Volume 31 • Number 4 • April 1998
Copyright © 1998 American College of Emergency Physicians
Mark A Brandenburg MD
Daniel J Dire MD
From the Section of Emergency Medicine & Trauma, University of Oklahoma Health Sciences Center, Oklahoma City, OK.
Study objective: To determine whether venous blood gas values can replace arterial gas values in the initial emergency department evaluation of patients with suspected diabetic ketoacidosis.
Methods: This prospective comparison was performed in an adult university teaching hospital ED. Samples for arterial and venous blood gas analysis were obtained during initial ED evaluations. The venous gas samples were collected with samples for other blood tests at the time of intravenous line insertion. Both arterial and venous samples were obtained before the initiation of treatment.
Result: Data from 44 episodes of diabetic ketoacidosis in 38 patients were analyzed. Laboratory findings of those patients with diabetic ketoacidosis were as follows (mean±SD): arterial pH, 7.20±.14; venous pH, 7.17±.13; serum glucose, 33.8±16 mmol/L (609±288 mg/dL); arterial HCO3 - , 11.0±6.0 mmol; venous HCO3 - , 12.8±5.5 mmol/L; serum CO2 , 11.8±5.0 mmol/L; and anion gap, 26.7±7.6 mmol/L. The mean difference between arterial and venous pH values was 0.03 (range 0.0 to 0.11). Arterial and venous pH results ( r=.9689) and arterial and venous HCO3 - results ( r=.9543) were highly correlated and showed a high measure of agreement.
Conclusion: Venous blood gas measurements accurately demonstrate the degree of acidosis of adult ED patients presenting with diabetic ketoacidosis.
[Brandenburg MA, Dire DH: Comparison of arterial and venous blood gas values in the initial emergency department evaluation of patients with diabetic ketoacidosis. Ann Emerg Med April 1998; 31:459-465.]
Received for publication September 26, 1996.
Revisions received May 20, 1997, and October 22, 1997.
Accepted for publication October 29, 1997.
Presented in part at the Society for Academic Emergency Medicine Annual Meeting, Denver, CO,
Copyright © 1998 by the American College of Emergency Physicians.
Diabetic ketoacidosis (DKA) is a metabolic derangement consisting of high blood glucose concentration, measurable ketone bodies, and metabolic acidosis, and is an endocrinologic emergency responsible for approximately 110,000 hospital admissions in the United States annually.  DKA It is the most common serious and life-threatening acute complication of diabetes. The mortality rate is currently estimated at 2% to 10% for patients hospitalized with DKA.   
Determination of arterial blood gas values is currently considered essential in the emergency department evaluation of patients with suspected DKA.   Sampling of arterial blood is a painful, sometimes technically difficult procedure that must be done in addition to the sampling of venous blood for testing for electrolytes, ketones, and other parameters.
Several authors have recommended the use of venous pH in the evaluation of DKA.   The correlation between arterial and venous pH measurements is well established         ; however, this relationship has not been established in DKA. If venous pH were found to be highly correlated and show a high level of agreement with arterial pH in patients who present to the ED with DKA, then it might be possible to eliminate arterial blood sampling in the initial diagnosis and evaluation of DKA.
The purpose of this study was to determine whether venous blood gas values are correlated and show a high level of agreement with arterial gas values in the initial ED evaluation of patients with
Patients presenting to the ED at University Hospital in Oklahoma City, the adult teaching hospital for the University of Oklahoma, College of Medicine, were eligible for this study if they had a fingerstick bedside blood glucose measurement in excess of 13.9 mmol/L (>250 mg/dL), urine dipstick result positive for ketones, and the emergency physician clinically suspected DKA before the results of further laboratory tests were known. This protocol was reviewed by the Office of Research Administration, University of Oklahoma Health Sciences Center and was determined to meet criteria for exemption from institutional review board review.
