Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
June 2, 2008

Prolonged QT Interval Corrected for Heart Rate During Diabetic Ketoacidosis in Children

Author Affiliations

Author Affiliations: Departments of Emergency Medicine (Dr Kuppermann) and Pediatrics (Drs Kuppermann, Park, Marcin, and Glaser), and Division of Cardiology, Department of Internal Medicine (Dr Glatter), University of California, Davis, School of Medicine.


Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008

Arch Pediatr Adolesc Med. 2008;162(6):544-549. doi:10.1001/archpedi.162.6.544

Objective  To evaluate the effect of diabetic ketoacidosis (DKA) on the QT interval corrected for heart rate (QTc) in children. Ketosis occurs in several conditions, including DKA and alcoholic ketoacidosis, and during use of very low-carbohydrate diets. Prolongation of the QTc has been described in a few children receiving ketogenic diets, but cardiac effects of ketosis have not otherwise been investigated.

Design  For this observational study, we performed electrocardiography during DKA and after recovery. We measured QTc as the QT interval divided by the square root of the R-R interval and correlated QTc with clinical variables.

Setting  The pediatric emergency department and intensive care unit of an academic medical center.

Patients  Thirty children with type 1 diabetes mellitus and DKA.

Main Outcome Measure  The QTc during DKA.

Results  The mean (SD) QTc during DKA was 450 (38) milliseconds (range, 378-539 milliseconds). After recovery from DKA, the mean (SD) QTc decreased to 407 (36) milliseconds (range, 302-485 milliseconds; difference, 43 milliseconds; 95% confidence interval, 23-63 milliseconds) (P < .001). Fourteen of the 30 children (47%) had prolonged QTc during DKA (range, 450-539 milliseconds). After recovery from DKA, only 4 children (13%) had persistent QTc prolongation (range, 451-485 milliseconds). The anion gap was significantly associated with QTc prolongation (correlation coefficient, 0.49; P = .006). Most patients had no electrolyte abnormalities or hypoglycemia to account for QTc prolongation.

Conclusions  Prolonged QTc occurs frequently during DKA and is correlated with ketosis. Current guidelines regarding cardiac monitoring of children during DKA should be strictly followed, and electrocardiographic screening of patients with other ketotic conditions should be considered.

Cardiac arrhythmias and cardiac arrest have been described in children with diabetic ketoacidosis (DKA).13 These rare complications of DKA generally have been attributed to electrolyte abnormalities during DKA treatment. Case reports, however, describe prolongation of the QT interval corrected for heart rate (QTc) in children receiving ketogenic diets in the absence of electrolyte abnormalities.4 In these children, there was a positive correlation between QTc prolongation and β-hydroxybutyrate concentrations and between QTc prolongation and systemic acidosis. Prolonged QTc and/or sudden death presumed to be caused by cardiac arrhythmias has also been described in adolescents and adults using very-low-carbohydrate, calorie-restricted diets.58 Likewise, sudden death has been documented in adults with alcoholic ketoacidosis.9

The consistent association between various ketotic conditions and prolonged QTc and/or sudden death raises the question of whether ketosis may directly affect cardiac repolarization. In this study, we tested the hypothesis that QTc prolongation occurs in children during DKA and returns to normal with resolution of ketosis. We also assessed correlations of clinical and biochemical variables with QTc prolongation.


Patients were eligible for participation in the study if they were younger than 18 years, had been diagnosed as having type 1 diabetes mellitus, and had DKA, defined as a serum glucose level of greater than 300 mg/dL, venous pH of less than 7.25 and/or a serum bicarbonate level of less than 15 mEq/L, and a positive test result for urine ketones or a serum ketone level of greater than 3 mmol/L. Patients were ineligible if they were taking medications known to affect QTc or if they had other underlying conditions known to affect QTc. (To convert glucose to millimoles per liter, multiply by 0.0555; and bicarbonate to millimoles per liter, by 1.0.)


