Hyperglycemia and other glycemic measures throughout therapy for pediatric acute lymphoblastic leukemia and lymphoma

Abstract

Transient hyperglycemia during induction chemotherapy is associated with increased morbidity and mortality in patients with acute lymphoblastic leukemia (ALL). Treatment with glucocorticoids, asparaginase, and stress are proposed causal factors. While these risks are not exclusive to induction, glycemic control throughout the remainder of ALL/lymphoma (ALL/ALLy) therapy has not been described. Further, prior research has been limited to transient hyperglycemia. This study aimed to characterize glycemic control throughout ALL/ALLy, and to evaluate risk factors and outcomes associated with increased mean glucose and glucose coefficient of variation (glucose CV) during induction chemotherapy. Records for 220 pediatric/young adult patients, age 1–26 years, who underwent treatment for ALL/ALLy from 2010–2014 at Children’s Hospital Colorado were retrospectively reviewed. Measures of glycemic control were calculated for each cycle. For the cycle with the highest mean glucose, induction (n=208), multivariable models were performed to identify potential risk factors and consequences of increased glucose. Highest mean glucose by cycle were: induction 116 mg/dL, pre-treatment 108 mg/dL, delayed intensification 96 mg/dL and maintenance 93 mg/dL; these cycles also had the most glycemic variability. During induction, patients with Down syndrome, or who were ≥12 years and overweight/obese, had higher mean glucoses; age and overweight/obese status were each associated with increased glucose CV. In multivariable analysis, neither induction mean glucose nor glucose CV were associated with increased hazard of infection, relapse, or death.

Introduction

Hyperglycemia is a well-described side effect of induction chemotherapy for acute lymphoblastic leukemia/lymphoma (ALL/ALLy), and in some studies has been shown to be associated with increased infection and mortality.^1–6 Hyperglycemia in this population is thought to be most commonly due to glucocorticoid therapy, but asparaginase can also contribute by decreasing insulin and other protein synthesis, as well as being a risk factor for pancreatitis.^1,7–9 Despite use of these agents in other phases of treatment (such as asparaginase in delayed intensification (DI) and some interim maintenance II (IMII cycles), and glucocorticoids (GC) in DI and maintenance), there is minimal research exploring hyperglycemia during these intervals. A single study, Yoshida et., describes transient hyperglycemia (single glucose excursions >250 mg/dL) during maintenance chemotherapy, while another evaluates insulin resistance specifically.^7,10 Further, physiologic stress is well-described risk factor for hyperglycemia, putting patients undergoing chemotherapy for ALL/ALLy at potential risk beyond exposure to asparaginase and GC.¹¹ Because of the potential implications of abnormal glucose control on adverse outcomes, we sought to describe glycemic control throughout ALL/ALLy therapy.

In addition to focusing on induction chemotherapy, prior studies to date define hyperglycemia in varying manners of transient hyperglycemia, such as 1–2 measurements above a certain cutoff. Further, while a few studies evaluated the effects of “mild” versus “severe” hyperglycemia, no studies have evaluated glucose continuously to assess for more mild effects of increased glucose on outcomes. Our prior research demonstrated that mean glucose alone may be associated with increased risk of infection and death in the hematopoietic stem cell transplant population.¹² As such, this study explores adverse effects associated with, and risk factors for, persistent subtle increases in glucose during the cycle found to have the highest mean glucose, induction.

In summary, this study aims to broaden existing knowledge of glycemic control in ALL/ALLy through describing glycemic measures throughout therapy and to explore induction chemotherapy as a test case for the clinical relevance of sustained or mild increases in glucose.

Materials and Methods

We performed a retrospective cohort study of patients (ages 1–26 years), diagnosed with acute lymphoblastic leukemia or lymphoma (ALL/ALLy) who were treated at a single academic children’s hospital between 2011–2014. Patients with pre-existing diabetes mellitus or Philadelphia chromosome-positive disease were excluded. This study was approved by the University of Colorado Institutional Review Board.

