Relationships among biochemical measures in children with diabetic ketoacidosis

Abstract

Objectives: Investigating empirical relationships among laboratory measures in children with diabetic ketoacidosis (DKA) can provide insights into physiological alterations occurring during DKA. We determined whether alterations in laboratory measures during DKA conform to theoretical predictions.

Methods: We used Pearson correlation statistics and linear regression to investigate correlations between blood glucose, electrolytes, pH and PCO₂ at emergency department presentation in 1,681 pediatric DKA episodes. Among children with repeat DKA episodes, we also assessed correlations between laboratory measures at the first vs. second episode.

Results: pH and bicarbonate levels were strongly correlated (r=0.64), however, pH and PCO₂ were only loosely correlated (r=0.17). Glucose levels were correlated with indicators of dehydration and kidney function (blood urea nitrogen (BUN), r=0.44; creatinine, r=0.42; glucose-corrected sodium, r=0.32). Among children with repeat DKA episodes, PCO₂ levels tended to be similar at the first vs. second episode (r=0.34), although pH levels were only loosely correlated (r=0.19).

Conclusions: Elevated glucose levels at DKA presentation largely reflect alterations in glomerular filtration rate. pH and PCO₂ are weakly correlated suggesting that respiratory responses to acidosis vary among individuals and may be influenced by pulmonary and central nervous system effects of DKA.

Keywords: acid-base balance; diabetes; diabetic ketoacidosis; electrolytes.

Figure 1:

Correlations between biochemical measures in children with diabetic ketoacidosis, ordered by descending correlation estimates (p<0.001 for all correlations)*†‡. We computed Pearson correlation statistics to describe associations between baseline laboratory values recorded for each visit, and the probability of observing the same or higher correlation coefficient under the null hypothesis of zero correlation (p-value). We also calculated slope estimates to describe the relationship between laboratory measures using simple linear regression. To address possible effects related to multiple DKA episodes in the same individuals, we re-calculated Pearson correlation statistics and p-values for baseline laboratory value associations after excluding repeat enrollments so that each patient was represented only once. Statistical analyses were performed using SAS/STAT software, version 9.4 (SAS Institute; Cary, NC). *Units are as follows: bicarbonate, chloride and potassium: mEq/L; PCO₂: mmHg; glucose, BUN: mg/dL; creatinine: z-score adjusted for age; sodium: mEq/L adjusted for glucose: measured sodium + 1.6 ([blood glucose-100]/100). †r: Pearson correlation coefficient. All p-values are <0.001; β₁: slope estimate from a linear regression model; 95% prediction ellipses are shown which would be expected to contain 95% of new observations from the same population. ‡ Some participants have either pH or bicarbonate levels above usual cutoffs for diagnosis of DKA because either pH<7.25 or serum bicarbonate concentration <15 mmol/L was used to define DKA. In addition, although the first measured biochemical values for study participants were recorded in the database, some patients had greater acidosis on subsequent measurements, meeting criteria for diagnosis of DKA sometime after the first biochemical measurements were made.

Figure 2:

Correlations between biochemical measures among individuals presenting with two separate DKA episodes, ordered by descending correlation estimates. Among patients with two DKA episodes (n=128), we calculated Pearson correlations and p-values comparing baseline laboratory values at the first vs. second visit. *Graphs show 95% prediction elliptical curve. Note that some participants have either pH or bicarbonate levels above usual cutoffs for diagnosis of DKA because either pH<7.25 or serum bicarbonate concentration <15 mmol/L was used to define DKA. In addition, although the first measured biochemical values for study participants were recorded in the database, some patients had greater acidosis on subsequent measurements, meeting criteria for diagnosis of DKA sometime after the first biochemical measurements were made. †r: Pearson Correlation Coefficient; 95% prediction ellipses are shown which would be expected to contain 95% of new observations from the same population.