The study was approved by the institutional review board of the participating institution. A convenience sample of patients was enrolled into the study after written informed consent was obtained from parents or guardians. We treated enrolled patients according to a standardized DKA protocol. All patients received an initial infusion of 10 to 20 mL/kg of 0.9% isotonic sodium chloride solution, depending on the assessed degree of hypovolemia. Patients with persistently poor perfusion or hemodynamic instability after the initial fluid infusion were given additional infusions of 0.9% isotonic sodium chloride solution until normal perfusion and hemodynamic stability were established. Subsequent intravenous fluids were administered as 0.67% isotonic sodium chloride solution (sodium concentration, 112 mEq/L) with the rate calculated to replace an estimated deficit of 70 mL/kg spread across 48 hours. Insulin was administered via continuous intravenous infusion at an initial rate of 0.1 U/kg/h. Patients were not treated with bicarbonate. Potassium replacement therapy was initiated with intravenous fluid containing potassium chloride, 20 mEq/L, and potassium phosphate, 20 mEq/L, with the dosage adjusted to maintain serum potassium concentrations within the reference range. Glucose was added to the intravenous fluids when the serum glucose concentration was less than 300 mg/dL. For patients referred from outside hospitals, we changed to our treatment protocol on their arrival to the pediatric intensive care unit.


Each patient had an initial electrocardiogram (ECG) recorded during DKA. A second ECG was recorded after recovery from DKA (defined as a serum bicarbonate concentration of > 18 mEq/L). Follow-up ECGs were recorded at the time of hospital discharge, or at the patient's first outpatient visit after discharge from the hospital (median, 45 days after the initial ECG; range, 24 hours to 180 days). A standard 12-lead ECG was recorded at 25 mm/s, along with a lead II rhythm strip recorded at 50 mm/s. The QT and R-R interval measurements were recorded from each ECG by a single pediatric cardiologist/electrophysiologist (J.P.) who was masked to the patients' clinical and biochemical data. The QT interval was measured from the onset of the QRS complex to the end of the T wave. The end of the QT interval was defined as the intersection of a tangent to the steepest downslope of the dominant repolarization wave with the isoelectric line. The QTc was calculated according to the Bazett formula as the QT interval divided by the square root of the R-R interval, where the R-R interval is measured in relation to the previous QRS complex.10 Three separate measurements by the same pediatric cardiologist/electrophysiologist were obtained from each ECG, and the mean of these measurements was used as the value for the QTc. Lead II was used preferentially for QTc measurement.11 If a T wave could not be clearly discerned in lead II, then the next limb lead with a clearly discernible T wave was used. Prolongation of QTc was defined as a QTc of at least 450 milliseconds. To evaluate interobserver agreement, QTc measurements were repeated by a second cardiologist/electrophysiologist (K.G.) who was masked to the clinical and biochemical patient data and to the ECG measurements of the first cardiologist. These repeated QTc measurements by a second cardiologist were used for evaluation of interobserver agreement only. The initial QTc measurements by the primary study pediatric cardiologist were used in all subsequent analyses.


We compared QTc measurements during DKA with QTc measurements after recovery from DKA using the Wilcoxon signed rank test. We compared clinical and biochemical findings for patients with and without QTc prolongation using the Wilcoxon rank sum test for continuous variables and the Fisher exact test for categorical variables. For these comparisons, we calculated the anion gap as Na+ − (Cl + HCO3), where Na+ indicates sodium; Cl, chloride; and HCO3, bicarbonate. We used the weighted κ statistic to assess interobserver agreement in QTc measurement (rounded to 2 decimal places) between the 2 study cardiologists.12 We also evaluated the association between the anion gap and QTc after adjusting for age and heart rate using linear regression. We performed a similar analysis to adjust for serum potassium and calcium concentrations at the time of the ECG. We considered P < .05 to represent statistical significance. Unless otherwise indicated, data are expressed as mean (SD).


To convert bicarbonate to millimoles per liter, multiply by 1.0; calcium to millimoles per liter, multiply by 0.25; glucose to millimoles per liter, multiply by 0.0555; and potassium to millimoles per liter, multiply by 1.0.


In 30 children, QTc was measured during and after DKA (Table 1). Fifteen of these children had new onset of type 1 diabetes mellitus, and the other 15 had known type 1 diabetes. The ECG measurements were recorded a mean of 6.2 (5.0) hours after the initiation of treatment for DKA.

Table 1. 
Image not available
Clinical and Biochemical Data in 30 Children With Type 1 Diabetes Mellitus and DKAa

The mean QTc for these 30 children during DKA was 450 (38) milliseconds (range, 378-539 milliseconds) (Figure 1). After recovery from DKA, the mean QTc decreased to 407 (36) milliseconds (range, 302-485 milliseconds; difference, 43 milliseconds; 95% confidence interval [CI], 23-63 milliseconds) (P < .001). Fourteen of the 30 children (47%) had QTc measurements of 450 milliseconds or longer during DKA (range, 450-539 milliseconds).