A combination of an existing patient database, electronic health record, and manual review was used for data extraction. Disease characteristics were divided by cell line and grouped into post-induction risk categories using Children’s Oncology Group (COG) definitions at the time of therapy, typically: 1) pre-B ALL/Ly standard risk (white blood cell count less than 50,000 cells/mm3, age 1 to 10 years old, no unfavorable genetics, central nervous systemic (CNS) disease is less than CNS 3, no testicular involvement, and day 29 minimal residual disease less than 0.01%; 2) pre-B ALL/Ly high/very high risk (any factors not meeting standard risk); 3) T-cell ALL/Ly. Chemotherapy regimens were defined by COG protocols at the time of treatment, with GC type as follows: either dexamethasone and prednisone during induction and maintenance (depending on the study and/or patient age), and dexamethasone during DI. Data collected included demographic data, diagnosis and treatment details, and whole blood glucose (BG) values. BG data included fasting and nonfasting values from up to seven days prior to diagnoses through end of treatment, relapse, induction failure or death (whichever occurred first). All whole BG values were included; glucose values >600 were set to 600 (n=13). Patients with <3 BG values for any given cycle were excluded from that cycle.

Data Analysis

Prior to calculating measures of glycemic control, interval 3-hour mean glucoses were calculated to avoid bias from repeat testing. Each cycle’s glycemic measures were then calculated using these interval means. Glycemic measures were calculated, including mean, median, interquartile range (IQR), and coefficient of variation (CV). CV, calculated as mean divided by standard deviation, was chosen to represent glycemic variability, or fluctuation of glucose within a single patient, as it is emerging as the preferred metric due to its advantage of being relative to mean glucose; it is therefore reflective of both hyperglycemia and hypoglycemia, whereas SD is largely only reflective of hyperglycemic excursions.^13–16

Odds ratios and corresponding p-values were used to assess the effect of overt hyperglycemia before treatment and, separately, during induction, on the risk of overt hyperglycemia in a future cycle. Overt hyperglycemia was defined as any patient-window for which the average glucose value was >140 mg/dL as this target has been suggested for other high risk populations such as pregnancy.¹⁷ Similar analyses were performed to assess the effect of the presence of abnormal glucose CV during earlier cycles on the risk of abnormal glucose CV in later cycles. Because no normal glucose CV ranges have been described in pediatric patients, we exploratorily categorized abnormal glucose CV two ways: 1) levels above the healthy adult range (≥20%) and 2) levels consistent with poorly controlled diabetes (≥36%).^14,18–20

Analyses were performed to explore the clinical relevance of mean glucose and glucose CV during induction chemotherapy. For this analysis, the above cohort was used, however excluding patients with <6 BG values during induction. A multivariable linear regression model was used to identify potential risk factors for increased mean glucose and increased glucose CV. Potential covariates were categorized: age (approximating puberty: ≤11, >12), ethnicity, Down syndrome, body mass index (BMI) percentile categorized by standard World Health Organization definitions and further grouped according to typical pediatric diabetes literature (underweight/normal and overweight/obese), ALL/ALLy subtype and post-induction risk category (standard risk, high/very high risk, T-cell); treatment with total parenteral nutrition (TPN) during induction. Multicollinearity was assessed for using variance inflation factors (VIF).Glucose CV was log-transformed due to a nonnormal distribution. Estimates from the model for log-transformed glucose CV were exponentiated, subtracted from 1 and multiplied by 100 to report effects as a percentage change in glucose CV.

We also assessed for associations between increased mean glucose and glucose CV and adverse outcomes, time to serious infections, relapse, and death. For time-to-event analyses, patients who died during induction or had induction failure were excluded (n=6), as were patients with missing predictor values (n=1). Cox proportional hazards models were used to evaluate associations between mean glucose and glucose CV and time to infection, death and relapse after induction. The model for time to infection was adjusted for age category, Down syndrome, ethnicity, body habitus, and ALL/ALLy subtype/risk. Because there were fewer relapses and deaths, these models adjusted only for covariates significant in univariate analyses evaluating relationships between covariates and mean glucose; the following variables were considered for these analyses: age, ethnicity, race, BMI category, Down syndrome, ALL/ALLy subtype/risk, and insulin use during induction. Kaplan-Meier curves were used to estimate event probabilities.

