Drug-induced hyperglycemia is a clinical condition that can occur as a result of impaired insulin secretion or action or the destruction of pancreatic beta cells [1]. The administration of glucocorticoids (GC) is a common cause [2]. Gulliford et al. found 2% of newly diagnosed cases of diabetes mellitus (DM) due to orally administered GC [3]. Its use is associated with multiple side effects with DM being one of the most important [4, 5]. In cases of DM prior to the use of GC, worsening of glycemic control is typical; however, in patients without DM, glucocorticoids may transiently or permanently induce hyperglycemia [5–7].

Two mechanisms are predominantly responsible for hyperglycemia secondary to GC: a decrease in insulin secretion and insulin sensitivity [5, 7, 8]. Although the link between GC and hyperglycemia was identified more than 50 years ago, clinically relevant questions remain unanswered [2, 5, 7, 9]. The incidence of DM secondary to the use of GC, the profile of subjects in whom it occurs, the consequences of comorbidities, the moment it becomes apparent, its intensity, and its evolution in the short and medium term are not well defined. An incidence of 2% to 50% has been reported [3, 10]. Published studies have used methodologies accountable for this broad range. Most are retrospective with a very large GC dose range, different DM diagnostic criteria, comorbid conditions that affect insulin sensitivity, and with no description of the course of hyperglycemia [3, 5, 9–16]. It is important to find answers to clinical scenarios in which GC are used. A common setting is the administration of high dose (>1 mg/kg/day) GC for one to three months in a patient without comorbidities and in whom DM can occur. Obtaining this information would allow appropriate diagnostic and therapeutic strategies [5, 9, 17].

With the aim of defining the incidence of secondary DM, we designed a prospective, longitudinal, observational study in patients with a recent diagnosis of hematological diseases without DM and in whom treatment with GC would be initiated. Secondary end points were to identify the intensity of hyperglycemia, the moment it develops, its progression, and the risk factors, possible mechanisms responsible for the occurrence of hyperglycemia, and the impact of a continuous glucocorticoid regimen (CGC) vs. a cyclic glucocorticoid regimen (CyGC).

In a noncritically ill adult population, using a high dose of prednisone (>1 mg/kg/day) for 6 to 12 weeks, in two classical schemes of GC administration, under close monitoring of fasting and postprandial glycemia, we identified a high incidence of prediabetes and DM. At eight weeks, by fasting and postprandial glycemia, a DM incidence of 40.6% was found. For prediabetes, the incidence was 34.3%. The incidence of DM did not increase at 12 weeks. Unlike other series, in this study fasting glycemia was more sensitive in identifying DM than postprandial determinations [5, 11, 12, 20]. Only two cases that had DM by postprandial hyperglycemia did not meet fasting plasma glucose DM criteria. Series of patients on high doses of GC have found a DM incidence in the range of our study, between 12.6% and 52% [11, 13–16]. Despite methodological problems such as: different populations with heterogeneous comorbidities, diverse diagnostic methods for DM, a lack of close monitoring of glycemia and varying patterns of high doses of GC, the incidence of DM in most of these studies was between 15% and 25%, less than the incidence of DM of 40.6% that we found in our study [11, 13–16]. This difference can be attributed mainly to the retrospective design and lack of close systematic monitoring of fasting and postprandial glycemia. Contrary to our study, the study with the highest incidence (52%) is a retrospective series in which all cases of DM were detected solely by postprandial glycemia [11].

When starting high-dose GC in a subject without DM, there are two interesting clinical issues for making appropriate decisions: the intensity that hyperglycemia can reach and when it may occur. With regard to the intensity, acute complications of DM have been reported in a few critically ill cases [14, 21–24]. In this study, most of the time hyperglycemia was not intense, deciding, under close supervision, not to start management of hyperglycemia, which allowed recording its evolution. The average fasting glycemia of patients with DM was 8.89 mmol/L and only four of 32 patients had a fasting glycemia greater than 8.33 mmol/L. Besides our study, no other has prospectively and closely evaluated the intensity of hyperglycemia. The second issue is the time it takes for hyperglycemia to occur. Of the 13 cases of DM, none occurred in the first or after the eighth week, and half appeared between the second and fourth week. Therefore, it is advisable to search for it during the second, fourth and sixth week of exposure to high doses of GC. This finding is consistent with the results of other authors that have also identified most cases of DM in early stages of exposure to GC [12, 14, 15]. Recently, a clinical practice guideline on the management of hyperglycemia in hospitalized patients, suggested that monitoring of glycemia in individuals without diabetes exposed to GC can be interrupted if glycemia is less than 7.78 mmol/L for a period of at least 24 to 48 hours [25]. Our results in noncritically ill patients show that hyperglycemia occurs after the first week or even until the eighth week of exposure to GC, coinciding with findings in other retrospective series [12, 14, 15]. It is likely that other comorbidities that are present in hospitalized patients alter the action or secretion of insulin and anticipate the onset of hyperglycemia.

Another question that has not been carefully evaluated in previous studies is the course of hyperglycemia. In this study, two-thirds of subjects with prediabetes or DM normalized spontaneously by the eighth week. The rest, at 12 weeks, showed normal glycemia or were in the range of prediabetes. This spontaneous remission in the majority of the participants could be attributed to the absence of other comorbidities, which might impair insulin secretion or action. Many cases of severe hyperglycemia related to GC use, occurred in subjects with an aggregated acute disease [14, 23, 24].

Some authors have proposed that in individuals with a prediabetes state, GC could precipitate the onset of DM [2, 5, 6]. In our study, mean basal serum insulin and HOMA-IR were not different between participants with CGC or CyGC or when comparing subjects with and without DM. These findings suggest that at the beginning there was no difference in insulin resistance between the groups. However, in participants on CGC and CyGC who developed DM there was a significant increase in insulin secretion compared with those who did not. This finding suggests that the primary mechanism involved in the onset of DM is most likely, increased insulin resistance, associated to a limited compensatory ability of the pancreas.

In a noncritically ill population, without other comorbidities, the use of high doses of GC for two to three months, as a CGC or CyGC dose, caused a high incidence of DM with non-severe hyperglycemia, most often with CGC, which appeared after the second week of GC treatment and resolved spontaneously without treatment in the next six weeks. Even though these results reflect glucose behavior in two specific hematological disorders, the clinical and therapeutic characteristics of our participants, who were otherwise healthy, could represent the clinical setting of many patients of other medical areas that might require high doses of GC for six to twelve weeks. It would be useful to perform prospective studies in other clinical situations that use GC, such as patients with multiple comorbidities, long-term low dose GC, or in critically ill patients.