Neither Kappler et al, Simonton et al, nor Weber and Janicki expressed reproducibility in terms of standard deviation. Instead they correlated the results obtained in many patients from one test with those of another; a slope close to one and an intercept close to zero, a significant correlation >0.77, or a low standard error of the estimate was considered to indicate good reproducibility. Thus, the results of this study cannot be compared directly with their results. However, one could compute the 95 percent confidence limits for a given value of a variable as was done in the above example and compare these limits with the range of values reported by these investigators. For example, the range of Vo2max (ie, test two results) for 14 ml/min/kg (ie, test one value) was reported to be 11.5 to 18.0 ml/min/kg by Kappler et al and 13 to 14.5 ml/min/kg by Weber and Janicki. There was only one patient in the Kappler et al report and none in the Weber and Janicki report who exceeded the upper value of 16.6 ml/min/kg that the reproducibility results of this study would have predicted for 14 ml/min/kg.
Similar findings were obtained when the 95 percent confidence limits of maximum HR (determined for 150 beats per minute), maximum systolic BP (160 mm Hg), and exercise duration (800 s) were compared with the range of two test values reported by Kappler et al. That is, in no case did more than one patient he outside the confidence limits.
As discussed above the 95 percent confidence limits associated with HR, systolic BP, respiratory rate, VT, VE, Oa uptake, and C02 production can be calculated at rest and each stage of exercise from the results of this study, as well as the confidence limits for Vo2max, AT, and exercise duration. In general, the CVAR values at rest and the early stages of exercise were found to be greater than CVAR values for the latter stages. This in all likelihood reflects two things: (1) anticipation and apprehension of the test, for it is not unusual for the patient to be hyperventilating and to have an elevated HR at the onset of the test; and (2) a degree of voluntary control over respiration. As the level of work increases ventilation becomes progressively an involuntary process. Beyond exercise stage 4, which is associated with a Vo2 of 5 met, maximum CVAR was below 10.5 percent for HR, VT, and 02 uptake. Maximum CVARs for the remainder of the variables were between 11.9 percent (AT) and 16.9 percent (exercise duration). The existence of a poorer repro-ducibility at low levels of work should be considered when designing or utilizing submaximal aerobic exercise tests.
Since gas exchange variables were monitored continuously throughout the exercise program, we knew how close the patient was to attaining Vo2max. This may have resulted in an underestimation of CVAR for exercise duration. That is, exercise durations from multiple tests are expected to be more consistent in a patient who is monitored and consequently encouraged to attain his or her Vo2max as opposed to a situation where the physician must depend primarily on feedback from the patient to determine when to end the exercise test. Obviously, in the latter case, exercise duration is a function of patient motivation and perception and, as a result, may not be as reproducible. Also, exercise duration was the only variable in this study whose CVAR was dependent on the severity of heart failure.
Finally, it should be stressed that when patient results obtained before and after some perturbation are statistically compared, the difference (be it paired or unpaired variates) does not have to be greater than the 95 percent confidence limits associated with reproducibility in order for it to be significantly different from zero. Herein the statistical test (ie, analysis of variance) takes into account the variance from patient to patient that exists at both instances.