Results

Glycemic control throughout therapy

Across 220 included patients, 53% were male, 79% were <12 years old at diagnosis, 33% were Hispanic, 6% had Down syndrome, 15% were underweight, and 22% were overweight/obese (Table 1). Disease characteristics were: 42% standard risk pre-B ALL/Ly, 46% high risk/very high risk pre-B ALL/Ly, and 12% T-cell ALL/Ly. Insulin therapy was used in 16.8% of patients at any point in therapy, most commonly during induction (13.6%). Treating each cycle independently, 75% percent of all patients with overt hyperglycemia (mean glucose >140 mg/dL) were managed with insulin therapy; the proportion of patients with overt hyperglycemia who were treated with insulin was higher in induction (88%) and delayed intensification (100%).

TABLE 1.

Study Cohort Demographics

Characteristic	N=220 (%)
Age at Diagnosis
1–11 years	173 (78.6)
≥12 years	47 (21.4)
Sex
Female	103 (46.8)
Male	117 (53.2)
Race
Asian	6 (2.7)
Black/African-American	5 (2.3)
Multiple Races	8 (3.6)
Other	11 (5)
Native Hawaiian/Pacific Islander	1 (0.5)
White	189 (85.9)
Ethnicity
Hispanic	73 (33.2)
Non-Hispanic	145 (65.9)
Unknown	2 (0.9)
BMI category at diagnosis
Underweight	33 (15)
Normal	138 (62.7)
Overweight	23 (10.5)
Obese	26 (11.8)
Down syndrome
No	207 (94.1)
Yes	13 (5.9)
Diagnosis/risk category
B-ALL/Ly – SR	92 (41.8)
B-ALL/Ly – HR/VHR	101 (45.9)
T-ALL/Ly	27 (12.3)
Induction treatment protocol
AALL0932	95 (43.2)
AALL0232	38 (17.3)
AALL1131	36 (16.4)
AALL0434	26 (11.8)
AALL0331	21 (9.6)
Other	4 (1.8)
Consolidation treatment protocol
AALL0932	73 (33.2)
AALL1131	48 (21.8)
AALL0232	47 (21.4)
AALL0434	24 (10.9)
AALL0331	18 (8.2)
Other	10 (4.5)
TPN exposure during induction
No	214 (97.3)
Yes	6 (2.7)
Insulin therapy during cycle
Induction	30 (13.6)
Consolidation	6 (2.7)
Interim Maintenance I	1 (0.5)
Delayed Intensification	13 (5.9)
Interim Maintenance II	3 (1.4)
Maintenance	10 (4.5)
Any Cycle	37 (16.8)

The distribution of patients’ mean glucose, per cycle, is depicted in Figure 1. Induction, followed by DI, pre-treatment, and maintenance, had the highest mean BG values (Table 2).While median BG is again highest during induction, the second highest median BG cycle was interim maintenance I. Notably, across subjects, mean glucoses were most varied in induction, followed by pre-treatment and DI, demonstrated by wider IQRs (Table 2). Within-subject glycemic variability, represented by median CV, was highest during induction, followed again by DI, Pre-treatment and maintenance.

FIGURE 1. Distribution of per-patient glucose means by cycle.

The median glucose is depicted by the center line, while each box contains the 25^th-75^th percentiles of mean glucoses for each cycle. The whiskers denote the 5^th and 95^th percentiles, and values beyond these upper and lower bounds are considered outliers, marked with open circles. For ease of depiction, one extreme outlier (mean 452) in Delayed Intensification was excluded from the figure.

TABLE 2.

