Quantifying and Characterizing the Presence of Insulin Overbasalization in a Family Medicine Practice Free

The Institutional Review Board of the Heritage Valley Health System in Pennsylvania reviewed and approved this project as exempt research and waived participant informed consent because the study involved de-identified existing patient health information and posed minimal risk to patients, as no intervention was performed. This project did not receive funding support nor did study participants receive compensation.

The study population included all patients ≥18 years of age with a diagnosis of type 2 diabetes who were prescribed any basal insulin product by providers at the Heritage Valley Family Medicine Center (FMC) in Beaver Falls, PA, between 1 June 2018 and 30 June 2021. The FMC houses the Heritage Valley Family Medicine Residency Program and provides primary care services for a medically underserved, rural community located in Southwestern Pennsylvania. Patients were excluded if a prandial (i.e., short- or rapid-acting) insulin product was listed as an active medication, they were deceased at the time of data collection, or they were found to have a diagnosis of type 1 diabetes. Although insulin overbasalization can occur among patients using prandial insulin, previous research has focused on insulin overbasalization among patients on basal-only insulin regimens.

Observational and retrospective in nature, this study used data collected through chart review from the electronic health record (EHR) system. A report generated from the EHR identified potential charts for review using a search of orders for both brand and generic names of basal insulin products (e.g., glargine, NPH, detemir, and degludec).

Demographic and clinical data were collected in a de-identified manner using a standardized data collection form that included the following: age at the conclusion of the study period; sex; highest level of education completed; insurance coverage (uninsured, Medicare, Medicaid, dual Medicare/Medicaid, or commercial/private); weight prior to basal insulin initiation and five consecutive visits thereafter; baseline BMI; basal insulin product prescribed, starting insulin dose, and most recent dose; most recent A1C; use of other medication classes for the treatment of type 2 diabetes (metformin, sulfonylurea [SU], thiazolidinedione, dipeptidyl peptidase 4 [DPP-4] inhibitor, glucagon-like peptide 1 [GLP-1] receptor agonist, or sodium–glucose cotransporter 2 [SGLT2] inhibitor); use of home blood glucose monitoring (BGM) and results (if available); patient-reported hypoglycemia; previous hospitalization for diabetic ketoacidosis (DKA) or hyperglycemic hyperosmolar syndrome (HHS); polypharmacy (defined as use of >9 medications); order for endocrinology referral; type of prescriber who last adjusted basal insulin dose (postgraduate year 1 [PGY1] resident, postgraduate year 2 [PGY2] resident, postgraduate year 3 [PGY3] resident, attending physician, or pharmacist); and presence of a diagnosis of depression or anxiety (International Classification of Diseases, 10th revision, codes F30–F48), neuropathy, or diabetic kidney disease.

Data were entered into a Microsoft Excel spreadsheet and analyzed using descriptive statistics to quantify the frequency of overbasalization. Total daily insulin doses in units/kg (calculated from the most recent insulin dose and weight) were used to categorize patients into one of two groups:

• Basal insulin dose <0.5 units/kg/day: not overbasalized (NOB)

• Basal insulin dose ≥0.5 units/kg/day: overbasalized (OB)

Although, as previously noted, several factors have been proposed as potential indicators of the presence of overbasalization, basal insulin doses expressed in units/kg/day present a practical method for identification that can be widely adopted in a variety of practices. Although there is value to the inclusion of other possible indicators (e.g., BeAM differentials and fasting glucose levels), these data are not consistently or readily available in EHR systems and would have greatly limited the generalizability of the findings. The use of A1C as an indicator of overbasalization without additional data related to hypoglycemia could have resulted in an undercounting of overbasalization cases.

To characterize the potential variables related to overbasalization, associations between the overbasalization categories and the characteristics described above were analyzed using the Fisher exact test (Fisher Exact Test Calculator, Social Science Statistics) with an a priori α <0.05. Where relationships were identified, relative risk (RR) of overbasalization was calculated (Relative Risk Calculator, MedCalc Software Ltd.). Odds ratios (ORs) were calculated to assess the likelihood of achieving an A1C goal <7% for overbasalized patients compared with patients on lower daily doses (Odds Ratio Calculator, MedCalc Software Ltd.).

An additional category of “approaching overbasalization” was defined by a basal insulin dose >0.3 but <0.5 units/kg/day. In a post hoc analysis, patients approaching overbasalization who were not at the A1C goal of <7% were further analyzed. Differences between the most recent and goal A1C values were calculated and divided by the difference between their current insulin dose (units/kg/day) and the threshold for overbasalization of 0.5 units/kg/day to estimate the change in A1C required to obtain the A1C goal without exceeding 0.5 units/kg/day.

