Minimal Functional β-Cell Mass in Intraportal Implants That Reduces Glycemic Variability in Type 1 Diabetic Recipients

Patients (median age 44 years and BMI 24 kg/m²) had type 1 diabetes for >20 years, with variable glucose control before transplantation (median HbA_1c 7.7% [interquartile range {IQR} 7.0–8.4]; 46 mmol/mol [53–68]). All subjects experienced large variability of fasting glycemia (coefficient of variation of fasting glycemia [CVfg] >25% [8]) despite the use of a subcutaneous insulin pump (n = 38) or long-acting insulin analogs (n = 3) (Supplementary Table 1).

At the time of their first clamp posttransplantation (median 6 months PT [IQR 6–18]), their diabetes was better controlled compared with pretransplantation, as illustrated by an HbA_1c ≤7% (≤53 mmol/mol) in 35 of 42 (P < 0.001) and a CVfg <25% in 27 of 42 patients (P < 0.001) (Supplementary Table 1). This better glycemic control was due to a partial restoration of endogenous insulin secretion (random C-peptide >0.5 ng/dL in 34 of 42 patients). At the time of the clamp, 18 patients were off insulin, 14 on insulin pump (median insulin dose 0.32 units/kg/day), and 10 on insulin analogs (median insulin dose 0.28 units/kg/day).

Cultured islet cell grafts were prepared as previously described. Patients received in total 2.2–9.0. 10⁶ β-cells/kg body weight either in one (n = 9) or two (n = 33) injections with a median interval of 2.7 months (IQR 2.3–3.1). The islet cell graft was infused into the portal vein over 5–6 min. Access to the portal vein occurred through a catheter inserted in the umbilical vein (n = 19) (8) or through subcutaneous transhepatic puncture under ultrasound guidance (n = 23) (8). Thus far, we did not find a different loss of FBM after infusion into the portal vein by laparoscopic or radiological access (P.G., G. Delvaux, G. Maleux, B.K., unpublished observations). The immune-suppressive regimen consisted of induction therapy with antithymocyte globulin (ATG; Fresenius) at the time of the first implant, followed by a maintenance immune-suppression regimen of mycophenolate mofetil (Cellcept) plus tacrolimus (Prograft) (8).

The clamp procedure was a modification of the method of DeFronzo, Tobin, and Andres (9) as previously described (12). In short, the evening before the clamp a standard meal was taken between 5.30 and 6.30 p.m. In patients treated with long-acting insulin, the day before the clamp, this insulin type was replaced by intermediate-acting insulin, while in patients treated with pump short-acting insulin, insulin was stopped 1 h before start of the clamp so that virtually no exogenous insulin was present during the clamp procedure. Subjects remained fasted until the test with free water intake. Intake of all medications except the immunosuppressive drugs was deferred to after the clamp procedure. During the first 14 min of the clamp, a priming dose of glucose was infused to reach a glycemia level of 180 mg/dL. This priming dose was adjusted to body surface area and was, respectively, 870 mg/m² ⋅ min from 0 to 5 min, 350 mg/m² ⋅ min from 5 to 10 min, and 210 and 180 mg/m² ⋅ min after 14 min. From 15 to 170 min, glycemia was maintained at 180 mg/dL. Whole blood glucose was measured with Hemocue (Hemocue, Ängelholm, Sweden) every 5 min. The glucose infusion rate was adapted at 5- to 10-min intervals. At 150 min, 1 mg glucagon (GlucaGen HypoKit; Novo Nordisk, Bagsvaerd, Denmark) was injected intravenously to complementarily stimulate insulin secretion via the cAMP pathway. Blood was drawn for measurement of glycemia and C-peptide at −30, −15, and 0 min (basal); at 5, 7.5, and 10 min (first-phase release); at 120, 135, and 150 min (second-phase release); and at 155 and 160 min (third-phase insulin release).

C-peptide secretion was measured during the different phases of the HGC test in β-cell graft recipients. The same test was conducted in 12 age- and weight-matched healthy volunteers, and different measures were expressed as a percentage of the mean of healthy control subjects. Plasma C-peptide levels were used to calculate area under the curve (AUC) during the four phases (basal, first, second, and postglucagon phases). The AUC was divided by its time interval resulting in a concentration/minute to facilitate comparison between intervals. In a post hoc analysis of relationship with glycemic variability, phase 2 HGC as index of FBM was divided in 4 quartiles (Q 1, 2, 3 and 4), from the lowest quartile of secretion (Q1) to the highest quartile (Q4). CVfg was calculated on the basis of 120 home blood glucose measurements in a two month period around the HGC. CVfg was expressed by an absolute value in %. A decrease by >50% of the pretreatment CVfg value was indicated as “reduction by more than 50%”.