Glycemic indices throughout acute lymphoblastic leukemia/lymphoma therapy

		Cycle
		Pre- treatment	Induction	Consolidation	Interim Maintenance I	Delayed Intensification	Interim Maintenance II	Maintenance	All Cycles
Patients	N	105	220	163	194	186	112	202	220
# glucose measurements (3hr) per person	Median	4.00	16.00	9.00	8.00	7.00	7.00	27.00	75.50
Mean Glucose (mg/dl) Glucose	Mean	107.59	116.38	85.83	91.44	95.74	87.60	93.05	99.79
	Median	104.00	110.36	84.00	90.13	86.88	86.60	88.93	94.81
	Min	79.64	64.70	63.86	66.43	59.56	59.38	63.17	73.69
	Max	171.14	290.07	230.09	157.23	452.00	186.87	232.56	198.06
	IQR	17.83	24.09	11.79	13.73	15.75	10.60	10.56	13.59
	N (%) > 140	6 (5.7)	25 (11.4)	2 (1.2)	2 (1.0)	11 (5.9)	2 (1.8)	4 (2.0)	10 (4.5)
Coefficient of Variation (within subject)	Median	15.63	28.26	15.90	13.76	19.34	13.87	16.49	24.80
	Min	2.92	8.62	0.84	2.09	2.51	2.82	5.07	11.26
	Max	85.03	74.16	61.47	40.91	92.24	70.72	92.81	94.68
	IQR	14.31	11.05	10.18	7.76	15.13	11.08	10.40	10.74

IQR; interquartile range

Sustained overt hyperglycemia, defined as a mean glucose >140 mg/dL, occurred in 11.4% of patients during induction, 5.9% during DI, and 5.7% pre-treatment (Table 2). Mean glucose >140 mg/dL during induction was associated with having overt hyperglycemia in any future cycle (OR 56; p<0.001). None of the 9 patients who had a mean glucose >140 mg/dL prior to treatment had any future cycles with mean glucose >140 mg/dL. Neither abnormal CV (>20%, see Methods) before treatment (OR 1.3; p=0.63). or during induction (OR 0.89; p=0.88) was associated with future abnormal glucose CV.

Implications of, and risk factors for, increasing mean glucose during induction

To evaluate the importance of sustained increased mean glucose throughout therapy, we evaluated associated outcomes, and risk factors for, increased mean glucose during induction. This cycle was chosen as it had the highest mean glucose of all treatment cycles.

After excluding patients with insufficient induction data, across 208 included patients, 54% were male, 77% were <12 years old at diagnosis, 33% were Hispanic, 6% had Down syndrome, 16% were underweight, and 22% were overweight/obese. Disease characteristics were 41% standard risk, 46% high risk, and 13% T-cell. Median follow-up from end of induction was 70.2 months (IQR 27.2).

During induction chemotherapy, 12.0% of patients experienced sustained overt hyperglycemia, with a mean glucose >140 mg/dL. Among presumed post-pubertal patients (≥12 years old), 34% had overt hyperglycemia, compared to 5.6% in the younger cohort (p<.001). Overt hyperglycemia was present in 16.8% of females and 8.0% of males (p=0.050), and 13% of Hispanic patients and 11.6% of non-Hispanic patients (p=0.84). In multivariable analysis for potential risk factors for increased mean glucose, patients ≥12 years old who were overweight or obese had a mean glucose 38 mg/dL higher than patients who were ≥12 years old and normal or underweight (p<0.001) (Figure 2). Patients with Down syndrome also were at increased risk for higher mean glucoses, with a mean glucose 20 mg/dL higher than patients without Down syndrome (p=0.016) (Figure 2); 23.1% of patients with Down syndrome had overt hyperglycemia. Due to the association between ALL/ALLy subtype/risk and age >10 years, we assessed for collinearity between age and ALL/ALLy subtype/risk, as well as all potential risk factors, and the variance inflation factors were less than 1.9 suggesting minimal collinearity.

FIGURE 2. Risk factors for increasing mean glucose during induction.