A total of 105 charts were included for review; patient characteristics are summarized in Table 1. Sixty-six participants (62.9%) were Medicare and/or Medicaid recipients. The most common basal insulin product used was U-100 glargine. None of the study participants were prescribed NPH insulin. Table 2 summarizes the products used by study participants. At insulin start, the mean BMI was 34.8 kg/m² (range 20.4–55.4 kg/m²), and the mean weight was 102.2 kg (range 60–159.5 kg). Patients were on basal insulin for an average of 59 months at the time of data collection (range 2–168 months). Average weights after initiation of insulin were relatively unchanged (−0.27 kg). The average most recent A1C was 8.8% (range 5–16.9%). Metformin was the most commonly used agent in addition to basal insulin (Table 3). Insulin dose adjustments were most frequently made by PGY3 residents (48 [46%]) followed by PGY2 residents (36 [34%]). Seventeen patients (16%) met the criteria for the OB category (basal insulin dose >0.5 units/kg/day), and 25 patients (24%) were approaching overbasalization (0.3 units/kg/day) (Figure 1).

The mean A1C was 9.4% in OB patients, 9.9% in patients approaching overbasalization, 8.8% in patients on <0.5 units/kg/day of insulin, and 8.4% in patients on <0.3 units/kg/day. Having an A1C >7% was associated with overbasalization (P = 0.0202). Although attaining an A1C goal of <7% was not observed in any patients with a basal insulin dose >0.5 units/kg/day, this result did not achieve statistical significance (OR 0.08, P = 0.0893, 95% CI 0.0049–1.4621). Similarly, among the five patients on doses >0.7 units/kg/day, although none met A1C goals, this association was not statistically significant (P = 0.5813). However, attaining an at-goal A1C level was significantly less likely with any basal insulin dose >0.3 units/kg/day (OR 0.18, P = 0.0089, 95% CI 0.0490–0.6482). A weak positive correlation between A1C and insulin dose (units/kg/day) was observed (R = 0.2221, P = 0.022779).

Among overbasalized patients, the median changes in weight between the time of basal insulin initiation and five consecutive visits thereafter were 0, +0.9, +0.6, +0.4, and +0.9 kg, respectively. Correlation between change in weight at visit five and insulin dose (units/kg/day) was not significant (R = 0.0969, P = 0.711405). Duration of insulin use was an average of 86 months (range 12–142 months) among overbasalized patients.

The lack of use of SUs was found to be associated with overbasalization; none of the 17 overbasalized patients (P = 0.0395) were prescribed an SU compared with 20.5% in the NOB group. Use of metformin (P = 0.7639), a GLP-1 receptor agonist (P = 0.5889), an SGLT2 inhibitor (P = 1), or a DPP-4 inhibitor (P = 1) was not related to the presence of overbasalization.

History of DKA was related to overbasalization (P = 0.0225); patients prescribed doses of basal insulin >0.5 units/kg/day were five times more likely to have been hospitalized for DKA (RR 5.18, P = 0.0121, 95% CI 1.4321–18.7108) compared with those who were not overbasalized.

Referral to endocrinology (P = 0.4584), sex (P = 0.1187), age ≥65 or <65 years (P = 0.7904), history of depression or anxiety disorders (P = 0.7929), history of hypoglycemia (P = 0.7066), BGM (P = 0.0622), history of HHS (P = 0.1219), neuropathy (P = 1), education level (P = 0.7729), and polypharmacy (P = 0.7913) were not related to the presence of overbasalization.

The post hoc analysis of patients not at an A1C goal <7% on doses of insulin between 0.3 and 0.5 units/kg/day identified 21 patients with a median A1C of 10.1% (range 7.3–16.5%) and a median dose of insulin of 0.36 units/kg/day. The median change in A1C expected from each 0.1 unit/kg/day increase in dose was 2.1%, with nine patients requiring a dose response in excess of 2.5%.

In this study, the prevalence of overbasalization (using the generally accepted definition of basal insulin dose >0.5 units/kg/day) was lower than the expected 40% reported previously in the literature (8). The lack of glycemic control and higher-than-average A1C found among overbasalized patients supports the current literature, which identifies overbasalization as a barrier to achieving glycemic control in type 2 diabetes. However, this study’s findings of an 82% lower incidence of glycemic control at lower thresholds (0.3–0.5 units/kg/day) for overbasalization add to the current body of literature.

