Over the past two years, two anti-PD-1 antibodies [nivolumab (Opdivo, Bristol-Myers Squibb), pembrolizumab (Keytruda, Merck)] and one anti-PD-L1 antibody atezolizumab (Tecentriq, Genentech)] have been approved to treat patients with advanced non-small cell lung cancer (NSCLC). These drugs target receptors which provide inhibitory signals to T-cells, thus amplifying T-cell activity in an attempt to generate anti-tumor immune response. As expected, the side effects associated with ICI are often autoimmune, including pneumonitis, colitis and endocrinopathies (e.g. thyroiditis, hypophysitis). Autoimmune or type 1 diabetes mellitus (T1DM) has rarely been reported as a side effect of anti-PD1 therapy, primarily in case reports [2,3,4,5,6,7,8,9,10]. T1DM is caused by destruction of pancreatic beta cells by autoreactive T-cells. Non-obese diabetic (NOD) mice have been studied extensively as an experimental model since they feature many aspects of T1DM similar to humans. PD-1 interaction with its ligands PD-L1 and PD-L2 is crucial for regulation of CD4/CD8 auto-reactive T cells. In transgenic mice, PD-1 expression was associated with resistance to the precipitation of autoimmune diabetes; however, blockade of the PD-1/PD-L1 axis caused diabetes in the already pre-diabetic NOD mice [11,12,13].
Nivolumab was approved by the FDA in 2015 for the second line treatment of advanced NSCLC based on the results of two large randomized open-label phase III trials which demonstrated a survival benefit for patients treated with nivolumab versus docetaxel [14, 15]. The first of these to be published (Checkmate 017) included 135 patients with squamous cell lung carcinoma randomized to treatment with nivolumab or docetaxel. Treatment related adverse events included pneumonitis (5%) and hypothyroidism (4%) . A subsequently published trial (Checkmate 057) included 292 patients with advanced NSCLC randomized to treatment with nivolumab or docetaxel. Treatment related adverse events in this trial included hypothyroidism (7%) and pneumonitis (1%) . No episodes of treated-related hyperglycemia or DKA were noted in either of these studies.
Our patient’s autoimmune diabetes was abrupt in onset and very difficult to control due to the complete absence of insulin secretion as evidenced by undetectable C-peptide levels at the time of diagnosis. Blood glucose levels were normal prior to the initiation the treatment with nivolumab. Two weeks after the first dose of nivolumab, a random glucose was elevated at 193 mg/dL; two weeks after the second dose, the patient presented with DKA, a blood sugar of 739 mg/dL. There was no ‘honeymoon period” of preserved beta cell function and good glycemic control often observed at the onset of juvenile Type 1 DM. At the time of diagnosis of autoimmune diabetes, the HbA1C value was 7.1% correlating with a three month blood sugar average of 154 mg/dL, further suggesting a rapid onset of acute hyperglycemia after treatment with nivolumab. The presence of GAD 65, IA-2, and ZnT8 antibodies prior to treatment may have predisposed the patient to the development of autoimmune diabetes. After treatment nivolumab and at onset of diabetes (after treatment with exogenous insulin), IAA seroconverted from negative to positive, perhaps demonstrating the enhanced immune activation against islet antigens. ZnT8 Ab measured 13 months after onset of diabetes fell below the threshold of positivity in agreement with the findings of Vaziri-Sani et al of declining ZnT8 Ab titers rapidly after onset among a longitudinal cohort of young type 1 diabetes patients .
Some HLA-I and HLA-II haplotypes are associated with increased susceptibility to T1DM.
Class II haplotypes HLA-DR3-DQ2 (DR3) and HLA-DR4-DQ8 (DR4) are associated with increased risk of T1DM especially in Caucasians . HLA typing of our patient revealed HLA-I A30 and HLA-II DR9 (Table 1); neither is a high risk haplotype associated with the development of T1DM. In contrast to our findings, several case reports have shown an established high risk allele for T1DM (HLA-II DR4 haplotype) present in the majority of patients for whom HLA typing was available [5, 7, 8, 10] while other case reports did not observe an association between HLA and the development of ICI associated T1DM [4, 6, 9].
