are potentially less pro-thrombogenic, it is also possible that (depending on type or stage of pancreatitis) these patients are experiencing systemic inflammation and are more susceptible to thrombosis once the islet product is infused.

Consistent procedures and care are critical to minimizing complication risks. Currently at our institution, the islet manufacturing team calculates the theoretical threshold TV for each patient based on body weight prior to the islet isolation and uses this value as a guideline throughout the process. Should the recovered tissue exceed 0.25cc/kg, purification is considered based on the likelihood of a good recovery (mean = 85 +/− 20%; unpublished) versus the predicted ΔPP. If purification is performed, fractions can be combined based on individual tissue volumes to ensure the overall islet product conforms to the threshold. However, we routinely provide more tissue than the calculated threshold to the OR when available, in the event that the pressure is lower than predicted for a given patient. To protect against aggregation of the tissue, the manufacturing team uses 35 U/kg heparin in the islet product, limits the tissue volume in non-conical-bottom bags, and frequently mixes the bag with agitation when possible. Addition of heparin in the islet product bags is a recent protocol change, and was not used in the cohort studied in this report.

Islet infusion and care protocols are as described in the Methods, with a few additional points. Islet product is infused currently with real-time flow monitoring in addition to the intermittent pressure monitoring, which will likely prove to be a superior method. Some surgeons prefer to use a heparin drip prior to the end of surgery that may be adjusted based on the observed portal hypertension (5-10U/kg/hr), while other patients are transitioned to enoxaparin at 0.5 mg/kg through approximately POD7 when routine duplex ultrasound is performed to verify that no thrombosis is present. This remains the only protocol difference at the current time, but current data suggests that these are equivalent anticoagulation approaches (unpublished data).

In the event that portal pressures exceed the threshold or flow is severely reduced, infusion is interrupted for 10-30 minutes but may be continued if pressures recover. Purification during the manufacturing procedure is balanced with preserving islet mass therefore, because when all cannot be infused intraportally the remaining islets are placed alternatively, principally in the peritoneal cavity or submucosa. We do not yet know the capacity for function and longevity of these islets, and the ideal alternate site is an important area of future research, particularly for this patient population.

In addition to those mentioned already in the text, limitations of this study are that it was a retrospective, single-center design with some variations in inter-surgeon care protocols, other era effects, a low event number, and possible detection biases due to different imaging modalities. We also recognize that unexplained variability in the models adds complexity, particularly to the sensitivity and specificity in predicting complication rates based on the proposed thresholds. As there is no perfect cut-off, we have recommended what we feel is the most relevant and useful thresholds to guide decision making based on current data, with an emphasis on reduction of false positives.

In conclusion, it should be remembered that the goal of IAT is to minimize the impact of postoperative diabetes and that increasing evidence suggests that IEQ/kg is the predominant factor determining this outcome. As our islet manufacturing protocols have also continued to evolve (16, 20), islet yields and tissue volume have increased, placing more emphasis on transplanting larger volumes, and by extension the best possible anticoagulation practices and/or need for improved alternate sites. We suggest that a larger tissue volume than has traditionally been considered can be infused with minimal complications in a controlled setting. However, this will be guided by each clinician’s individual perspective on the balance of complication risks versus the possible metabolic benefit.

Abbreviations

AUC

Area under the curve

BID

Twice a day

BSA

Body surface area

CP

Chronic pancreatitis

IAT

Islet autotransplantation

IEQ

Islet equivalent

IPN

Islet particle number

POD

Post-operative day

PP

Portal pressure

PVT

Portal venous thrombosis

ROC

Receiver operating characteristic

TP

Total pancreatectomy

TV

Tissue volume