470
ANESTH ANALG 1992;75:461-71
LETTERS TO THE EDITOR
tions, medians, ranges, and percentiles. Inferential statistics include standard errors (SEs), confidence intervals (CIS),and various tests that yield P values for differences in means, proportions, and other descriptive data (1). In this communication I would like to suggest some methods to improve the quality of presenting and summarizing data in our journals. I examined the articles published in Anesthesia G. Analgesia in the year 1991 to determine how data are presented and summarized. There were 208 articles related to laboratory and clinical investigations. In 90 (43.27%) of these articles, the data were summarized as mean t SD and in 91 (43.75%)as mean SEM. The data in the rest of the articles were summarized as medians, ranges, percentages, etc. The suggestion that all journals discontinue the use of the 2 sign was made in 1986 by Altman and Gardner in a letter to the The Lancet (2). But this sign continues to appear in the Anesthesia 6 Analgesia journal regularly. It was also stated in the same letter that SEM should not be used to indicate the variability of a set of observations. As mentioned above, in 43.75% of the articles published in 1991, authors used SEM to indicate the variability. Variability is best indicated by SD or CV (SD/mean %) according to the situation (3,4). The CV is most useful as a descriptive tool in situations in which a change in the conditions under which measurements are made alters the SD in the same proportion as it alters the mean (3). Glantz suggested that SD, not SE, should be used to summarize the data (5). The fact that SE is inappropriate for indicating the variability of a sample must be clearly understood. By itself, SE provides limited information. It is used to estimate the CI, which is very informative (6,7). The CI is a range of values that is likely to cover the true, but unknown, population value. Thus SE is an inferential statistic, not a descriptive statistic (8). Confidence intervals can be computed from SEs for every sample statistic (means and their differences, proportions and their differences, correlation coefficients and regression coefficients, etc.) using the standard formulas (3,6,7). There are several advantages of presenting the data (in tables and figures) as CIS wherever appropriate. Confidence intervals provide more information than hypothesis testing by the use of P values (6). Differences of no real interest can be statistically significant with large samples, whereas clinically important effects may not be significant when the sample size is small. Confidence intervals provide useful information in studies with large as well as small samples (9) and help us to avoid some misinterpretations that result from dichotomous conclusions from the significance tests based on P values (10,ll). In addition, CIS also have a link to the outcome of hypothesis tests. For example, when the result of difference between two means is significant at the 5% level, the 95% CI for difference in their means will not encompass zero (7). Difference in proportions summarized in 2 x 2 tables are usually tested by Fisher’s exact test or ,$ analysis with Yates‘ correction. But 95% CI for difference in proportions gives more information than hypothesis testing by either of the above two methods. In addition, meaningful interpretations of results of different studies can be derived when the difference in proportions are given in 95% CIS rather than when the results are interpreted by significant testing based on
*
P values (12). Finally, according to Snedecor ”the purpose of an experiment is to produce a sample of observations which will furnish estimates of parameters of the population together with measures of uncertainty of these estimates” (13). Confidence intervals satisfy the requirements for a meaningful inference from an experiment because they indicate the degree of imprecision of the sample as an estimate of the population value. In summary, variability in a sample may be presented as SD without the use of the ? sign, or as CV according to the situation. Confidence intervals, if appropriate to the type of study, should be used for major findings in both the main text of the paper and its abstract (6). In the present day and age of rapidly growing medical literature in a plethora of journals, we generally rely on computer-assisted literature search programs to access different publications. As these services allow us to get only the abstracts, the suggestion that the more informative CIS be presented in the abstracts appears very attractive. This letter is essentially an echo of what statistical experts have been expressing during the past few years. The suggestions if followed will further improve the quality of scientific communication in Anesfhesia 6 Analgesia and in other journals of our speciality. Srinivas Mantha, MD Department of Anesthesiology Nizam’s Institute of Medical Sciences Hyderabad 500482 lndia
References 1. Feinstein AR. Demeaned errors, confidence games, nonplussed mi-
nuses, inefficient coefficients, and other statistical disruptions of scientific communication. Clin Pharm Ther 1976;65:617-31. 2. Altman DG, Gardner MJ. Presentation of variability. Lancet 1986;ii:639. 3. Armitage P, Berry G. Statistical methods in medical research. 2nd ed. Oxford: Blackwell Scientific Publications, 1987. 4. Fisher DM. Statistics in anesthesia. In: Miller RD, ed. Anesthesia. 3rd ed. New York: Churchill-Livingstone, 1990685-712. 5. Glantz SA. Biostatistics: how to detect, correct and prevent errors in the medical literature. Circulation 1980;61:1-7. 6 . Gardner MI, Altman DG. Confidence intervals rather than P values: estimation rather than hypothesis testing. Br Med J 1986;292:74&50. 7. Bulpitt CJ. Confidence intervals. Lancet 1987;i:494-7. 8. Brown GW. Standard deviation, standard error. Which ‘standard’ should we use? Am J Dis Child 1982;136:937-41. 9. Braitman LE. Confidence intervals extract clinically useful information from data. Ann Intern Med 1988;108:296-8. in. Rothman KJ. A show of confidence, N Engl J Med 1978;299:1362-3. 11. Simon R. Confidence intervals for reporting results of clinical trials. Ann Intern Med 1986;105:429-35. 12. Rothman KI.Sianificance auestine. Ann Intern Med 1986:105:445-7. 13. Snedecor GW. f h e statistical part gf the scientificmethod. Ann NY Acad scl 1950;52:792-9.