In contrast to other autoimmune sequelae of anti-PD1 therapy such as pneumonitis or colitis that are routinely treated with high dose steroids per clinical trial protocol specifications and published guidelines, there is little data regarding the use of steroids in the setting of autoimmune diabetes. High dose steroids are usually avoided since they exacerbate hyperglycemia and complicate the management of insulin dependent diabetes. Standard irAE immunosuppression with prednisolone in an attempt to reverse pembrolizumab-induced T1DM did not salvage beta cell function in one case reported by Aleksova et al. . Current management includes traditional treatment strategies for DKA, including intravenous (IV) insulin therapy, IV hydration and frequent monitoring of labs to correct the anion gap and electrolyte derangements. Co-management and communication between medical oncology and endocrinology is essential to coordinate education and close outpatient follow-up to optimize patient outcomes. Glycemic targets must be individualized to the patient, taking into consideration numerous other factors such as patient’s nutritional status, health literacy and overall prognosis. Development of DKA is not an absolute contraindication to continuing nivolumab therapy in patients with advanced NSCLC once the DM is well-controlled, given the paucity of treatment options for these patients. In keeping with a shared decision making model, careful discussion of risks and benefits between the patient and clinician is required. Of note, our patient has experienced a significant ongoing response to nivolumab and has not received antineoplastic treatment since December 2015.
Since it is not standard practice to check for the presence of diabetes autoantibodies prior to ICI treatment, the pre-treatment diabetes antibody status was typically unavailable for the previously reported cases. However, Lowe et al reported a case of autoimmune diabetes following combination therapy with ipilimumab and nivolumab that demonstrated anti-GAD antibody seroconversion from negative to positive . It is not known whether diabetes antibodies were present prior to ICI treatment in the other reported cases to date. To our knowledge, this is the first reported case to describe a patient who was found to have GAD 65, IA-2 and ZnT8 antibodies prior to treatment with nivolumab who then subsequently developed new insulin antibodies and autoimmune diabetes following treatment. It is likely that our patient was in the “preclinical” phase of the development of autoimmune diabetes characterized by the presence of one or more diabetes related autoantibodies and sufficient beta cell function to maintain euglycemia. PD-1 inhibition may have simply accelerated a pre-existing autoimmune process that ultimately led to the development of T1DM in this patient. Indeed, the presence of both GAD and IA-2 antibodies in first degree relatives of patients with T1DM has been shown to confer a 61% risk of developing TD1M in 10 years . Although our patient did not have any first degree relatives with diabetes, she almost certainly was at increased risk of developing T1DM prior to treatment with nivolumab due to the presence of these antibodies.
Reported cases of T1DM related to anti PD-1 have shown conflicting results regarding the presence of diabetes related autoantibodies after development of T1DM. Hughes et al. reported five cases of new-onset insulin-dependent diabetes after receiving anti-PD1 therapy. Three of the five patients described developed at least one positive diabetes related autoantibody after onset of T1DM . At least half of the reported cases of ICI related autoimmune diabetes show no detectable diabetes autoantibodies at onset [2, 4, 5, 8,9,10]. Similar results have been found in the NOD mouse model of autoimmune diabetes. Ansari and colleagues found no correlation between insulin autoantibody levels and development of autoimmune diabetes in NOD mice treated with PD-1–PD-L1 blockade; some mice developed diabetes with no antibodies and others developed antibodies but did not develop diabetes . Similarly, HLA alleles associated with increased risk of T1DM have been present in some but not all reported cases of ICI related T1DM.
As the approved indications for anti-PD1 therapy expand in the coming years, the number of patients receiving these drugs will dramatically increase, as will the frequency of autoimmune adverse events. Medical oncologists must be aware of the possibility of anti-PD1 therapy induced autoimmune diabetes and counsel their patients to report symptoms that may be related to DKA, as this is a medical emergency. The presence of diabetes autoantibodies prior to anti-PD1 therapy likely predisposed our patient to developing T1DM. It is possible that the presence of diabetes antibodies is a risk factor for ICI related T1DM. However, the absence of diabetes autoantibodies after onset of T1DM in some cases of ICI related T1DM suggests other factors are involved. Future correlative studies in this small patient population are needed to delineate genetic and immunologic biomarkers to identify those at highest risk for autoimmune sequelae, such as autoimmune DM.