Left Main Coronary Artery Stenosis

In their article on left main coronary artery stenosis (Circulation 57: 1111, 1978), Campeau et al. discussed the overall survival of their operated and unoperated cohorts. In attempting to interpret their data I find a discrepancy which I am unable to resolve. Their unoperated cohort included 114 patients entered at 0 time; total deaths in the unoperated patients during the first 5 years were 34. If those patients lost to follow-up were excluded, this would yield a 5year cumulative survival rate of 70.1%. If, on the other hand, all 14 patients are presumed to have died an obviously unwarranted assumption this places the cumulative survival rate at 57.9%. 1 suspect it lies somewhere in between these two figures. In their table 2, however, the authors cite a 5-year cumulative survival rate of 48.5%. This number seems to be the basis for the graph of the survival curve of unoperated patients as portrayed in figure 2. Perhaps I am missing an important point; however, I cannot logically accept any amount of statistical manipulation which produces a higher mortality rate than would be generated by the inclusion of all patients lost to follow-up in the non-surviving group. ARTHUR C. SGALIA, M.D. Milford-Whitinsville Hospital Milford, Massachusetts 01757


Left Main Coronary Artery Stenosis
In their article on left main coronary artery stenosis (Circulation 57: 1111(Circulation 57: , 1978, Campeau et al. discussed the overall survival of their operated and unoperated cohorts. In attempting to interpret their data I find a discrepancy which I am unable to resolve. Their unoperated cohort included 114 patients entered at 0 time; total deaths in the unoperated patients during the first 5 years were 34. If those patients lost to follow-up were excluded, this would yield a 5year cumulative survival rate of 70.1%. If, on the other hand, all 14 patients are presumed to have diedan obviously unwarranted assumptionthis places the cumulative survival rate at 57.9%. 1 suspect it lies somewhere in between these two figures. In their table 2, however, the authors cite a 5-year cumulative survival rate of 48.5%. This number seems to be the basis for the graph of the survival curve of unoperated patients as portrayed in figure 2. Perhaps I am missing an important point; however, I cannot logically accept any amount of statistical manipulation which produces a higher mortality rate than would be generated by the inclusion of all patients lost to follow-up in the non-surviving group.
ARTHUR C. SGALIA, M.D. Milford-Whitinsville Hospital Milford, Massachusetts 01757 The author replies: To the Editor: Dr. Sgalia's calculation of the 5-year survival is based on the assumption that all 114 unoperated patients entered in our study had a potential follow-up period of 5 years. In fact, the duration of potential follow-up periods varied from I month to 8 years and, as shown in table 2 of our article, only 21 patients completed the 49-60-month interval after entry in the study. Thus, the 5-year total survival of 70.1% (80 survivors out of 114 patients) mentioned by Dr. Sgalia is erroneous. The method of calculating cumulative survival rates in our study considers the various potential follow-up periods (interval survival proportion for each yearly interval) as well as patients lost to follow-up during the yearly intervals when their course is known.
We agree that patients lost to follow-up represent a problem in the calculation of survival. We believe it is acceptable, however, to exclude them whenever their course becomes unknown, rather than to consider them alive or dead, or to assume on a purely empirical basis that a particular proportion had died (50% for instance, as suggested by Dr. Sgalia). If the exclusion of these patients appears unwarranted, it would then appear more realistic to assume that the survival of these patients lost to follow-up be equal to that of patients whose course is known. The 5-year cumulative survival for the unoperated cohort calculated on the basis of this latter assumption would be 47.3%, a value not significantly different from the 48.5% obtained by excluding these patients during the yearly interval when their course became unknown.

Propranolol and Enzyme Levels
To the Editor: Peter et al. have recently reported a study in which they showed that early propranolol treatment of patients with acute myocardial infarction (MI) resulted in lower total cumulative CPK appearance, a lower rate of CPK appearance and also lower peak levels of CPK activity.' Although controversial, there is evidence that total CPK appearance reflects the size of the myocardial infarct, and thus the implication from their work is that propranolol reduced the size of myocardial infarcts in the treated group. Although propranolol may be useful in patients with acute MI, I think that this paper has a number of potential sources of error.
Theoretically, the method of calculating KD from the terminal portion of the CPK vs time curve is unsound. The assumption using this method is that when the terminal portion of CPK vs time is linear on semilog paper the CPK appearance function, (fit]), is zero.
However, it has recently been shown that this assumption may not necessarily be correct,2 so their method may underestimate KD in some cases. A major problem with this type of study is to ensure that the intervention by itself (in this case, propranolol) does not interfere with CPK kinetics, consequently resulting in lower values for the estimated CPK functions, without altering the size of the myocardial infarct. I do not feel that Peter et al. have ruled out this possibility. Cairns and Klassen3 have shown that in dogs propranolol increases K0 significantly and any intervention using propranolol might produce an artificial reduction of infarct size solely on the basis of this change. Peter et al. stated that propranolol did not effect KD, using as evidence the fact that the mean KD for treated and control cases were not statistically different. However, they calculated K0 from the terminal portion of the CPK vs time plot. Since the last dose of propranolol was given well before this time in many of their patients, the effect of propranolol on KD may have diminished and therefore they may have missed it. For example, KD is calculated in figure 1 55-80 hours after the patient's entry into trial. Propranolol, however, was administered only for the first 27 hours, and thus, between 55-80 hours the mean serum propranolol was much below the mean level for the initial 27 hours ( fig. 2). Therefore, the KD was calculated at a time when the effect of propranolol on its kinetics could not be assessed. Peter et al. compared peak CPK activity levels of the control group to the treated group because they felt that this comparison did not involve assumptions about CPK kinetics. This is not correct.
Peak CPK activity is affected by K,. In the extremes, if KD = 0, peak CPK would be equal to the integral of f(t), and if K = , then peak CPK would be zero. Thus changes in KD induced by propranolol would also effect peak CPK levels.
Although propranolol may be useful in limiting infarct size, the results of this study should be interpreted with caution as they may simply represent the effect of propranolol on CPK kinetics. It is imperative in studying the effect on infarct size of any intervention using the mathematical model of Shell et al. that the effects of the intervention on the parameters of the model be assessed.
ARTHUR SLUTSKY, M.D. Pulmonary Division Peter Bent Brigham Hospital Boston, Massachusetts 02115