Esterase Nomenclature: A Confusing Topic To the Editor: The classification of cholinesterases is so unclear that even leading anesthesiology textbooks carry errors in their description. Page 473 of Clinical Anesthesia edited by Barash, Cullen, and Stoelting states that “pseudocholinesterase degrades acetylcholine released at the neuromuscular junction” (1). Instead, it should read ”acetylcholinesterase” or
ANESTH ANALG 1992;75:461-71
LETTERS TO THE EDITOR
471
any of the other multiple terms used to designate this enzyme. This letter is not to point out a typographical mistake but to illustrate that esterase nomenclature is very confusing. True cholinesterase (E.C.3.1.1.7) is also known as choline esterase I or specific, cholinesterase, acetylcholine hydrolase, acetylcholinesterase, or erythrocyte cholinesterase. Whereas pseudocholinesterase (E.C. 1.1.8)is also known as choline esterase I1 or unspecific, butyrylcholinesterase, benzoylcholinesterase, acylcholine acyl-hydrolase, serum cholinesterase, or plasma cholinesterase (2,3). Mariano Mikulic, MD Rafael Ortega, MD Department of Anesthesiology Boston University Medical Center 88 East Newton Street Boston, M A 02130
References Rosenberg H, Seitman D. Pharmacogenetics. In: Barash P, Cullen 8 , Stoelting R, eds. Clinical anesthesia. Philadelphia: JB Lippincott, 1989: 459-83. Whittaker M. Plasma cholinesterase variants and the anaesthetist. Anaesthesia 1980;35:174-97. International Union of Biochemistry. Enzyme nomenclature. Recommendations 1964, of the International Union of Biochemistry. Amsterdam: Elsevier. 1965.
A Simple Replacement for the Pulmonary Artery Catheter Balloon Syringe To the Editor: During the use of a pulmonary artery catheter, the 3.0-mL balloon inflation syringe may be misplaced. This may necessitate replacement with a regular 3.0-mL syringe and the expense of opening a new kit to obtain the special syringe. Manufacturers recommend that no more than 1.5 mL of air be used to inflate the pulmonary artery catheter balloon. If too much air is injected, it can cause balloon or pulmonary artery rupture. We describe a simple
Figure 1. Pulsator blood gas syringe with the 16-gauge needle at the 1.5-mL mark.
