Clinical consequences of hyperglycemia during remission induction therapy for pediatric acute lymphoblastic leukemia

Abstract

Hyperglycemia adversely affects outcomes in adult patients with acute lymphoblastic leukemia (ALL), but its impact on children with this disease is unknown. We evaluated the relationship between hyperglycemia during remission induction therapy and clinical outcomes among pediatric patients with ALL. We reviewed the records of patients enrolled on four consecutive ALL protocols (Total Therapy protocols XIIIA, XIIIB, XIV, and XV) at St. Jude Children's Research Hospital from 1991 to 2007 and identified those who experienced hyperglycemia (glucose ≥200 mg/dL) during remission induction. Complete remission rates at the end of induction, event free survival, overall survival, cumulative incidence of relapse, and occurrence of infections were compared between those who did and did not experience hyperglycemia. Of 871 patients analyzed, 141 (16%) experienced hyperglycemia during remission induction. Patients with hyperglycemia were significantly older than the other patients (p<0.0001). There was no significant difference in complete remission rate (p=0.92), event-free survival (p=0.80), overall survival (p=0.28), cumulative incidence of relapse (p=0.59), or in the probability or types of infection between patients who did and did not experience hyperglycemia. Pediatric patients with or without hyperglycemia during remission induction for ALL have similar clinical outcome. Occurrence of hyperglycemia does not warrant alteration of the antileukemic regimen.

Introduction

Hyperglycemia occurs in approximately 10% of patients receiving remission induction therapy including a gluoccorticoid and L-asparaginase for acute lymphoblastic leukemia (ALL) (1-12) and can be associated with acute complications such as non-ketotic hyperglycemic hyperosmolar syndrome or diabetic ketoacidosis. Hyperglycemia is associated with increased mortality in critically-ill adult and pediatric patients without a previous history of diabetes, though no specific threshold level of hyperglycemia has been defined that correlates with the increased mortality in these patients (13-17). In hospitalized adults with non-critical illness, hyperglycemia has been associated with increased in-hospital mortality, longer length-of-stay, and more frequent admissions to intensive care units (18).

In addition to increased mortality rates, more infectious complications have also been noted in patients without preexisting diabetes who experience hyperglycemia during illness, as observed in burn and post-operative patients. (19, 20). In pre-term neonates, hyperglycemia is also seen more frequently prior to the onset of invasive fungal (but not bacterial) infections, though a causal relationship has not been established (21).

Several mechanisms may explain the association between hyperglycemia and poor clinical outcomes. In studies of known diabetics, phagocytosis and microbicidal activity of white blood cells is impaired during periods of poor glucose control (22). Hyperglycemia may suppress immune function by inhibition of endogenous production of interleukins 2, 6 and 10 (23-25). Extremely elevated glucose concentrations, such as those seen in diabetic ketoacidosis and non-ketotic hyperglycemic hyperosmolar syndrome, affect T-cell proliferation and response to Candida albicans in vitro, and patients with poorly-controlled diabetes have an increased risk for invasive fungal infection (24). Furthermore, rabbit models using lipopolysaccharides to simulate septic shock showed that hyperglycemia altered the inflammatory responses, and negatively affected hemodynamic status (26). This is consistent with clinical observations in pediatric patients after cardiac surgery, necrotizing enterocolitis, traumatic brain injury, and meningococcal septic shock (27-30). The fact that infection remains a common complication during treatment for ALL (31-34) suggests that these inflammatory and immune effects of hyperglycemia may be particularly relevant to ALL patients. In addition, some studies suggest that hyperglycemia may directly affect cell growth and thus promote malignant cell proliferation (35, 36). Adult patients who develop hyperglycemia during remission induction therapy for ALL have a decreased rate of complete remission, shorter duration and lower probability of event-free survival, and an increased number and severity of infections (37). However, the impact of hyperglycemia during remission induction therapy in children with ALL is not known, and is the subject of this study.

