Post-Liver Transplantation Diabetes Mellitus: A Review of Relevance and Approach to Treatment

This article is based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors.

Risk factors for PLTDM can be classified into two groups: those associated with the development of DM in the general population and those specifically associated with increased DM risk among LT recipients (Fig. 1).

In the first group, “classical” risk factors for PLTDM with robust evidence support include older age [16‐25], male sex [17, 19, 24, 25], high body-mass index (BMI) [19, 20, 24, 25], pre-transplant impaired fasting glucose [26‐28], family history of DM [26] and African American or Hispanic ethnicity [20, 29].

Conditions that predispose particularly to development of DM after a LV include hepatitis C virus (HCV) [15‐20, 24, 26, 28, 30‐33] or cytomegalovirus [22, 25] infection and immunosuppressive therapy with high-dose corticosteroids [14, 15, 20, 24, 25, 34] or calcineurin inhibitors (CNIs; tacrolimus or cyclosporine) [16, 20, 21, 23, 24, 26, 35, 36]. Nonalcoholic steatohepatitis (NASH) is a strong risk factor for the development of type 2 diabetes (T2DM) in the general population [37], and there is no reason to believe that this association would be any different in patients who have received a LT.

Some less well-established determinants of PLTDM in the recipient are statin therapy [27], central body fat distribution prior to transplantation [32], low magnesium levels before and 1 month after surgery [22], hyperglycemia in the first post-transplant month [25, 36] and stay in the intensive care unit (ICU) > 15 days [25].

Donor characteristics also play a key role in predisposing or protecting from PLTDM. Factors associated with increased risk are age > 60 years [19, 20], male gender [25], computed tomography scan- or biopsy-diagnosed liver steatosis [14, 35] and deceased liver donor [15, 25]. In a study of Japanese recipients of living donor liver transplants, cholinesterase plasma levels of < 185 IU/L (as a measure of liver function in the donor) were an independent risk factor [21].

Among the variables related to the transplant procedure itself, a cold ischemia time of > 9 h [25] is detrimental. The use of induction therapy with agents other than corticosteroids as part of the immunosuppressive regime has been found to be a protective factor against PLTDM in several studies, two of which used basiliximab, a monoclonal antibody directed to the interleukin-2 receptor [20, 24]. Acute graft rejection also predisposes to PLTDM [19, 22, 24, 38], but a causal link is hard to establish given that acute rejections are usually treated with high doses of corticosteroids, which induce hyperglycemia [30].

Calcineurin inhibitors are a family of highly effective immunosuppressive drugs that have revolutionized transplantation medicine over the last 40 years. Both cyclosporine and tacrolimus were developed by multidisciplinary research teams working at pharmaceutical companies and searching for immunosuppressants with a mild profile of cytotoxic adverse effects [39]. Cyclosporine is a hydrophobic cyclic undecapeptide with N-methylated amino acids that render it resistant to digestion by gastrointestinal system proteases [40, 41]. Tacrolimus is a macrolide antibiotic with a slightly better water solubility than cyclosporine. After intestinal absorption and entry into cells, both of these CNIs bind a cytoplasmic protein belonging to the immunophilin family: cyclophilins in the case of cyclosporine and FK-binding protein (FKBP) in the case of tacrolimus [41].

The cyclosporine–cyclophilin or tacrolimus–FKBP complex inhibits calcineurin, a calcium-dependent phosphatase involved in T-cell activation and regulation via dephosphorylation of nuclear factor of activated T-lymphocytes (NFAT) [42, 43]. However, calcineurin and NFAT are relevant not only in immune cells but also in many other tissues, including kidney, heart, spleen, liver, testes, brain and pancreas [44, 45]. In pancreatic beta-cells, calcineurin promotes the transcription of survival factors and stimulates the growth and expansion of the beta-cell mass [46]. Calcineurin is also involved in metabolic signaling pathways in adipose [47] and skeletal muscle tissue [48]. Widespread inhibition of calcineurin by CNIs during immunosuppressive therapy [45] may therefore interfere with its action in all these tissues and lead to potential metabolic side effects (Fig. 2).