Plasma glucose was measured on Vitros 950 IC until 2006 and Vitros 5.1FS from 2007 onwards (Ortho Clinical Diagnostics, Rochester, NY), C-peptide and proinsulin by time-resolved fluoroimmunoassay (11). The proinsulin assay was considered to measure total proinsulin immunoreactive material (13). As there is 100% cross-reactivity of proinsulin in the C-peptide assay, free C-peptide concentrations were obtained by subtracting the proinsulin concentration from the measured total C-peptide result.

Since the aim of this study was to examine the relationship between FBM, measured with HGC, and parameters of glycemic control and glycemic variability we have chosen to use only 1 data point per β-cell recipient in a cross-sectional design. All patients who received a HGC posttransplantation (n = 42) were included with a wide range of FBM (between 1 and 71% of healthy controls). Data are presented as median (IQR) unless otherwise stated. Normality was tested for by using the Shapiro-Wilk test. Comparison of patient subgroups used Kruskal-Wallis H test and if significant the Mann-Whitney test for quantitative variables and Fisher exact test for binary variables. Statistical difference between repeated measurements were examined with the Friedman test and, if significant, with the Wilcoxon signed rank test. Analysis of correlations was performed using the Spearman rank correlation test. Multiple linear regression including all parameters with P < 0.25 was used for multivariate analyses using HbA_1c, CVfg, mean fasting glycemia, insulin dose (U/d and units/kg/d) and presence of insulin independence as explanatory variables and second phase C-peptide secretion as outcome variable. Both a complete model and a forward selection approach were used. Differences were considered significant for P values < 0.05. The use of AUC C-peptide during the second phase of the HGC test as predictor of CVfg <25% and reduction of pretransplant CVfg by >50% was evaluated using receiver operating characteristic (ROC) curves. All analyses were performed using SPSS statistics, 20.0. No adjustments were made for multiplicity.

In this cross-sectional study in 42 patients, secretory capacity of the implants was measured during the different phases of the HGC test. During the basal (nonstimulated) phase, secretory capacity reached 71% of that in healthy control subjects (Fig. 1). During the stimulatory phases of the HGC, median C-peptide secretion was, however, significantly lower than measurements in healthy control subjects (Fig. 1). After acute (first phase) and chronic (second phase) glucose stimulation, secretory capacity reached 45% and 18% of that in healthy control subjects, respectively. Adding glucagon showed the same secretory capacity compared with healthy control subjects as during the second phase. The secretory capacity during the second phase of the HGC was therefore used as an index of FBM. The observation that basal secretory capacity approached that of healthy control subjects is attributed to its response to higher basal glucose levels: median 110 mg/dL (IQR 97–126) vs. 84 mg/dL (79–89); P < 0.001. In contrast, glycemic levels achieved during the first phase (168 mg/dL [IQR 143–189] vs. 145 mg/dL [126–152]; P = 0.14), second phase (209 mg/dL [189–229] vs. 193 mg/dL [178–196]; P = 0.09), and postglucagon phase (214 mg/dL [189–232] vs. 193 mg/dL [177–213]; P = 0.64) were not different from levels of healthy control subjects.

The average phase 2 HGC value was 18% of that in healthy control subjects—however, ranging from 1 to 71%. A wide variation was also observed in associated metabolic parameters such as HbA_1c, fasting glycemia, CVfg, and insulin dose (Supplementary Table 1). A correlation was observed between the phase 2 HGC values and HbA_1c, CVfg, and insulin need but not with fasting glycemia (Table 1 and Supplementary Fig. 1). In the multivariate analysis, only CVfg remained independently correlated with phase 2 HGC (Table 1). The same was seen when only insulin-dependent patients were studied (Table 1). For further analysis of this correlation, patients were divided according to their FBM values into four quartiles (Fig. 2), i.e., quartile 1 with a relative FBM of 6% (range 1–10), quartile 2 with 13% (11–18), quartile 3 with 23% (19–33), and quartile 4 with 38% (34–71). All quartiles correlated with a decrease in CVfg as measured pretransplantation (P = 0.02 for quartile 1; P = 0.005 for quartiles 2 and 3; P = 0.003 for quartile 4). Patients in quartiles 3 and 2 had undergone a stronger decrease in CVfg than, respectively, those in quartiles 2 and 1, while patients in quartiles 3 and 4 showed no statistical difference in the effects on CVfg (Fig. 2).