Materials and Methods

Study Population

We reviewed the records of 872 pediatric patients with newly diagnosed ALL enrolled on four consecutive clinical trials (Total Therapy protocols XIIIA (38), XIIIB (39), XIV (40), and XV (41)) at St. Jude Children's Research Hospital between 1991 and 2007. Informed consent from the parent or guardian and age-appropriate assent from the patient were obtained at the time of study enrollment. The remission induction regimens of all 4 protocols contained similar doses of agents known to cause hyperglycemia: oral prednisone 40 mg/m² for 28 consecutive days and 6 to 9 doses of L-asparaginase 10,000 units/m² administered every other day. One patient with a pre-existing diagnosis of diabetes mellitus at the time of ALL diagnosis was excluded. The median follow-up time was 8.5 years (range, 4 months to 16 years).

Assessment of Potential Risk Factors, Hyperglycemic Events, and Treatments

Medical records were reviewed to obtain demographic information; treatment protocol; leukemia risk category; dates of diagnosis, complete remission, relapse, death, and last follow-up; hospitalizations and infections.

Hyperglycemia was defined as one or more glucose values ≥ 200 mg/dL during remission induction therapy. Glucose levels were collected on the date of scheduled chemotherapy and as clinically indicated. Patients with single glucose values ≥ 200 mg/dL whose results were felt to be spurious and whose repeated glucose values within 1 hour without intervention were <200 mg/dL were not considered to have had hyperglycemia (n=12, 1.3% of elevated values). Resolution of the hyperglycemic episode was defined as a 72-hour period during which all blood glucose measurements were <200 mg/dL and no insulin treatment was administered. The date of resolution was defined as the first day of three days in which no insulin was given and no glucose value exceeded 200 mg/dL. To reduce the possibility that spurious values misclassified patients as having hyperglycemia, repeat analysis was performed in which a hyperglycemia episode was defined as 2 or more glucose values ≥200 mg/dL.

Hospital admission and discharge dates and reason for admission were recorded. In the event of fever or infection during hospital admission, type and location of infection and any identified causal organism were recorded. It is our standard practice to admit all patients with fever or suspected infection during remission induction therapy; thus, no infectious complications during induction therapy were initially managed in the outpatient setting.

Response Criteria and Statistical Considerations

Primary endpoints included comparisons of the rates of complete remission at the end of remission induction therapy, event-free survival, overall survival and cumulative incidence of relapse between patients who did or did not experience one or more episodes of hyperglycemia during remission induction. Complete remission was defined as a bone marrow with less than 5% blasts and with restoration of normal hematopoiesis (38-41). Event-free survival duration was measured from the date of diagnosis until the date of relapse, death or most recent follow-up. Failure to enter remission was considered an event at time zero. Data for patients who were alive at the most recent contact date were censored at the time of that contact. The duration of overall survival was measured from the date of diagnosis until the date of death or most recent follow-up. Event-free survival and overall survival distributions were estimated by the Kaplan-Meier method (42) and compared between patients with or without hyperglycemia using the log-rank test (43). Multivariable Cox regression models (for event-free survival and overall survival) and multivariable logistic regression model (for complete remission) were used to control for the effect of patient and leukemia characteristics on the relationship between hyperglycemia and clinical outcomes (44). Factors significant at the 0.10 level in univariate analysis were entered into the multivariable models.

In the analysis of the cumulative risk of relapse, second cancer or death due to any cause were regarded as competing events. Patients who did not enter remission were excluded, and those who remained alive and in remission were censored at the time of last contact. The risk of relapse and competing risks were estimated as described by Kalbfleisch and Prentice (45) and compared by the method described by Gray (46).

Secondary endpoints included comparisons of the number of hospital admissions for infection and severity of those infections, and the number of fungal infections in those with or without hyperglycemia during remission induction. Infections were categorized in an analogous manner to the study by Weiser et al (37) for ease of comparison: any infection, uncomplicated infection (upper respiratory infection, urinary tract infection, or fever without documented infectious source), complicated infection (sepsis/symptomatic bacteremia, infection affecting pulmonary, cardiovascular, renal, central-nervous-system or gastrointestinal system or soft tissues), or fungal infection (including all documented fungal infections except oropharyngeal candidiasis). The mean number of infections per patient in each group was compared using the t-test. The numbers of infections in each infection category were compared using the chi-squared test. All statistical tests were performed using SAS version 9.1 (Cary, NC). Statistical analyses were considered significant at a type I error rate of p=0.05. All reported p-values are 2-sided.