In their large, retrospective analysis of real-world data to assess the probability of attaining A1C targets after initiation of basal insulin, Blonde et al. (9) observed that, although there were reductions in mean A1C (of ∼1.5 percentage points) during the first 6 months of treatment, only 21.5% of patients achieved A1C goals within this time frame. Further clinically significant reductions in A1C after 6 months were not observed. Although mean doses of insulin were not reported in this study, these data suggest that the majority of patients do not achieve glycemic goals despite the use of basal insulin.

In the DUNE (Diabetes Unmet Need with Basal Insulin Evaluation) study, Meneghini et al. (10) reported similar findings of A1C reductions of 0.8–1.4% at 12 weeks and only 27% of patients attaining individual A1C targets. Patients in this study were titrated to a mean basal insulin dose of 0.31 units/kg/day. The authors proposed titration inertia as a cause for the failure to attain A1C targets despite the initial robust response of A1C to insulin initiation.

Although not meeting A1C targets on 0.3–0.5 units/kg/day may be the result of clinical inertia with regard to inadequate titration of basal insulin doses in some patients, more than one-third of the patients in this subgroup of the current study would require dose responses greater than the 2.5% per 0.1 unit/kg/day that was observed in this insulin dose range by Umpierrez et al. (4). These data suggest that, although clinical inertia is likely in many cases, as suggested in the DUNE study (10), this inertia may not be related to lack of basal dose titration alone, but rather may also indicate a failure to intensify therapy (e.g., adding mealtime insulin or a GLP-1 receptor agonist). This possibility supports suggestions to begin considering a basal insulin–associated effect ceiling at this lower dose limit. Overbasalization may be preventable in a meaningful subset of patients with basal insulin doses between 0.3 and 0.5 units/kg/day, who could benefit from earlier implementation of strategies to target postprandial hyperglycemia and the anticipated clinical inertia that could arise despite continued basal dose titrations. For example, a patient taking 0.45 units/kg/day of basal insulin with an A1C >1.5% above target would likely become overbasalized in the attempt to reduce the A1C with basal dose increases alone.

The current study was similar to the study by Cowart et al. (8) in setting and inclusion criteria; however, there were several notable differences in practice and patient population. Our site is presumably smaller based on patient enrollment numbers and, although limited baseline patient characteristics were reported, it appears that the current study population was also older (mean age 61 vs. 57 years) and predominately male (60 vs. 48%). The higher incidence of OB (40%) could be partially explained by the previous study’s observation period from January 2015 to December 2018, which was before literature highlighting overbasalization was widely available. Although both studies included time frames before the first mention of overbasalization in the ADA Standards of Care, clinician awareness of overbasalization in our later study could have contributed to lower observed rates. Furthermore, although no difference was seen in our study regarding the use of GLP-1 receptor agonists, the previous study excluded patients with any GLP-1 receptor agonist use. This study suggests that rates of overbasalization may be variable across practices and supports the need for future research in this area.

Although other literature has indicated the risk of weight gain with insulin use, an average weight loss of 0.27 kg was observed in this study. Although overbasalized patients experienced an average weight gain, these changes in weight were modest (<1 kg) and occurred later in therapy compared with what has been previously reported in the literature (11). The authors hypothesized several possible explanations for these observations, including healthy lifestyle changes prompted by initiation of basal insulin or caloric wasting related to uncorrected glycosuria despite insulin use. These findings could also be affected by limitations resulting from the study methods if insulin initiation occurred before the start date captured in the EHR (e.g., patients who were already using insulin at the time of establishing care within the health system would have appeared in the EHR as a new insulin user). Although the study was likely underpowered for assessing the frequency of hypoglycemic events, the numerically higher rate of hypoglycemia among patients on doses of basal insulin >0.5 units/kg/day suggests possible safety concerns, as suggested previously in the literature. However, it is possible that hypoglycemia rates were underestimated because they were assessed by patient reports documented in provider notes. Rates could be higher with the use of continuous glucose monitoring (CGM) or a more systematic method of documenting hypoglycemia.