A single sample of arterial blood (.5 to 1.0 mL) was obtained from the radial artery of each patient by an ED staff member or by one of the investigators. A single (.5 to 1.0 mL) sample of peripheral venous blood was obtained at the time of intravenous line placement or peripheral venipuncture for laboratory testing and was used for the venous blood gas analysis. The two separate blood gas samples were obtained as temporally close to each other as possible and before the initiation of treatment with intravenous fluids and insulin, and were transported to the laboratory on ice. The arterial and venous blood gas determinations were performed with an AVL-995 (AVL Scientific Corporation, Roswell, GA) or GEM Premier-5300 analyzer (Instrumentation LABs, Ann Arbor, MI), and the serum chemistry tests were performed on a Hitachi 747 Automatic Analyzer (Boehringer Mannheim Corporation, Indianapolis, IN) or a Beckman CX3 analyzer (Beckman Instruments, Inc, Brea, CA). Serum ketones were measured with Ketostix Reagent Strips (Bayer Corporation, Elkhart, IN) in the hospital laboratory.
TABLE -- Descriptive statistics.
Variable Mean SD Range
Serum glucose 33.8 16.0 13.9-81.3
(mmol/L) [mg/dL] [609.8]  [250-1,464]
Arterial pH 7.20 0.14 6.78-7.39
Venous pH 7.17 0.13 6.89-7.38
Arterial HCO3 - (mmol/L) 11.0 6.0 2-23
Venous HCO3 - (mmol/L) 12.8 5.5 3-24
Serum CO2 (mmol/L) * 11.8 5.0 2-24
Anion gap (mmol/L) * 26.7 7.6 15-47
Serum sodium (mmol/L) * 134.2 4.5 121-143
Serum potassium (mmol/L) * 5.1 0.9 3.6-8.8
Serum chloride (mmol/L) * 95.4 7.6 68-109
*From venous specimens.
A patient was considered to have DKA if the arterial pH was less than 7.35 or the serum CO2 was less than 20 mmol/L, the serum glucose was less than 13.9 mmol/L (>250 mg/dL), and results for serum ketones were positive. Patients without DKA were subsequently excluded. Demographic and laboratory data were recorded on a database form initiated in the ED and subsequently entered into a computer for statistical analysis using Microsoft Excel for Windows (version 5.0c, 1985-1994, Microsoft Corporation) and True Epistat (version 4.01, 1987, 1991, Epistat Services).
The strength of association between arterial and venous pH, arterial and venous HCO3 - , and arterial HCO3 - and venous serum CO2 results was assessed with Pearson's correlation coefficient ( r). The coefficient of determination ( r2 ) was used to measure the proportion of the variance in arterial levels that could be "accounted for" by the venous levels using a linear model. The degree of agreement between the arterial and venous pH measurements, arterial and venous HCO3 - values, and the arterial HCO3 - and venous serum CO2 values was evaluated by plotting the difference between the paired determinations against the mean of any two determinations, as described by Bland and Altman. 
This type of plot is bounded by "limits of agreement," defined as the mean of the arteriovenous differences ±2 SD.
Because the intent of our study was to establish the utility of the venous blood gas pH (as the independent variable) as a predictor of arterial blood gas pH (the dependent variable), a linear equation was developed to estimate the mean value of the dependent variable (arterial pH) for each value of the independent variable (venous pH). The slope of the regression line indicates the amount the mean of the dependent variable changes for each unit change in the numerical value of the independent variable. To examine the amount of variation in the dependent variable that cannot be explained with the linear equation, we tested the null hypothesis that the regression equation does not allow us to explain the value of the dependent variable given a value of the independent variable. We used the F distribution to test this hypothesis with an alpha level of .05. If the P value (reported with the regression equations) is significant, we reject the null hypothesis and conclude that the independent variable can be used to provide an accurate estimate of the dependent variable.
A convenience sample of 61 patients with 68 suspected episodes of DKA was enrolled into this study during a 14-month period (no patient was enrolled more than twice). Data on 23 patients with 24 episodes were omitted from analysis for the following reasons: patient did not have acidosis, 14; serum ketone results were negative, 5; no venous blood gas analysis was done, 2; no arterial blood gas analysis was done, 1; no serum ketone analyses were done, 1; and patient was not hyperglycemic, 1. This left 44 episodes of DKA involving 38 patients with a mean (±SD) age of 35.4±12.9 years (range 12 to 69 years). There were 20 men and 18 women.
The main results are summarized in the Table. The mean difference between arterial and venous pH values is .03 (range 0 to .11). Arterial and venous pH results ( r=.9689, r2 =.94, Figure 1 ), arterial and venous HCO3 - results ( r= .9543, r2 =.91, Figure 2 ), and arterial HCO3 - and serum CO2 results ( r=.8985, r2 =.81, Figure 3 ) demonstrate strong correlations.