Results

Of the 871 patients evaluated, 141 (16%) experienced hyperglycemia during remission induction (of which 70 patients had more than one glucose value >200mg/dL) while 730 (84%) did not. Baseline characteristics between patients with and without hyperglycemia during induction are shown in Table I. Only age greater than 10 years was associated with hyperglycemia on multivariable analysis (p<0.0001).

Table 1.

Comparison of Baseline Characteristics in Patients With vs. Without Hyperglycemia During Remission Induction

Patient Characteristics	Total n=871 (%)	Hyperglycemia No n=730 (%)	Hyperglycemia Yes n=141 (%)	p-value Univariable	p-value Multivariable	Odds Ratio (95% CI)
Age at Diagnosis [n (%)]
0 to 9 years	628 (70)	548 (87)	80 (13)	<0.0001	<0.0001	1**
≥ 10 years	243 ( 28)	182( 75)	61( 25)			2.3 (1.5 to 3.4)
Gender [n (%)]
Male	485 ( 56)	401( 83)	84( 17)	0.31
Female	386 ( 44)	329( 85)	57( 15)
Race [n (%)]
White	659 (76)	554 (84)	105 (16)	0.80
Black	151 (17)	124 (82)	27 (18)
Other	61 (7)	52( 85)	9 (15)
Risk Group* [n (%)]
Lower-risk	347 (40)	300(87)	47( 13)	0.09	0.88	1**
Higher-risk	524 (60)	430 (82)	94 (18)			1.0 (0.68 to 1.6)
Study number [n (%)]
Total XIIIA	165 (19)	140 (85)	25( 15)	0.39
Total XIIIB	246 (28)	213 (87)	33( 13)
Total XIV	53 (6)	42 (79)	11 (21)
Total XV	407(47)	335 (82)	72( 18)

^*Risk group was determined by a combination of age, white blood cell count at diagnosis, immunophenotype, DNA index, and response to initial therapy.

^**Reference group

The relationship between hyperglycemia during remission induction and treatment-related infection during induction is shown in Table 2. Patients who experienced hyperglycemia had no increase in the number of infections overall or in any severity category. The mean number of infections per patient in patients who experienced hyperglycemia was 1.4 versus 1.2 in those who did not experience hyperglycemia (p=0.34). The mean number of infections in those who experienced 2 or more abnormal glucose values was 1.4 versus 1.3 in those who did not experience hyperglycemia (p=0.66).

Table 2.

Comparison of Infectious Subtypes In Patients With and Without Hyperglycemia During Remisson Induction Therapy

Outcomes	Hyperglycemia
	Yes (n=141)	No (n=730)	p-value
Achieved complete remission	135(96)	719 (95)	0.92
*5-year estimated event-free survival	77.6 ± 4.8%	82.9 ± 1.7%	0.80
*5-year estimated overall survival	81.9 ± 4.3%	89.1 ± 1.4%	0.28
Toxicities
Any infection	104 (73%)	500 (68%)	0.71
Uncomplicated infection (URI, UTI, Fever without documented infection)	62 (44%)	316 (43%)	0.88
Complicated infection (Sepsis/symptomatic bacteremia, infection of lungs, kidney, heart, soft tissue, CNS, GI)	35 (25%)	159 (22%)	0.42
Fungal infection	7 (5%)	25 (3%)	0.37

^*Percent probability of EFS and OS with standard error, p-values obtained from Cox Proportiona Hazard multivariable analysis

Comparisons of event-free survival and overall survival are shown in Table 2 and Figures 1 and 2, respectively. In univariable analysis, patients with hyperglycemia during induction were significantly less likely to achieve complete remission by the end of induction (p=0.03). However, after adjustment for other baseline characteristics that independently correlate with ALL remission rates (e.g., age), the effect of hyperglycemia was not significant (p=0.92). Pediatric patients who experienced hyperglycemia also did not exhibit significant differences in event-free suvival (p=0.80) and overall survival (p=0.28) compared with those who did not experience this complication. Similarly, hyperglycemia was not associated with cumulative risk of relapse (p=0.59, Table 2 and Figure 3). When hyperglycemia was defined as 2 or more glucose values >200 mg/dL; the results did not change: p=0.89 for achievement of CR, p=0.76 for EFS, and p=0.46 for OS. Sensitivity analysis was also performed to determine whether results would differ if different threshold glucose values were used to define hyperglycemia (>250, 300, 350, 400 and 450 mg/dL), but the occurrence of hyperglycemia was not associated with any of the clinical outcomes, regardless of the threshold chosen.