This forest plot demonstrates a multivariable linear regression model evaluating for factors associated with increased mean glucose. All variables present were included in the model, including an interaction between age and body habitus (p<0.001), as well as unknown ethnicity (n=2; estimate=−15.0, 95% CI: −70.2, 40.2, p=0.6). Risk factors for increasing mean glucose included overweight/obese body habitus in patients ≥12 (p<0.001), and Down syndrome (p=0.016). Other demographic and disease-related factors were not associated with increasing mean glucose.

Among patients who achieved remission with induction chemotherapy, 48 patients developed infections during follow-up (25.6% at 3 years). In univariate (p=0.065) and multivariable (p=0.95) analyses, mean glucose was not significantly associated with risk of infection (Figure 3).

FIGURE 3. Hazard ratios for adverse events after induction for every 10 mg/dL increase in mean glucose during induction.

This forest plot demonstrates Cox proportional adjusted hazard ratios for time-to first infection, relapse and death after induction. There were no significant associations between increasing mean glucose during induction and adverse events after induction. Time-to-first infection was adjusted for age, Down syndrome, ethnicity, body habitus and ALL/ALLy subtype/risk. Time-to-death was adjusted for body habitus. Time-to-relapse was adjusted for Down syndrome diagnosis.

There were 22 patients who had a relapse after induction (10.8% at 3 years). Mean glucose was not significantly associated with risk of relapse in univariate or multivariate analysis (Figure 3). After induction, 16 patients died (6.6% at 5 years). In univariate analysis, every 10 mg/dL increase in mean glucose was associated with a 10% increased hazard of death (95% confidence interval [CI] 1.0%, 20.9%; p=0.048). However, after adjusting for body habitus, there was no significant association between mean glucose and death (Figure 3).

Implications of, and risk factors for, increasing glucose CV during induction

As with mean glucose, we evaluated potential risk factors and associated outcomes with increased glucose CV during induction. This cycle was chosen as it had the highest glucose CV of all treatment cycles. The sample was the same as that evaluated for mean glucose above.

During induction chemotherapy, 188 (88%) had an elevated glucose CV (≥20%) and 44 (21%) had a glucose CV comparable to an adult with labile diabetes (≥36%). In univariate analysis, age >12 years, Down syndrome, and high or very high risk disease were risk factors for increased glucose CV during induction. In multivariable analysis (Figure 4), age ≥12 years was associated with a 22% increase in glucose CV (p=0.002) and Down syndrome was associated with a 26% increase (p=0.018). Similar to mean glucose, there was a relationship between age and BMI category, though not significant (p=0.057); patients who were ≥12 years and overweight/obese had a higher glucose CV than those ≥12 years who were normal or underweight, and both were higher than patients ≥12 years. As with mean glucose, we evaluated for an collinearity between ALL/ALLy risk/category and age in association with glucose CV, which was minimal (<1.9)

FIGURE 4. Risk factors for increasing glucose CV during induction.

This forest plot demonstrates a multivariable linear regression model evaluating for factors associated with increased glucose CV (by percentage of change). All variables present were included in the model, as well as unknown ethnicity (n=2; estimate=7.5, 95% CI: −44.3, 107.5, p=0.8). Risk factors for increasing glucose CV included age ≥12 (p=0.002) and Down syndrome (p=0.018).

In univariate analysis, induction glucose CV was associated with increased risk of death after induction in time-to-event analysis (HR=2.91; p=0.034); this association was no longer statistically significant after adjusting for BMI category (HR=2.43; p=0.094). Induction glucose CV was not associated with time to infection (HR=1.71; p=0.09) or relapse (HR=1.51; p=0.39) after induction in univariate analysis, or in multivariable analysis (adjusting for the same factors as in mean glucose analysis).