Graphic depictions of the scatter of data and the analysis of agreement for arterial and venous pH measurements, arterial and venous HCO3 - values, and arterial HCO3 – and serum CO2 values are shown in Figures 4 , 5 , and 6 , respectively.
At present, there is no consensus definition in the medical literature regarding the diagnostic criteria for DKA.  Our study criteria are most similar to the diagnostic guidelines used by Fleckman.  
In DKA, mortality is related to the age and state of consciousness of the patient, and the degree of acidosis, hyperglycemia, and azotemia.  Determination of acid-base status, therefore, has prognostic as well as diagnostic value. To clearly characterize the acidosis, a measurement of the arterial P co2 or pH is needed.   
In 1961 Gambino  compared "arterialized" capillary blood with samples from the brachial artery. The capillary blood was collected from the ear or finger after heating to 40 to 45° C. He found no significant difference in pH or CO2 content in 13 patients undergoing routine pulmonary function testing. Although Gambino did not calculate the strength of association for his findings, we calculated a Pearson's correlation coefficient of .9 for both his pH and his CO2 measurements. Subsequently, there have been many studies on the concordance between arterial and capillary blood with respect to acid-base parameters.         
In 1988 Hale and Nattrass  compared arterial and "nonarterialized" capillary blood gases in 20 patients presenting in DKA. They found that the arterial and capillary pH and bicarbonate levels were strongly correlated ( r=.98 and .97, respectively). The differences in pH measurements ranged from -.09 to +0.02. They concluded that capillary samples were a reliable indicator of the acid-base status in patients with DKA and were preferable to repeated arterial puncture.
We could find no previous studies that specifically compared arterial and venous blood gases in patients with DKA. However, in 1948 Kety et al  reported on the blood flow and oxygen consumption of the human brain in diabetic acidosis and coma. In their report, they presented data on 14 patients with "severe" DKA who had both arterial and central venous blood gases (samples obtained from the internal jugular vein) that were obtained before the initiation of treatment. Although they did not perform any statistical comparisons, we determined that there was a strong correlation between their two sets of samples ( r=.99).
Our study is the first report comparing peripheral venous blood gas values with arterial blood gas values in patients with DKA. We did not specifically measure or control for time that the tourniquets were in place on the extremity during venipuncture. A recent study showed that the use of a tourniquet over a clinically relevant time (3 minutes) does not increase venous lactate levels. 
Our goal was to determine whether the two paired pH measurements agree sufficiently to allow one to be substituted for the other in patients with DKA. The Pearson's correlation coefficient is frequently used to demonstrate the linear relationship between two continuous variables, but does not necessarily establish that they are in agreement. An editorial concerning correlation versus agreement was recently published in this journal.  The analysis of agreement, as described by Bland and Altman,  uses a plot of the differences between the two measurements against the average of the two measurements. The mean of the differences is shown by a horizontal "line of agreement" bounded above and below by two parallel lines containing 95% of the differences between the two measurements. As seen in Figure 4 , the small range of pH differences inside these two boundaries is not clinically important, and in 41 cases (93%), the venous pH was equal to or greater than the arterial pH. In 40 cases (91%), the values fell within the boundaries of agreement. Regardless of the plotted boundaries, in only 1 case (2%) did the venous pH differ from the arterial pH by greater than 0.1 (the difference in this case was .11, which had no diagnostic or therapeutic implications and did not change the patient's outcome). Therefore we conclude that the peripheral venous pH measurement is a valid and reliable substitute for arterial pH. Similarly, the venous HCO3 - measurement could replace the arterial HCO3 - . However, the HCO3 - measurement that was reported by our hospital laboratory is not a true measurement, but rather a calculated value. Therefore we chose to use the venous serum CO2 value that is actually measured.
A potential disadvantage of the use of the venous blood gas values in DKA is that it may be more difficult to determine when mixed acid-base disorders are present. Patients with DKA often have conditions such as vomiting, diarrhea, dehydration, and hyperventilation that may cause mixed acid-base disturbances.  A mixed acid-base disorder is suggested when the anion gap increase does not equal the serum bicarbonate decrease.  In patients who have respiratory depression, venous blood gas values may not accurately reflect the degree of respiratory acidosis or hypoxia, if present. Nomograms have been developed that enable the calculation of arterial P co2 after the direct measurement of pH, total CO2 of the plasma, and hemoglobin concentration.   However, the use of these nomograms needs to be validated in patients presenting with DKA. It seems reasonable to continue to perform arterial blood gas analysis in patients with DKA in whom a mixed acid-base disorder is suspected, in those with respiratory depression, and in those in whom hypoxia is suggested by abnormally low pulse-oximetry measurements.