The mechanistic factors by which CNIs contribute to the development of PLTDM are diverse and include deregulation of insulin secretion, apoptosis of insulin-producing beta cells and induction of peripheral insulin resistance. Studies in cultured beta-cells have reported a reduction of basal and glucose-stimulated insulin secretion after exposure to cyclosporine, an effect less consistently found with tacrolimus [49, 50]. In addition to the inhibition of calcineurin [51], other targets have been implicated in CNI-induced insulin secretory dysfunction. Of note, tacrolimus treatment has been observed to reduce the number and oxygen consumption of beta-cell mitochondria in vitro, limiting the availability of ATP and other metabolites essential for secretory pathways [52]. Similarly, studies in isolated murine islets revealed cyclosporine-mediated inhibition of the permeability transition pore, a mitochondrial protein important for the regulation of cytoplasmic calcium oscillations and therefore vesicle-mediated exocytosis [53]. Tacrolimus also affects the normal closure of ATP-sensitive potassium channels [54] and reduces glucokinase (but not hexokinase) activity [55].

Experiments with reporter genes in hamster beta cells found that CNIs suppress the transcription of genes whose promoters contain cAMP-responsive elements [56], many of which are essential to the survival, replication and function of beta-cells [57, 58]. In fact, the administration of tacrolimus to Sprague–Dawley rats negatively impacts their beta cell mass [59]. The impact of CNIs on peripheral insulin action does not seem to be as crucial as their impact on beta-cells, but calcineurin inhibition does have an effect on insulin sensitivity. When primary human adipocytes were incubated with high concentrations of a CNI (100 nM), the amount and phosphorylation of key intermediaries of the insulin signaling pathway (IRS1/2, p85-PI3 K, PKB, AS160, and mTORC1) did not change [60], but the membrane content of glucose transporter 4 (GLUT4) transporters and the uptake of C14-labeled glucose did diminish. This effect was apparently due to a slower vesicle-to-plasma membrane recycling of GLUT4. Finally, calcineurin inhibition seems to promote transformation of type I skeletal muscle fibers to less insulin-sensitive type II fibers [61] (Fig. 2).

Clinical data suggest that the impact of specific CNIs on glucose metabolism may not be the same. A meta-analysis of 16 studies involving 3813 patients found a greater degree of glucose impairment with tacrolimus than with cyclosporine [62]. A more recent meta-analysis estimated a PTDLM risk ratio of 0.60 (95% confidence interval [CI] 0.47–0.77) for cyclosporine relative to tacrolimus [63].

Corticosteroids continue to be part of the standard treatment for immunosuppression during the early post-transplant period, and they are known to promote hyperglycemia through a host of different mechanisms. The corticosteroid dexamethasone exhibits cytotoxic [64] and anti-proliferative [65] effects on beta-cell lines in vitro. Also, in vitro exposure to corticosteroids impairs insulin secretion, an effect mediated via the upregulation of serum-and-glucocorticoid-inducible kinase-1 and deterioration of membrane depolarization necessary for glucose-mediated vesicle exocytosis [66]. Corticosteroids interfere with the insulin signaling pathway, thereby reducing insulin-mediated glycogen synthesis [67], GLUT4 translocation and glucose uptake in muscle [68]. Lastly, corticosteroids increase hepatic glucose output, further contributing to fasting hyperglycemia [69] (Fig. 2).

Mammalian target of rapamycin (mTOR) inhibitors exert their immunosuppressive action by forming a complex with FKBP (the target of tacrolimus) and inactivating the protein mTOR. Inhibition of the mTOR signaling pathway leads to a decrease of cytokine-mediated T-cell activation and proliferation [70]. Only everolimus is approved by both Federal Drug Administration (FDA) and European Medicines Agency (EMA) for use in LT recipients. However, sirolimus is still used in certain LT patients [71]. The metabolic effects reported for these agents are mainly hypertriglyceridemia and hypercholesterolemia [72, 73]. The impact of sirolimus on glucose regulation seems to be less pronounced than that of CNIs. A study in 23 LT patients evaluated the impact of conversion to sirolimus after 4 weeks of CNI therapy. Three patients had developed PLTDM after the use of CNI and had insulin requirements between 80 and 130 IU per day. After conversion to sirolimus, daily insulin requirements dropped to 24–32 IU while blood glucose levels remained stable [74]. A review of data from 227 patients with hepatocarcinoma as an indication for liver transplantation showed a significantly higher incidence of PLTDM in patients treated with tacrolimus + mycophenolate (12.3%) versus those treated with sirolimus (0%; p < 0.001) [75].