Figure 1.
Image not available

Measurements of the QT interval corrected for heart rate (QTc) in 30 children during diabetic ketoacidosis (DKA) and after recovery.

After recovery from DKA, only 4 children (13%) had persistent QTc of 450 milliseconds or longer, and the degree of prolongation was mild (range, 451-485 milliseconds). Two of these 4 children had ECGs repeated several weeks after recovery from DKA, and in both cases the QTc was less than 450 milliseconds. The other 2 patients were lost to follow-up before the ECGs could be repeated in the outpatient setting. None of the children enrolled in the study experienced cardiac arrhythmias during DKA treatment. Determination of QTc by the second cardiologist showed substantial interobserver agreement with the first cardiologist (weighted κ, 0.62; 95% CI, 0.43-0.80; P < .001).12

Prolongation of QTc occurred with equal frequency among children with new-onset diabetes and those with known diabetes (Table 2). The anion gap at presentation was significantly higher in children with QTc prolongation than in those without (P = .008). In addition, the initial anion gap was significantly correlated with QTc values during DKA (correlation coefficient, 0.49; P = .006), suggesting a strong association between ketosis and QTc prolongation (Figure 2). In contrast, the anion gap at the time of ECG recording was not significantly correlated with QTc. There were no other significant differences in biochemical or clinical variables between the 2 groups (Table 2).

Figure 2.
Image not available

Correlation of serum anion gap with the QT interval corrected for heart rate (QTc) during diabetic ketoacidosis. Anion gap is calculated as Na+ − (Cl + HCO3), where Na+ indicates sodium; Cl, chloride; and HCO3, bicarbonate. The solid line represents the linear regression fit across all subjects. The correlation coefficient and P value are shown.

Table 2. 
Image not available
Clinical and Biochemical Values at Presentation of Patients With and Without QTc Prolongation During DKAa

Age-adjusted blood pressure values were normal in all children throughout the study, and no study participant experienced hypoxia. None of the children had hypoglycemia (defined as a blood glucose concentration of < 70 mg/dL) at the time of recording of the ECG. Hypokalemia (defined as a serum potassium concentration of < 3.3 mEq/L) was present in 5 children at the time of ECG recording (range, 2.4-3.2 mEq/L). Three of these 5 children had serum potassium concentrations of 3.2 mEq/L, just slightly below the reference range. Four of the 5 children with hypokalemia at the time of ECG had a QTc of at least 450 milliseconds. Six children had low serum calcium concentrations (< 8.8 mg/dL) at the time of ECG recording (range, 7.9-8.6 mg/dL). Two of these 6 children had a QTc of at least 450 milliseconds. Serum magnesium concentrations were within the reference range in all patients at the time of ECG recording. The QTc measurements were not significantly correlated with the serum potassium concentrations (correlation coefficient, −0.15; P = .44), the serum calcium concentration (correlation coefficient, 0.12; P = .58), or the serum magnesium concentration (correlation coefficient, 0.02; P = .93) at the time of ECG. To further adjust for any effects of electrolyte concentrations on QTc, we conducted a multiple linear regression analysis that included the initial anion gap and the serum potassium and calcium concentrations at the time of ECG as covariates. In this analysis, only the initial anion gap was significantly associated with QTc (P = .04) (Table 3).

Table 3. 
Image not available
Multivariate Analysis of Biochemical Factors Associated With QTc Prolongation During DKA

The patients' heart rates were on average more rapid during DKA than after recovery. Because some studies suggest that formulas used to calculate the QTc are inexact,13,14 we conducted further analyses to assess any residual correlation between heart rate and QTc. In univariate analyses, heart rate was not significantly correlated with QTc during DKA (correlation coefficient, 0.23; P = .25) or after recovery (correlation coefficient, 0.32; P = .11). To further adjust for any effect of heart rate on QTc not captured in the formulas, we conducted a multiple linear regression analysis that included heart rate, age, and anion gap as covariates, with QTc duration as the outcome variable. The anion gap remained significantly associated with QTc (P = .04), but heart rate was not (P = .13), suggesting that QTc measurements and their association with the anion gap were not significantly biased by heart rate differences.


