In a previous communication the results of duplicate estimations of cardiac output on 50 persons, the majority hospital patients in the basal condition, have been ret)orted (1). The opportunity to extend this investigation rapidly was provided by the development, by Donal and Gamble, of a physical method for the estimation of ethyl iodide by means of its thermal conductivity in a katharometer (2). This improvement so increased the rapidity with which the cardiac output could be estimated, that about two hundred hospital cases, four hundred estimations of cardiac output, were added to the series with the expenditure of less time and effort than had been required to secure the results in the first fifty cases by the chemical technique. When the new series was combined with the old, over 200 cases were secured in which satisfactory estimations of basal cardiac output, metabolism, blood pressure, and pulse rate, had been made on resting patients 15 or more hours after the last meal; and in which orthodiagrams had been secured also. The analysis of these results forms the subject of this paper.As soon as results, secured by any cardiac output method, are examined a difficulty appears which can best be set forth by an example. The average cardiac output of 31 healthy persons is 2.9 liters per minute per 100 pounds, that of 8 cases of anemia 3.2 liters; should the difference be considered significant or not? An estimation of the validity of differences is based on knowledge of the relative accuracy of the methods involved. One may try to ascertain the accuracy of a cardiac output method, when applied to man, by the agreement of duplicate estimations, and by comparison of the results with those obtained by other methods, preferably based on different physiological principles. But it should be emphasized that cardiac output procedures have not attained the position of those methods the accuracy of which can be tested by estimation of known quantities. Therefore, we have fallen back on another way of approaching the problem and have estimated the significance of our differences by statistical procedures.
561