This study is limited by the fact that subjects were taken from a convenience sample (usually when the investigators were on duty in the ED), raising the possibility of selection bias. Second, the sample size was small and may not have included the full spectrum of patients with DKA, such as those with hypotensive or hypoperfusion states. Only four patients in our study were hypotensive (systolic blood pressure less than 90 mm Hg) on arrival to the ED. Also, there were very few patients with severe acidosis (only five patients had an arterial pH<7.0). It is possible that in an occasional patient, the venous blood gas value might underestimate the degree of acidosis. A larger sample size would be needed to look specifically at these subgroups. Third, the ketone test we used reacts with acetoacetic acid and not acetone or beta-hydroxybutyric acid. Thus some excluded patients (who were excluded post hoc because they did not meet the study definition for DKA) might have had predominantly beta-hydroxybutyric acid, making the serum ketone test result falsely negative. However, when we repeated data analysis without excluding any of the 14 patients with negative serum ketone results, the results (for correlation and agreement) did not change. Finally, we did not look at the final patient outcomes (after admission from the ED); thus we cannot make any conclusions as to whether use of venous instead of arterial gas values might have any effects on outcome.
In the present study, the costs of arterial and venous blood gas analyses were the same; however, it should be noted that arterial blood gas sampling is more painful to patients and sometimes requires multiple attempts. Arterial puncture is more time consuming and labor intensive when compared with venipuncture, which must be performed anyway to measure serum electrolytes and glucose levels, as well as to establish intravenous access for fluid resuscitation and insulin administration. Further studies should be conducted to prospectively validate the use of venous blood gas values in the management of patients with DKA in the ED and subsequently in the inpatient units.
1. Wetterhall SF, Olson DR, DeStafano F, et al: Trends in diabetes and diabetic complications, 1980-1987. Diabetes Care 1992;15:960-966.
2. Lebovitz HE: Diabetic ketoacidosis. Lancet 1995;345:767-772.
3. Bagdade JP: Endocrine emergencies. Med Clin North Am 1986;70:1111-1128.
4. Kitabchi AE, Wall BM: Diabetic ketoacidosis. Med Clin North Am 1995;79:9-37.
5. Ragland G: Diabetic ketoacidosis, in Tintinalli JE, Ruiz E, Krome RL (eds): Emergency Medicine:A Comprehensive Study Guide, ed 4. New York: McGraw-Hill, 1996:947-951.
6. Becker DJ, Drash Al: Diabetic ketoacidosis, in Dickerman JD, Lucey JF (eds): Smith's: The Critically Ill Child: Diagnosis and Management, ed 3. Philadelphia: WB Saunders, 1985:113-128.
7. Jordan RM: Endocrine emergencies. Med Clin North Am 1983;67:1193-1213.
8. Brooks D, Wynn V: Use of venous blood for pH and carbon-dioxide studies. Lancet 1959;1:227-230.
9. Forster HV, Dempsey JA, Thomson J, et al: Estimation of arterial Po2 , P co2 , pH, and lactate from arterialized venous blood. J Appl Physiol 1972;32:134-137.
10. Gambino SR: Comparison of pH in human arterial, venous, and capillary blood. Am J Clin Pathol 1959;32:298-300.
11. Gilbert R: Spirometry and blood gases, in Henry JB (ed): Todd, Sanford, Davidsohn Clinical Diagnosis and Management by Laboratory Methods, ed 17. Philadelphia: WB Saunders, 1984:96-117.
12. Moore EE, Good JT: Mixed venous and arterial pH: A comparison during hemorrhagic shock and hypothermia. Ann Emerg Med 1982;11:300-302.
13. Paine EG, Boutwell JH, Soloff LA: The reliability of "arterialized" venous blood for measuring arterial pH and P co2 . Am J Med Sci 1961;242:431-435.