Prolongation of the QT interval is a serious condition that provides the substrate for development of the potentially life-threatening arrhythmia torsade de pointes.15 In the present study, we demonstrate that QTc prolongation is present in almost half of children with DKA. After recovery from the DKA episode, QTc was documented to be normal in all but 2 patients (who were lost to follow-up). Most patients had no evidence of electrolyte abnormalities or hypoglycemia to account for the observed QTc prolongation, suggesting that ketoacidosis per se might lead to delayed cardiac repolarization.

Arrhythmia and cardiac arrest are rare complications of DKA and generally have been presumed to be caused by electrolyte abnormalities.1,2 However, the QTc prolongation observed in the present study might also play a role in these complications, or QTc prolongation due to ketoacidosis might make children more prone to arrhythmias when hypokalemia, hypocalcemia, or hypomagnesemia are superimposed. In addition, sudden death during sleep in children and young adults with type 1 diabetes mellitus (“dead in bed” syndrome) has been described. This phenomenon has been attributed to hypoglycemia leading to prolonged QTc16,17; however, data from the present study raise the question of whether ketosis might also be a factor involved in some of these deaths.

In addition to the present data, studies of other ketotic conditions provide evidence supporting the association between ketosis and QTc prolongation. Deaths have been reported in patients receiving ketogenic diets for the treatment of seizure disorders, but the causes of these deaths have been unclear.18 One previous case series showed QTc prolongation in 3 of 20 children adhering to a ketogenic diet for management of seizures.4 The QTc measurements in the 20 children enrolled in that study were found to correlate with β-hydroxybutyrate and serum bicarbonate concentrations. Sudden death has also been described in adults with alcoholic ketoacidosis.9,19 Although these deaths are generally attributed to hypoglycemia, the present data raise the question of whether ketosis might also play a role. Finally, QTc prolongation and/or sudden death have been documented in adults and adolescents adhering to very-low-calorie, carbohydrate-restricted diets for weight loss.58 Many of these patients had no electrolyte abnormalities to account for QTc prolongation or arrhythmias, and sudden death has been attributed to myocardial atrophy caused by protein and micronutrient deficiencies and to other causes. Our study raises the question of whether ketosis may also play a role in these deaths.

Children with long-standing diabetes but without DKA have been shown to have a longer QTc and a greater frequency of other abnormalities in cardiac autonomic function compared with age-matched control subjects.2022 These abnormalities are thought to be manifestations of diabetic neuropathy. In the present study, QTc prolongation could not be attributed to preexisting manifestations of diabetic neuropathy because the QTc returned to normal after recovery from DKA in most of the children. In addition, the frequency of QTc prolongation was equivalent in patients with new-onset diabetes and those with known diabetes.

In the study cited previously4 describing children receiving a ketogenic diet, QTc measurements correlated with β-hydroxybutyrate measurements and serum bicarbonate concentrations, suggesting a direct relationship between ketosis and QTc prolongation. In the present study, serum ketone concentrations were not measured; however, the anion gap was significantly greater at presentation in children with a prolonged QTc. In addition, there was a significant positive correlation between QTc values during DKA and the initial anion gap. In the setting of DKA, the anion gap is mainly a reflection of ketosis.23 Therefore, these findings further support the possibility that ketosis may have a direct effect on the myocardium, causing QTc prolongation.

The anion gap at the time of ECG recording was not significantly associated with QTc, despite the very strong association between the initial anion gap and QTc. The reason for these findings is uncertain; however, it is possible that the effects of ketosis may persist for some time after serum ketone concentrations decrease. Alternatively, if β-hydroxybutyrate is primarily responsible for the observed changes in QTc, the association between the anion gap and QTc might be eliminated during therapy for DKA as β-hydroxybutyrate is converted to acetoacetate.