Discussion

This retrospective cohort study demonstrates that ALL/ALLy patients may be at risk for glycemic derangements beyond induction chemotherapy. We found that DI and the pre-treatment phase, in particular, may also be high risk timepoints for increased glucose and glycemic variability. This is not surprising based on the use of both GC and asparaginase during these two chemotherapy cycles; notably, though, increased mean glucose during DI was driven by a number of patients with more extreme BG during this period, as evidenced by the lower median during DI. Future analyses are necessary to determine if factors such as GC type (prednisone vs. dexamethasone), ALL/LLy subtype, asparaginase cumulative dose, or other factors affect risk. In prior studies, the most commonly identified risk factor for hyperglycemia during induction chemotherapy is age ≥10 years; numerous prior studies have not demonstrated an association between steroid type and hyperglycemia.^2,21,22

We hypothesize that abnormal glycemic control may be more prevalent during these two intensive cycles due to increased risk for acute illness and stress hyperglycemia. Maintenance chemotherapy also includes use of GC therapy but did not appear to have the same level of increased mean glucose or variability; this may be due to the shortened courses of GC (five days) during maintenance cycles as well as less acute illness/physiologic stress. Further, based on local practice, most glucose values collected during maintenance would be either before or weeks after completion of GC therapy; additional research is needed to explore glycemic control during and immediately after GC therapy during maintenance.

As mean glucose was highest during induction, we assessed risk factors for increased mean glucose and glucose CV during this cycle. While overweight or obese patients older than 12 years, and patients with Down syndrome, are at increased risk for increased mean glucose, age ≥12 and Down syndrome were each independently associated with increased variability. The interaction between age and BMI in association with mean glucose is in contrast to other studies, in which increased risk for hyperglycemia with older age has been independently demonstrated. This difference may be attributed to our focus on sustained hyperglycemia, whereas most prior studies evaluated transient hyperglycemia.^2,22–25 We are not aware of prior literature demonstrating the interaction between body habitus and age, which gives insight to the most at-risk population. Further, while steroid type was dependent on age in some protocols, numerous prior studies have not demonstrated an association between steroid type and hyperglycemia, this is unlikely.^2,21,22

Abnormal glycemic control is critical to recognize, as it may be associated with adverse events such as infection and death.^2–6 Importantly, we did not find persistent and subtle increased glucose during induction, represented by mean glucose, to be associated with risk of infection, relapse, or death after induction. In addition to evaluating more overt hyperglycemia, though, many prior studies included events during induction itself; as such, in some cases hyperglycemic events may have been directly associated with the adverse outcome (stress hyperglycemia), versus hyperglycemia that precedes and possibly predisposes the patient to adverse outcomes.

Due to its retrospective nature, this study is limited in its ability to evaluate whether glucose values were accurate and not contaminated; use of 3-hour mean glucoses was used to help reduce this potential bias. We were also unable to determine whether glucose values were fasting or nonfasting, so were assumed to be nonfasting. For this reason, we primarily avoided defining hyperglycemia, favoring a focus on descriptive statistics using continuous metrics.

As mentioned, while mean glucose and glucose CV during induction were not associated with increased adverse events, multiple studies suggest that transient hyperglycemia is associated with adverse events. Whether hyperglycemia serves as a biomarker for infection and other events, or it is directly related to underlying pathophysiology, is unknown. However, hyperglycemia has known detrimental effects on immune function.^26–34 Research describing the immunologic effects of hyperglycemia in ALL/ALLy is warranted. While mild persistent increases in mean glucose may not be associated with outcomes, research is necessary to evaluate the implications of persistent versus transient overt hyperglycemia. Future research is necessary to evaluate potential factors associated with increased glucose, such as age, ALL/Ly subtype and risk category, ethnicity, presence of Down syndrome, GC type, and other potential factors.¹

In summary, patients undergoing treatment for ALL/ALLy may be at risk for abnormal glycemic control beyond the oft-studied induction cycle. Because overt hyperglycemia has been associated with increased risk of infections, mortality and other adverse events, further work describing glycemic control and associated risk factors and outcomes throughout ALL/ALLy therapy is warranted.