14. Fleisher M, Schwartz MK: Blood gas analysis and acid-base balance: Principles and technics, in Sonnenworth AC, Jarett L (eds): Gradwohl's Clinical Laboratory Methods and Diagnoses, ed 8. St Louis: CV Mosby, 1980:351-366.
15. Sutton RN, Wilson RF, Walt AJ: Differences in acid-base levels and oxygen-saturation between central venous and arterial blood. Lancet 1967;2:748-751.
16. Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurements. Lancet 1986;1:307-310.
17. Fleckman AM: Diabetic ketoacidosis. Practical Diabetology 1991;10:1-8.
18. Fleckman AM: Diabetic ketoacidosis. Endocrinol Metab Clin North Am 1993;22:181-207
19. Kreisberg RA: Diabetic ketoacidosis: New concepts and trends in pathogenesis and treatment. Ann Intern Med 1978;88:681-695.
20. King AJ, Cooke NJ, McCuish A, et al: Acid-base changes during treatment of diabetic ketoacidosis. Lancet 1974;1:478-481.
21. Singer RB, Hastings AB: An improved clinicical method for the estimation of disturbances of the acid-base balance of human blood. Medicine 1948;27:223-242.
22. Schwartz WB, Relman AS: A critique of the parameters used in the evaluation of acid-base disorders. N Engl J Med 1963;268:1382-1388.
23. Gambino SR: Collection of capillary blood for simultaneous determinations of arterial pH, CO2 content, P co2 , and oxygen saturation. Am J Clin Pathol 1961;35:175-183.
24. Howland WS, Scweizer O, Murphy TW: Estimation of acid-base of venous and arterial blood from capillary samples. Acta Anaesthesiol Scand 1964;8:191-196.
25. Jung RC, Balchum OJ, Massey FJ: The accuracy of venous and capillary blood for the prediction of arterial pH, P co2 , and Po2 measurement. Am J Clin Pathol 1966;45:129-138.
26. Knudsen EJ, Hansen P: Carbon dioxide tensions in nonarterialzed capillary and arterial blood during anaesthesia. Acta Anaesthesiol Scand 1962;6:29-35.
27. Koch G: Comparison of carbon dioxide tension, pH and standard bicarbonate in capillary blood and in arterial blood with special respect to relations in patients with impaired cardiovascular and pulmonary function and during excercise. Scand J Clin Lab Invest 1965;17:223-229.
28. Koch G, Wendel H: Comparison of pH, carbon dioxide tension, standard bicarbonate, and oxygen tension in capillary blood and in arterial blood during the neonatal period. Acta Paediatr Scand 1967;56:10-16.
29. Langlands JHM, Wallace WFM: Small blood samples from earlobe puncture. A substitute for arterial puncture. Lancet 1965;11:315-317.
30. Maas AHJ, Van Heijst ANP: A comparison of the pH of arterial blood with arterialized blood from the ear-lobe with Astrup's micro glass electrode. Clin Chim Acta 1961;6:31-33.
31. Maas AHJ, Van Heijst ANP: The accuracy of the microdetemination of the P co2 of blood from the ear-lobe. Clin Chim Acta 1961;6:34-37.
32. Torjussen W, Nitter-Hauge S: pH and P co2 in capillary blood from the earlobe. Scand J Clin Lab Invest 1967;19:79-81.
33. Hale PJ, Nattrass M: A comparison of arterial and nonarterialized capillary blood gases in diabetic ketoacidosis. Diabetes Medicine 1988;5:76-78.
34. Kety SS, Polis BD, Nadler CS, et al: The blood flow and oxygen consumption of the human brain in diabetic acidosis and coma. J Clin Invest 1948;27:500-510.
35. Tortella BJ, Ferguson KP, Lavery RF, et al: Use of a venous tourniquet does not increase venous lactate. Acad Emerg Med 1996;3:531-532.
36. Gallagher EJ: Correlation versus agreement: Methods of measurement in medicine. Ann Emerg Med 1996;27:236-238.
37. Zonszein J, Baylor P: Diabetic ketoacidosis with alkalemia--a review. West J Med 1988;149: 217-219.
38. Paulson WD, Gadallah MF: Diagnosis of mixed acid-base disorders in diabetic ketoacidosis. Am J Med Sci 1993;306:295-300.
39. Siggaard-Andersen O: Blood acid base alignment nomogram. Scand J Clin Lab Invest 1963; 15:211-217.