This study has several limitations. Serum β-hydroxybutyrate and acetoacetate concentrations were not measured, and therefore direct correlations between serum ketone concentrations and QTc prolongation could not be verified. Although the anion gap in DKA is indicative of ketosis, it does not differentiate between β-hydroxybutyrate and acetoacetate concentrations. In addition, although the present study shows an association between ketosis and QTc prolongation, it does not provide any insights into the pathogenesis of QTc prolongation in this setting. Although we did not detect any association between serum electrolyte concentrations and QTc, variations in intracellular ion concentrations caused by DKA may play a role. Similarly, we cannot exclude the possibility that other physiological alterations associated with DKA may play a role, such as central nervous system dysfunction caused by DKA or alterations in sympathetic tone. Finally, the patients' heart rates during the 2 ECG evaluations were usually not equivalent, and many patients had tachycardia during DKA. It has been suggested that currently available formulas for correcting QT interval measurements for heart rate are imperfect and that some residual correlation between heart rate and QT interval may persist, despite the application of formulas for correction.13,14 In addition, there is disagreement regarding the precise cutoff value defining QTc prolongation.24,25 It is possible, therefore, that the proportion of patients with QTc prolongation during DKA may be somewhat lower or higher than that documented in our study. It is highly unlikely, however, that tachycardia during DKA is the sole or even a main factor responsible for QTc prolongation during DKA because subanalyses did not demonstrate a significant correlation between heart rate and QTc and because multivariate analyses demonstrated a persistent significant association between the anion gap and QTc prolongation, after adjusting for age and heart rate.

The present study has several possible implications. Although it is not certain that fatal events during DKA are caused by QTc prolongation, our findings suggest that cardiac monitoring of children with DKA is important, and investigation of QTc in other ketotic conditions should be considered. Current consensus statements recommending continuous cardiac monitoring of children with DKA should be strictly followed.26 In addition, the absence of ketosis should be ensured when using QTc to evaluate autonomic dysfunction as a manifestation of diabetic neuropathy. Finally, several very low-carbohydrate diets that have recently been popularized for weight loss often induce ketosis.27 These diets have been recommended for children with severe obesity.28 Further study is necessary to determine whether individuals with an underlying genetic predisposition toward delayed cardiac repolarization may have greater prolongation of QTc when adhering to such diets or whether healthy individuals who become more markedly ketotic while following these diets may have prolonged QTc.

Back to top
Article Information

Correspondence: Nathan Kuppermann, MD, MPH, Department of Emergency Medicine, University of California, Davis, School of Medicine, 2315 Stockton Blvd, PSSB Bldg, Ste 2100, Sacramento, CA 95817 (

Accepted for Publication: November 27, 2007.

Author Contributions: Drs Kuppermann and Glaser had full access to the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Kuppermann and Glaser. Acquisition of data: Kuppermann, Park, Glatter, Marcin, and Glaser. Analysis and interpretation of data: Kuppermann and Glaser. Drafting of the manuscript: Kuppermann and Glaser. Critical revision of the manuscript for important intellectual content: Park, Glatter, and Marcin. Statistical analysis: Kuppermann and Glaser. Obtained funding: Glaser. Administrative, technical, and material support: Park and Glatter. Study supervision: Kuppermann, Marcin, and Glaser.

Financial Disclosure: None reported.

Funding/Support: This study was supported by an American Diabetes Association Clinical Research Award and a University of California, Davis, Health System Research Award (both to Dr Glaser).

Role of the Sponsors: The study sponsors had no role in determining the study design, conduct, data collection, data interpretation, manuscript preparation, or any other aspect of the study.