Another hypothesis to explain the high coincidence of DM and LT is that liver diseases for which a transplant is required and DM have a common origin. HCV is both the most common cause of LT in the USA (affecting almost 50% of the recipients) and a risk factor for DM: a meta-analysis showed that HCV infection increased DM by a factor of 1.7 in both retrospective and prospective studies [76]. HCV infection is positively associated with the homeostasis model assessment-insulin resistance (HOMA-IR) index in humans [77], probably due to impairment of insulin signaling in hepatocytes [78, 79]. Mice transgenic for the HCV core gene exhibit overexpression of tumor necrosis factor-alpha (TNF-α), and TNF-α-dependent insulin resistance [80]. Similarly, alcoholic liver disease is the second-most common indication for LT [81] and a DM risk factor: a recent dose–response meta-analysis with over 1.9 million individuals concluded that alcohol intakes of > 120 g/day significantly increase the risk of DM [82].

NASH is another common cause of LT, and the percentage of LT secondary to this condition is estimated to have increased 35-fold between 2000 and 2005 in the USA [81]. NASH is strongly associated with metabolic disturbances known to be risk factors for DM, among which are overweight and abdominal obesity [83]. However, NASH in and of itself is a risk factor for DM and CVD [84], and many patients develop liver steatosis de novo after LT [85]. Thus, there is a bidirectional relationship between NASH and DM, also in the post-LT patient.

Not all cases of PLTDM persist over time; some resolve spontaneously. There is substantial disparity in the reported persistence of PLTDM across studies, probably due to the heterogeneity of diagnostic criteria and differences in follow-up length [10, 14, 86, 87]. In a study of 17,184 adult LT recipients followed for 5 years, 29.2% developed at least one episode compatible with PLTDM, but persistence for > 1 year was only 4.9% of the initially transplanted patients (7.6% for recipients with NASH) [88]. Characteristics associated with persistent PLTDM include African American race, HCV infection, NASH in the recipient, higher MELD score and acute cellular rejection [87, 89]. The authors of a long-term study of renal transplantation proposed a classification according to the temporal pattern of presentation, as follows: early-onset persistent DM (present in the first year of transplantation and during the next 7 years), late-onset DM (occurring after the first year) or transient DM (diagnosed within the first year but eventually recovering to normal glycemia) [90]. The implications of this temporal classification on the management, prognosis and survival of LT recipients have not yet been established [91].

Cardiovascular complications are a major cause of non-liver-related mortality after LT, representing 13–28% of all deaths [96‐98]. PLTDM has been identified as an independent predictor of post-transplantation CVD events. The CVD rates of patients after a LT were evaluated in a retrospective analysis of the Organ Procurement and Transplantation Network/UNOS database. A comparison of patients with persistent PLTDM, transient PLTDM, pre-LT DM or LT without DM, respectively, revealed that those with persistent PLTDM exhibited the highest risk (HR 1.95 vs. non-DM; p < 0.01) [87]. DM at 1 year after LT is much more frequent among recipients who develop CVD events than among those who do not (64 vs. 38%; p < 0.001) [99]. Therefore, the occurrence of PLTDM in LT recipients puts this group in a special condition of cardiovascular vulnerability.

In a small single-center matched case–control study, PLTDM was associated with higher rates of acute rejection, but not with long-term graft failure [100]. Yet in a larger, multicenter prospective study, patients with PLTDM showed increased risk for graft failure [93]. PLTDM is also linked to a higher number of rejection episodes [101]. Data from a large U.S. cohort showed that pre-transplant DM, PLTDM and acute rejection were significantly associated with increased risks for graft failure. However, after multivariate Cox regression adjustment, the association between PLTDM and adverse outcomes did not persist [102]. An important consideration is that the impact of PLTDM on survival of the transplanted organ may be underestimated: patients with PLTDM may develop rejection, graft failure and subsequent death. In these patients, however, the only recorded outcome is mortality. Due to these complexities, it is not possible based on current evidence to ascertain whether or not PLTDM has an independent effect on the risk of graft rejection or failure.

In addition to the above-mentioned adverse outcomes, PLTDM patients also have a significantly higher incidence of renal insufficiency and post-operative bacterial infection [10]. In a study of adults who required LT due to HCV infection in the USA, patients with PLTDM developed more HCV recurrences (59%) and stage 2 fibrosis than both patients with pre-LT DM and normal patients [103]. A similar study in the UK found PLTDM to be a strong independent risk factor for the development of a fibrosis score of ≥ 4 over a 6-year follow-up (HR 3.28; p = 0.004) [104]. In summary, infections and infection-related complications [26] seem to be higher in patients with PLTDM.