A strong correlation was measured between CVfg and episodes of fasting hypoglycemia <70 mg/dL (r = 0.78, P < 0.001).

An ROC analysis was performed in order to investigate whether the phase 2 HGC values are associated with a specific metabolic effect, in particular bringing the CVfg <25% or inducing a reduction by >50% in pretransplant CVfg (Supplementary Fig. 2). The latter two criteria were achieved in all patients with an implant FBM >37% of control subjects, while this occurred in none of the patients with an FBM <5% (Table 2). CVfg was consistently reduced <25% in all patients with FBM >18% of control subjects (positive predictive value 100%); this end point was never reached for an FBM <12% (negative predictive value 100%). As indicated by the area under the ROC curve (0.97 [95% CI 0.93–1.00]), PT FBM distinguishes each patient with CVfg <25% from one with CVfg >25% (Supplementary Fig. 2).

This cross-sectional study shows a correlation between the FBM of established islet cell implants and their ability to correct three parameters of metabolic control (HbA_1c, CVfg, and insulin dose), the strongest relationship being with glycemic variability. We selected glucose clamps to assess the secretory capacity of the implants and compare it with that in nondiabetic control subjects. Other β-cell stimulation tests exhibit a higher intraindividual variability (14–16).

Hyperglycemic glucose clamps have shown a wide range of residual β-cell function in type 1 diabetic patients (8,10,11). Their second secretory phase has indicated that intraportal islet cell implants exhibit a functional capacity that varies between 12 and 45% of that in age- and body weight–matched healthy control subjects (8). The current study confirms this variability in a larger group of recipients, including both insulin-dependent and -independent patients. It argues against using the basal phase or the first hyperglycemic phase as an index of FBM, as these appear significantly less sensitive for detecting differences from the control values, thus overestimating the relative potency of implants. The average secretory capacity during the second phase was significantly lower than in control subjects, with equal levels in the absence (18% of control subjects) and in the presence (16%) of glucagon. Both segments of the second phase can thus serve as an index of FBM. The higher-than-normal basal glycemia in some transplant patients may lead to an underestimation of the first-phase insulin release because of a presumed higher degree of degranulation of the insulin granules and to a lesser extent of the second phase of the hyperglycemic phase in these patients. Because of side effects after glucagon administration in some patients and no added value of the postglucagon response in this study, we conclude that in studies with intraportal human implants the glucagon segment can be omitted. It may, however, still be useful when porcine or human stem cell–derived cells will be tested in future trials.

In our protocol, we did not perform a euglycemic or isoglycemic-hyperinsulinemic clamp that allows fine assessment of glucose disposal and thus estimation of insulin resistance. It is assumed that in lean type 1 diabetic recipients, graft insulin secretion is more important than insulin resistance in determining overall glycemia. Moreover, an FBM with an IQR variability of 10–33% makes it difficult to study insulin action and insulin resistance with an iso- or euglycemic clamp technique. However, the HGC technique is an adequate method to determine graft insulin secretory capacity, which is the purpose of this study. In order to examine the effects of insulin secretion on variables like average glucose control, glycemic variability, and insulin needs, it is imperative to include subjects with an important variation in function.

The relative magnitude of the second phase was selected as an index of the FBM of the implants. It strongly correlates with the reduction in glycemic variability, which is clinically relevant since glycemic variability is widely recognized as the main limiting factor in obtaining and maintaining optimal HbA_1c with intensive insulin therapy in type 1 diabetic patients (17).

The current study was conducted in patients who were C-peptide negative before transplantation, with a history of hypoglycemia and important fasting glycemic variability, defined as CVfg >25%. Even on insulin analogs or insulin pump, most type 1 diabetic patients have CVfg >25% (18–22). They are at higher risk for hypoglycemia (23) and therefore frequently kept at higher HbA_1c targets with its associated risks for chronic complications. There is no universally accepted gold standard method for quantifying glycemic variability (24). We used daily home blood glucose monitoring because it has the advantage of providing real-life data on a continuous basis with the lowest risk of study-related changes in patient behavior. With these data, glycemic variability can be easily estimated by calculating the SD or, better, the CV (8,17,21,25). In future studies, continuous glucose monitoring may become an even better tool to study glycemic variability, although only 64% of a selected group of adult patients and 19% of adolescents are willing to use this new methodology on a permanent basis beyond month 6 (26).