Glaser  NBarnett  P McCaslin  I  et al. Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics, Risk factors for cerebral edema in children with diabetic ketoacidosis. N Engl J Med 2001;344 (4) 264- 269
McMullen  JK Asystole and hypomagnesaemia during recovery from diabetic ketoacidosis. Br Med J 1977;1 (6062) 690
McCormack  P Cardiac arrest complicating diabetes. Lancet 1977;2 (8028) 87- 88
Best  THFranz  DNGilbert  DLNelson  DPEpstein  MR Cardiac complications in pediatric patients on the ketogenic diet. Neurology 2000;54 (12) 2328- 2330
Drott  CLundholm  K Cardiac effects of caloric restriction: mechanisms and potential hazards. Int J Obes Relat Metab Disord 1992;16 (7) 481- 486
Fisler  JS Cardiac effects of starvation and semistarvation diets: safety and mechanisms of action. Am J Clin Nutr 1992;56 (1) ((suppl)) 230S- 234S
Sours  HEFrattali  VBrand  C  et al.  Sudden death associated with very low calorie weight reduction regimens. Am J Clin Nutr 1981;34 (4) 453- 461
Stevens  ARobinson  PTurpin  JGroshong  TTobias  J Sudden cardiac death of an adolescent during dieting. South Med J 2002;95 (9) 1047- 1049
Denmark  LN The investigation of beta-hydroxybutyrate as a marker for sudden death due to hypoglycemia in alcoholics. Forensic Sci Int 1993;62 (3) 225- 232
Bazett  H An analysis of the time relationships of electrocardiograms. Heart 1920;7353- 370
Viitasalo  MOikarinen  LSwan  H  et al.  Ratio of late to early T-wave peak amplitude in 24-h electrocardiographic recordings as indicator of symptom history in patients with long-QT syndrome types 1 and 2. J Am Coll Cardiol 2006;47 (1) 112- 120
Landis  JRKoch  GG The measurement of observer agreement for categorical data. Biometrics 1977;33 (1) 159- 174
Luo  SMichler  KJohnston  PMacfarlane  P A comparison of commonly used QT correction formulae: the effect of heart rate on the QTc of normal ECGs. J Electrocardiol 2004;37 ((suppl)) 81- 90
Malik  MFarbom  PBatchvarov  VHnatkova  KCamm  A Relation between QT and RR intervals is highly individual among healthy subjects: implications for heart rate correction of the QT interval. Heart 2002;87 (3) 220- 228
Roden  DM Drug-induced prolongation of the QT interval. N Engl J Med 2004;350 (10) 1013- 1022
Heller  SR Abnormalities of the electrocardiogram during hypoglycaemia: the cause of the dead in bed syndrome? Int J Clin Pract Suppl 2002;129 (129) 27- 32
Robinson  RTHarris  NDIreland  RHLee  SNewman  CHeller  SR Mechanisms of abnormal cardiac repolarization during insulin-induced hypoglycemia. Diabetes 2003;52 (6) 1469- 1474
Freeman  JMKossoff  EHHartman  AL The ketogenic diet: one decade later. Pediatrics 2007;119 (3) 535- 543
Brinkmann  BFechner  GKarger  BDuChesne  A Ketoacidosis and lactic acidosis: frequent causes of death in chronic alcoholics? Int J Legal Med 1998;111 (3) 115- 119
Massin  MMDerkenne  BTallsund  M  et al.  Cardiac autonomic dysfunction in diabetic children. Diabetes Care 1999;22 (11) 1845- 1850
Suys  BEHuybrechts  SJDe Wolfe  D  et al.  QTc interval prolongation and QTc dispersion in children and adolescents with type 1 diabetes. J Pediatr 2002;141 (1) 59- 63
Verrotti  AChiarelli  FMorgese  G Autonomic dysfunction in newly diagnosed insulin-dependent diabetes mellitus children. Pediatr Neurol 1996;14 (1) 49- 52
Taboulet  PHaas  LPorcher  R  et al.  Urinary acetoacetate or capillary β-hydroxybutyrate for the diagnosis of ketoacidosis in the emergency department setting. Eur J Emerg Med 2004;11 (5) 251- 258
Benatar  ADecraene  T Comparison of formulae for heart rate correction of QT interval in exercise ECGs from healthy children. Heart 2001;86 (2) 199- 202
Mason  JWRamseth  DJChanter  DOMoon  TEGoodman  DBMendzelevski  B Electrocardiographic reference ranges derived from 79 743 ambulatory subjects. J Electrocardiol 2007;40 (3) 228- 234
Dunger  DBSperling  MAAcerini  CL  et al. European Society for Paediatric Endocrinology; Lawson Wilkins Pediatric Endocrine Society, European Society for Paediatric Endocrinology/Lawson Wilkins Pediatric Endocrine Society consensus statement on diabetic ketoacidosis in children and adolescents. Pediatrics 2004;113 (2) e133- e140 Accessed March 24, 2008
Gardner  CDKiazand  AAlhassan  S  et al.  Comparison of the Atkins, Zone, Ornish, and LEARN diets for change in weight and related risk factors among overweight premenopausal women: the A to Z Weight Loss Study: a randomized trial [published correction appears in JAMA. 2007;298(2):178]. JAMA 2007;297 (9) 969- 977
Willi  SMMartin  KDatko  FMBrant  BP Treatment of type 2 diabetes in childhood using a very-low-calorie diet. Diabetes Care 2004;27 (2) 348- 353