and easily accessible method to prevent excessive air injection. Using a 3-mL pulsator blood gas syringe, the plunger is withdrawn and a 16-gauge needle is used to pierce two holes through the cylinder of the syringe at the 1.5-mL mark (Figure 1). The 16-gauge needle is then removed. These holes will permit balloon inflation with 1.5 mL of air, but will prevent overinflation as there will be an air leak above the 1.5-mL mark. Because this is a closed system, sepsis is not a concern. Like the original syringe, this one will also allow passive deflation rather than active deflation, which may cause balloon rupture. Jorge R. Villarreal, MD Devanand Mangar, MD Department of Anesthesiology University of South Florida College of Medicine MDC Box 59 12901 Bruce B. Downs Boulevard Tampa, F L 33612
ANESTH ANALG 1992;75:461-71
LETTERS TO THE EDITOR
tions, medians, ranges, and percentiles. Inferential statistics include standard errors (SEs), confidence intervals (CIS),and various tests that yield P values for differences in means, proportions, and other descriptive data (1). In this communication I would like to suggest some methods to improve the quality of presenting and summarizing data in our journals. I examined the articles published in Anesthesia G. Analgesia in the year 1991 to determine how data are presented and summarized. There were 208 articles related to laboratory and clinical investigations. In 90 (43.27%) of these articles, the data were summarized as mean t SD and in 91 (43.75%)as mean SEM. The data in the rest of the articles were summarized as medians, ranges, percentages, etc. The suggestion that all journals discontinue the use of the 2 sign was made in 1986 by Altman and Gardner in a letter to the The Lancet (2). But this sign continues to appear in the Anesthesia 6 Analgesia journal regularly. It was also stated in the same letter that SEM should not be used to indicate the variability of a set of observations. As mentioned above, in 43.75% of the articles published in 1991, authors used SEM to indicate the variability. Variability is best indicated by SD or CV (SD/mean %) according to the situation (3,4). The CV is most useful as a descriptive tool in situations in which a change in the conditions under which measurements are made alters the SD in the same proportion as it alters the mean (3). Glantz suggested that SD, not SE, should be used to summarize the data (5). The fact that SE is inappropriate for indicating the variability of a sample must be clearly understood. By itself, SE provides limited information. It is used to estimate the CI, which is very informative (6,7). The CI is a range of values that is likely to cover the true, but unknown, population value. Thus SE is an inferential statistic, not a descriptive statistic (8). Confidence intervals can be computed from SEs for every sample statistic (means and their differences, proportions and their differences, correlation coefficients and regression coefficients, etc.) using the standard formulas (3,6,7). There are several advantages of presenting the data (in tables and figures) as CIS wherever appropriate. Confidence intervals provide more information than hypothesis testing by the use of P values (6). Differences of no real interest can be statistically significant with large samples, whereas clinically important effects may not be significant when the sample size is small. Confidence intervals provide useful information in studies with large as well as small samples (9) and help us to avoid some misinterpretations that result from dichotomous conclusions from the significance tests based on P values (10,ll). In addition, CIS also have a link to the outcome of hypothesis tests. For example, when the result of difference between two means is significant at the 5% level, the 95% CI for difference in their means will not encompass zero (7). Difference in proportions summarized in 2 x 2 tables are usually tested by Fisher’s exact test or ,$ analysis with Yates‘ correction. But 95% CI for difference in proportions gives more information than hypothesis testing by either of the above two methods. In addition, meaningful interpretations of results of different studies can be derived when the difference in proportions are given in 95% CIS rather than when the results are interpreted by significant testing based on
*
P values (12). Finally, according to Snedecor ”the purpose of an experiment is to produce a sample of observations which will furnish estimates of parameters of the population together with measures of uncertainty of these estimates” (13). Confidence intervals satisfy the requirements for a meaningful inference from an experiment because they indicate the degree of imprecision of the sample as an estimate of the population value. In summary, variability in a sample may be presented as SD without the use of the ? sign, or as CV according to the situation. Confidence intervals, if appropriate to the type of study, should be used for major findings in both the main text of the paper and its abstract (6). In the present day and age of rapidly growing medical literature in a plethora of journals, we generally rely on computer-assisted literature search programs to access different publications. As these services allow us to get only the abstracts, the suggestion that the more informative CIS be presented in the abstracts appears very attractive. This letter is essentially an echo of what statistical experts have been expressing during the past few years. The suggestions if followed will further improve the quality of scientific communication in Anesfhesia 6 Analgesia and in other journals of our speciality. Srinivas Mantha, MD Department of Anesthesiology Nizam’s Institute of Medical Sciences Hyderabad 500482 lndia
References 1. Feinstein AR. Demeaned errors, confidence games, nonplussed mi-
nuses, inefficient coefficients, and other statistical disruptions of scientific communication. Clin Pharm Ther 1976;65:617-31. 2. Altman DG, Gardner MJ. Presentation of variability. Lancet 1986;ii:639. 3. Armitage P, Berry G. Statistical methods in medical research. 2nd ed. Oxford: Blackwell Scientific Publications, 1987. 4. Fisher DM. Statistics in anesthesia. In: Miller RD, ed. Anesthesia. 3rd ed. New York: Churchill-Livingstone, 1990685-712. 5. Glantz SA. Biostatistics: how to detect, correct and prevent errors in the medical literature. Circulation 1980;61:1-7. 6 . Gardner MI, Altman DG. Confidence intervals rather than P values: estimation rather than hypothesis testing. Br Med J 1986;292:74&50. 7. Bulpitt CJ. Confidence intervals. Lancet 1987;i:494-7. 8. Brown GW. Standard deviation, standard error. Which ‘standard’ should we use? Am J Dis Child 1982;136:937-41. 9. Braitman LE. Confidence intervals extract clinically useful information from data. Ann Intern Med 1988;108:296-8. in. Rothman KJ. A show of confidence, N Engl J Med 1978;299:1362-3. 11. Simon R. Confidence intervals for reporting results of clinical trials. Ann Intern Med 1986;105:429-35. 12. Rothman KI.Sianificance auestine. Ann Intern Med 1986:105:445-7. 13. Snedecor GW. f h e statistical part gf the scientificmethod. Ann NY Acad scl 1950;52:792-9.