Discussion

In contrast to the findings by Weiser et al (37) in adults, our results indicate that pediatric patients who experience acute hyperglycemia as a complication of remission induction therapy for ALL do not differ in their probability of complete remission, event-free survival, overall survival and cumulative risk of relapse, compared with patients who do not experience this complication. In addition, unlike in adults, pediatric patients who experience hyperglycemia during remission induction do not experience an increased number or severity of infections. Because of wide variation in the management of hyperglycemic patients within our institution, we cannot attribute the lack of increased infections in hyperglycemic patients to a particular approach to glucose control. Whether tight glucose control should be attempted in patients with ALL is controversial. Overuse of insulin poses a risk for hypoglycemia and the continuous use of glucocorticoids, L-asparaginase, and highly variable diet due to emetogenic chemotherapy make the response to insulin therapy particularly unpredictable (47). Six patients experienced diabetic ketoacidosis, and are discussed in detail in our previous publication (47); zero patients experienced non-ketotic hyperglycemic hyperosmolar syndrome. Our results suggest that occurrence of hyperglycemia in children with ALL does not require modification of antileukemic therapy.

There are several potential reasons for the difference in the impact of hyperglycemia on ALL treatment outcomes in adults versus children. Adults enter therapy with a much higher incidence of co-morbidities than children, which puts them at higher risk for multi-organ dysfunction during remission induction therapy (48). Recent data also illustrates significant inter- and intra- patient variability in dexamethasone pharmacokinetics, with increased systemic exposure to dexamethasone in older children and adults (49). It is known that older age adversely affects outcome in the adults with ALL and that older age is also related to increased insulin resistance and decreased insulin secretion, factors that may contribute to the effect of hyperglycemia on outcomes (50).

This retrospective study has several limitations. Glucose values were obtained as part of routine clinical care rather than in a protocol-controlled fashion. While it is our standard practice to assess serum chemistries at least twice per week during remission induction, it is possible that patients exhibiting signs and symptoms of illness underwent more blood chemistry studies, potentially introducing selection bias. Evaluation of infection was performed and treatment given based on clinical judgment in response to symptoms and thus some minor infections or fastidious organisms may not have been identified. In addition, this is a single-institution study at St. Jude Children's Research Hospital, where unique opportunities for aggressive supportive care and monitoring exist, so results may not be generalizable to other settings. We did not evaluate treatment delays or modifications related to infection or hyperglycemia. However, the absence of a difference in survival between hyperglycemic and non-hyperglycemic groups, suggests that any delays or modifications did not have a significant impact. In addition, hyperglycemia and its complications may result in more clinic visits and increased frequency of laboratory tests, which may increase medical costs and decrease quality of life, which were also not assessed in the current study. Finally, we did not specifically study risk factors for hyperglycemia nor management strategies, which were not the focus of this investigation.

In order to establish evidence-based guidelines for the management of hyperglycemia during pediatric ALL therapy, prospective, randomized trials may be necessary to establish the ideal range of glucose control and to assess factors that affect the persistence of hyperglycemia as well as its late effects. Until these guidelines are established, it is important to recognize that excessive insulin therapy and resulting hypoglycemia independently contribute to poor outcomes in other critically ill populations (14, 17). Further research is necessary to establish the ideal range for glucose control in ill pediatric patients without pre-existing diabetes. Adults studies and in-vitro data suggest that a glucose value of 250 mg/dL impairs neutrophil function even without preexisting diabetes (22, 37), while data by Wintergrest et al (17) suggests that values below 66 mg/dL are associated with increased morbidity and mortality Until these thresholds are confirmed, pediatric leukemia patients should be closely monitored for hyperglycemia and symptoms of diabetic ketoacidosis, and when managed with insulin, vigilantly monitored for the development of hypoglycemia.

Conclusion

In summary, pediatric patients who experience acute hyperglycemia as a complication of remission induction for ALL have the same treatment outcome as those who do not experience hyperglycemia. Occurrence of hyperglycemia does not warrant alteration of antileukemic therapy.