A retrospective review of 184 LT recipients showed that strict intraoperative glycemic control (< 150 mg/dL; actual mean glucose 135 mg/dL) results in a remarkably lower 1-year mortality compared with less strict control (≥ 150 mg/dL; actual mean glucose 184 mg/dL) (8.8 vs 21.9%; p = 0.05) [108]. In the same report, patients in the less strict group experienced a cumulative infection rate of 48%, compared to 30% to those in the tighter control group (p = 0.02) [108]. In a separate study in a tertiary care transplantation center, the attainment of a perioperative blood glucose level that was lower by 31 mg/dL was accompanied by a significant reduction in infection rates in LT patients (adjusted odds ratio [OR] 0.22, 95% CI 0.06–0.86) [109]. In a large case series, severe intraoperative hyperglycemia (≥ 200 mg/dL) during liver transplantation was an independent risk factor for postoperative surgical site infection (OR 2.25; 95% CI 1.26–4.03; p = 0.006) [110].

Poor glycemic control has also been related with longer stays the ICU on the first admission (5.6 vs. 3.2 days; p = 0.039) [109] and prolonged ventilation (OR 4.3, 95% CI 1.28–14.4) [104]. Lastly, a retrospective review of 144 liver and liver–kidney transplants found lower rejection rates in patients with in-hospital average glucose levels of < 200 mg/dL (35.1 vs 76.7%, p < 0.001) [111].

Several factors increase whole-body insulin requirements during the early post-transplant period, among them immunosuppression with steroids, acute pain and surgical stress [112]. Consequently, intravenous or intensive (three or more injections a day) insulin therapy is the standard of care during the immediate post-transplant phase. The safety of bedside glucose-based, sliding-scale intravenous insulin schemes in hospitalized LT recipients has been documented [113]. Nonetheless, it should be stressed that such schemes demand careful and frequent glucose monitoring by nursing personnel. Once patients have returned to a regular eating pattern, they can be transitioned into a subcutaneous basal/bolus regimen. For patients not receiving total parenteral nutrition, an initial total daily dose of between 0.2 and 0.4 U/Kg is reasonable, of which 50% should be administered as basal insulin and 50% as prandial insulin. Prandial (fast or ultra-fast acting insulin) is used at each meal, usually in a fixed ratio of dietary carbohydrate to insulin. A good initial ballpark is 1–2 U per each 15 g carbohydrate. Supplemental doses of fast-acting insulin should be administered when blood glucose measurements are outside treatment goals [114]. A practical initial approach is to measure blood glucose every 4–6 h and to administer 1–2 U of fast-acting insulin for every 40–50 mg/dL that the blood glucose is above 140–150 mg/dL.

While the transplanted patient is still hospitalized, capillary glucose goals are the same as those for other hospitalized patients: 140–180 mg/dL in the ICU, and < 140 outside the ICU for pre-meal and < 180 mg/dL for 2-h post-meal or random measurements [114]. For the PLTDM outpatient, capillary glucose goals are 70–110 mg/dL for pre-meal measurements and 70–140 mg/dL for post-meal measurements. Ambulatory HbA1c goals can be defined according to diabetes duration, presence of comorbidities and risk of side effects associated with antidiabetic therapy (i.e. hypoglycemia). Thus, an HbA1c of < 7% is a suitable goal for most PLTDM patients. For patients with a short disease duration, younger age and few comorbidities aside from their liver problems, an HbA1c of < 6.5% is both feasible and desirable. For patients with advanced age, multiple comorbidities, high risk of hypoglycemia and/or limited life expectancy, an HbA1c of < 8.0% is a more realistic and potentially less iatrogenic objective. Despite a paucity of long-term studies evaluating the impact of glycemic control in patients with PLTDM, a study in LT recipients with HCV infection found that patients with tight glycemic control (< 138 mg/dL) had a lower rate of Stage 2 fibrosis development relative to patients with a glycemic average above 138 mg/dL (78 vs. 60%; p = 0.027) [103].