Esterase Nomenclature: A Confusing Topic To the Editor: The classification of cholinesterases is so unclear that even leading anesthesiology textbooks carry errors in their description. Page 473 of Clinical Anesthesia edited by Barash, Cullen, and Stoelting states that “pseudocholinesterase degrades acetylcholine released at the neuromuscular junction” (1). Instead, it should read ”acetylcholinesterase” or
ANESTH ANALG 1992;75:461-71
LETTERS TO THE EDITOR
471
any of the other multiple terms used to designate this enzyme. This letter is not to point out a typographical mistake but to illustrate that esterase nomenclature is very confusing. True cholinesterase (E.C.3.1.1.7) is also known as choline esterase I or specific, cholinesterase, acetylcholine hydrolase, acetylcholinesterase, or erythrocyte cholinesterase. Whereas pseudocholinesterase (E.C. 1.1.8)is also known as choline esterase I1 or unspecific, butyrylcholinesterase, benzoylcholinesterase, acylcholine acyl-hydrolase, serum cholinesterase, or plasma cholinesterase (2,3). Mariano Mikulic, MD Rafael Ortega, MD Department of Anesthesiology Boston University Medical Center 88 East Newton Street Boston, M A 02130
References Rosenberg H, Seitman D. Pharmacogenetics. In: Barash P, Cullen 8 , Stoelting R, eds. Clinical anesthesia. Philadelphia: JB Lippincott, 1989: 459-83. Whittaker M. Plasma cholinesterase variants and the anaesthetist. Anaesthesia 1980;35:174-97. International Union of Biochemistry. Enzyme nomenclature. Recommendations 1964, of the International Union of Biochemistry. Amsterdam: Elsevier. 1965.
A Simple Replacement for the Pulmonary Artery Catheter Balloon Syringe To the Editor: During the use of a pulmonary artery catheter, the 3.0-mL balloon inflation syringe may be misplaced. This may necessitate replacement with a regular 3.0-mL syringe and the expense of opening a new kit to obtain the special syringe. Manufacturers recommend that no more than 1.5 mL of air be used to inflate the pulmonary artery catheter balloon. If too much air is injected, it can cause balloon or pulmonary artery rupture. We describe a simple
Figure 1. Pulsator blood gas syringe with the 16-gauge needle at the 1.5-mL mark.
and easily accessible method to prevent excessive air injection. Using a 3-mL pulsator blood gas syringe, the plunger is withdrawn and a 16-gauge needle is used to pierce two holes through the cylinder of the syringe at the 1.5-mL mark (Figure 1). The 16-gauge needle is then removed. These holes will permit balloon inflation with 1.5 mL of air, but will prevent overinflation as there will be an air leak above the 1.5-mL mark. Because this is a closed system, sepsis is not a concern. Like the original syringe, this one will also allow passive deflation rather than active deflation, which may cause balloon rupture. Jorge R. Villarreal, MD Devanand Mangar, MD Department of Anesthesiology University of South Florida College of Medicine MDC Box 59 12901 Bruce B. Downs Boulevard Tampa, F L 33612