FBM of established implants correlated with their ability to reduce CVfg but not with decreased mean fasting glucose concentrations, indicating that all patients received appropriate insulin doses overnight. In pathophysiological terms, FBM determines and thus predicts glycemic variability. This was indeed shown in the ROC analysis demonstrating a strong predictive value of FBM on reduction of glycemic variability. An FBM of at least 18% is needed to achieve a CVfg <25%. A value of at least 37% brings CVfg <25% plus reduces it by >50% of pretransplantation values in all patients. The latter criterion is currently used in our center to decide whether an additional implant is to be performed. It is only met in a small number of patients, which underlines the need for further improvements in the preparation of grafts and the transplant protocols.

It is not to be excluded that the exposure of B-cells to calcineurin inhibitors leads to impaired insulin secretion (27,28) while lower glomerular filtration rate in transplanted patients leads to lower C-peptide clearance and thus potentially falsely elevated basal C-peptide levels. Future studies in recipients of intraportal islet grafts under other immune-suppression or immune-suppressive doses and in de novo type 1 diabetic patients—with normal estimated glomerular filtration rate and residual endogenous pancreatic function—should determine if these observations are applicable to these populations. Likewise, we agree that the results here obtained with intraportal implants cannot be extrapolated to extrahepatic implants. However, today no clinical functional extrahepatic implants have been published in man.

Each transplant study should also determine what minimal β-cell transplant function is needed so that benefits of treatment overrule (risk of) adverse events. In our study, no clinical benefit in terms of glycemic variability was measured in patients with an FBM <5%. This finding helps us with the decision to stop tacrolimus and mycophenolate mofetil as maintenance immune suppression. As both drugs are associated with acute and long-term risks, continuation is discussed on an individual basis with patients who have a FBM <5%.

In conclusion, the FBM of islet cell implants as measured by a HGC test predicts their effect on glycemic variability in C-peptide–negative type 1 diabetic recipients of an intraportal graft. Cutoff values were identified above and below which clinical benefit was present or absent, respectively. Further studies can indicate to which extent these values can help guide and monitor decisions on conducting additional islet cell transplants or on adjusting or stopping immune therapy. The glucose clamp data may also serve to compare outcome of different islet cell transplant protocols and to assess β-cell replacement therapies using novel cell sources.

This study was supported by the JDRF (grant 4/2005/1327) and the W. Gepts fund of University Hospital Brussels. P.G. is funded by a clinical research foundation of the University Hospitals Leuven, Katholieke Universiteit Leuven. I.W., C.M., and B.K. are Senior Clinical Investigators of the Research Foundation Flanders (Fonds Wetenschappelijk Onderzoek-Vlaanderen).

No potential conflicts of interest relevant to this article were reported.

P.G. wrote the manuscript and researched data. R.H. researched data and reviewed the manuscript. U.V.d.V. researched data. Z.L. reviewed the manuscript. D.H.L. researched data. I.W., F.G., and C.D.B. reviewed the manuscript. L.K. advised in statistical analysis. C.M. reviewed the manuscript. D.P. contributed to the discussion and edited the manuscript. B.K. wrote the manuscript and researched data. P.G. and B.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Parts of this study were presented in abstract form at the 2nd Joint Artificial Insulin Delivery Pancreas and Islet Transplantation (AIDPIT)/European Pancreas and Islet Association Winter Symposium and 31st AIDPIT Workshop, Igls, Austria, 30 January 2012.

The authors thank Monique Robyn, Hilde Morobe (University Hospitals Leuven), Koen Verbeeck, Sofie Vandenhoeck, Soniya Thomas, Jenny Van Den Brande (University Hospitals Brussel), and Rie Braspenning (University Hospital Antwerp) for their work in completing all the clamp tests. The authors also thank the Eurotransplant Foundation and its transplant surgeons and coordinators for organ procurement. Finally, the authors thank the staff of the Beta-Cell Bank, the Belgian Diabetes Registry, the Diabetes Research Center of the Vrije Universiteit Brussel, and the Clinical Biology Department of University Hospital Brussels.