Other relevant treatment goals include CVD risk factors other than glycemia, as CVD continues to be a major cause of mortality among LT patients in general and PLTDM patients in particular [115]. Smoking should be strongly discouraged, and support for smoking cessation should be provided when needed. The level of low-density lipoprotein (LDL) cholesterol should be kept at < 100 mg/dL in PLTDM patients without prior clinical CVD and at < 70 mg/dL in those with clinical CVD. For patients with progressive CVD despite receiving cholesterol-lowering therapy, an LDL cholesterol level of < 55 mg/dL is recommended by some guidelines [116]. The goal for plasma triglycerides is < 150 mg/dL. However, the management of plasma triglycerides may be particularly difficult in patients who receive the mTOR inhibitors sirolimus or everolimus, as these agents are clearly associated with marked hypertriglyceridemia [117]. Since most PLTDM patients will be receiving a statin for LDL cholesterol management, it must be borne in mind that fenofibrate is the fibrate of choice for combined therapy with statins. Blood pressure goals can be set according to the most recent American Heart Association guidelines [118] to < 130/80 mmHg. Patients with PLTDM, like all other patients with diabetes, must undergo periodic evaluations of eye and foot care, as they are not only at increased risk of the usual complications of DM, but also are particularly prone to cataracts (due to corticosteroid use) and soft tissue infections (due to prolonged immunosuppression).

The 2014 international consensus guidelines on post-transplantation DM recommend a stepwise approach for the management of late post-transplantation DM that consists of lifestyle modification followed by oral anti-diabetic therapy and then insulin therapy [11]. Different factors influence the decision of which oral anti-diabetic drug should be the first choice, and every patient should be evaluated individually.

Maintenance of caloric balance and control of body weight are essential components of diabetes management. In patients with kidney transplants, weight gain during the months following transplantation is proportional to the risk of new-onset diabetes, independent of the pre-transplant BMI [119]. Also in kidney transplant recipients, a 6-month intensive lifestyle modification program that included referral to a dietitian, a structured exercise program and weight loss advice induced regression to normoglycemia in up to 44% of IGT patients [120]. Randomized trials are needed in which the overall impact of structured diet and exercise can be specifically assessed in the PLTDM population.

Despite the advent of safer and more effective immunosuppressive agents, steroids remain the most commonly used therapy for the induction and treatment of rejections, based on their efficacy and low cost. However, steroids have a well-recognized diabetogenic effect [161]. The tapering or withdrawal of steroids in order to diminish their metabolic effects is a controversial issue. In one clinical trial, 502 LT recipients with HBV infection were randomized to one of four steroid-minimization protocols (steroid withdrawal after 6 months, 3 months or 14 days, or complete avoidance of steroid therapy). After 3 years of follow-up, hyperglycemia and diabetes were significantly more frequent in patients on the two longer protocols, without any difference in patient survival, graft survival or chronic rejection among the four groups [162]. Another randomized trial of 82 adult LT patients found that “almost avoidance” of steroids (withdrawal within 24 h of receiving the transplant) was associated with much lower insulin requirements 1 week post-transplant compared to a more conventional regime (20.5 vs. 39.6 Units; p < 0.05) [163]. Other studies suggest that patients with older age or HCV as a cause for LT would specially benefit from an early steroid withdrawal protocol [164, 165]. Further, a retrospective analysis of 330 patients found no difference in PLTDM, hyperlipidemia or cardiovascular events at 4 years after the LT between patients with steroid withdrawal by 3 months or those with steroid withdrawal by 3–12 months post-procedure [166].

Of note, a recent meta-analysis of 16 clinical trials reported that steroid avoidance or withdrawal was accompanied by increased acute rejection (relative risk [RR] 1.33, 95% CI 1.08–1.64) and steroid-resistant rejection (RR 2.14, 95% CI 1.13–4.02), while achieving only a modest reduction in the risk of PLTDM (RR 0.81, 95% CI 0.66–0.99). These results clearly illustrate that glycemic control should not jeopardize patient and graft survival [167]. That does not mean, however, that steroids should be used liberally: one study assessed the impact of reducing the methylprednisolone dose from 10–15 mg per day to 5 mg per day during the first year in renal transplant recipients and found positive effects on insulin sensitivity [168]. Unfortunately, the optimal steroid dose needed to achieve a balance between prevention of LT rejection episodes and avoidance of glycemic impairment is still unknown.

Concerning CNIs, tapering of tacrolimus may have a positive effect on glucose tolerance [169, 170], but there are no trials directly aimed at assessing the influence of different CNI dosing regimens on glycemic control in PLTDM. A few trials have assessed metabolic outcomes after the change in treatment regimen from tacrolimus to cyclosporine, reporting notable benefits [171, 172]. Whether persistence of PLTDM is a sufficient indication for a change of immunosuppressant is still not well established.