Although SU use is decreasing in favor of newer medications with cardiometabolic and cardiorenal benefits and more favorable adverse effect profiles, they remain important antihyperglycemic therapies where cost is an issue (1). The observation in this study that the use of SUs appeared to have a relationship with a lower incidence of overbasalization is an unanticipated finding and, in the context of relatively higher use of GLP-1 receptor agonists in the overbasalized group, supports the potential utility of SUs as adjunctive agents, a strategy suggested in previous literature on this topic (2). Alternatively, despite the use of GLP-1 receptor agonists in >40% of patients in the overbasalized group, higher GLP-1 receptor agonist utilization would likely provide the needed A1C reductions as a strategy for managing overbasalization. GLP-1 receptor agonist underutilization demonstrates an unmet need for therapy intensification among both those who were overbasalized and those who were not, in particular, to support weight management, given the elevated mean BMI in this patient population (1).

The authors suspected that patient variables such as increased age, lower levels of education, or the presence of depression or anxiety may have been factors that providers considered when deciding to titrate basal insulin in place of initiating prandial insulin, in an attempt to simplify the therapeutic regimen. Although there were no statistically significant findings on these patient variables related to overbasalization, the study might be underpowered to detect differences in these factors and suggests a need for further research on prescriber behaviors and decision-making related to insulin dose adjustments that might lead to overbasalization.

This study’s findings are not without limitations. Patients prescribed prandial insulin were excluded from this study; therefore, we were unable to assess the presence of overbasalization despite initiation of prandial insulin. Overbasalization could still be present with the use of prandial insulin, a potential area for further exploration. Given the retrospective nature of the study, potential missing or incorrect data from EHRs may not have accurately reflected the frequency of hypoglycemia, blood glucose values, actual insulin start dates, or patients’ doses of insulin actually being taken. For example, patients given verbal orders to increase or decrease their insulin dose might not be captured in EHRs. Also, because this study was completed at just one practice, the findings may not be generalizable to all practices or patient populations. Additionally, the small sample size prevented post hoc analyses to evaluate for potential confounders in the results.

Although pharmacokinetic changes in the dose-response relationship between basal insulin and glycemic control are observed beginning at doses of 0.3 units/kg/day, it is unknown at which dose of basal insulin treatment changes are warranted (2). It is possible, in theory, that patients in the range of 0.3–0.5 units/kg/day were perhaps not receiving adequate doses of basal insulin and not attaining glycemic targets as a result of clinical inertia. Although prescribed BGM, this group had the lowest documentation of BGM values in the EHR, preventing further assessment of possible indicators of overbasalization (e.g., calculation of BeAM differentials, elevations in postprandial glucose, and above-goal A1C despite controlled fasting blood glucose). Further studies are needed in this range of basal insulin doses to understand how changes in the dose response contribute to the lack of glycemic control.

Our study highlights the need for strategies to identify patients who are overbasalized and to prevent overbasalization as insulin doses are titrated. One strategy implemented by the authors since the completion of this study is to place a maximum dose on titration instructions when initiating basal insulin (e.g., “inject 10 units daily and increase by 2 units every 3 days until the morning blood glucose is 80–30 mg/dL, to a maximum of 30 units”). It is not yet known whether this strategy has been effective.

This study also emphasizes the need for increased monitoring upon initiation of basal insulin. With increased patient access to CGM for patients on basal-only insulin therapy, there is the potential to overcome clinical inertia in overbasalized patients by recognizing patterns of postprandial hyperglycemia and thereby prompting initiation of prandial insulin or therapies specific to postmeal glucose elevations. This study also supports ongoing investigation of the outcomes associated with using a lower threshold to define overbasalization in day-to-day practice among primary care providers.

This study’s findings suggest that, although overbasalization was not found to be as prevalent as anticipated based on a previous study, the presence of overbasalization in patients with type 2 diabetes was associated with a higher-than-average A1C and an overall decreased ability to achieve optimal glycemic control. This observation was consistent with previous work in this area and highlights the importance of identifying overbasalization at 0.5 units/kg/day as a practical threshold for clinicians involved in the management of type 2 diabetes. However, this study also supports the potential threshold of 0.3 units/kg/day as indicative of an opportunity to begin considering therapy intensification while purposely adjusting basal insulin doses.

The authors thank Michael C. Grimes, MD, from the Center for Diabetes and Endocrine Health at Allegheny Health Network for his review and feedback on a manuscript draft.

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

All authors substantially contributed to the study conception and design, participated in acquisition and/or analysis and interpretation of data, wrote and/or revised the manuscript critically for important intellectual content, and approved the manuscript for submission. A.S.-L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Portions of this work were previously published in abstract form in American Pharmacists Association. APhA 2022 Annual Meeting and Exposition contributed papers program abstracts. J Am Pharm Assoc 2022,62:947–1112.