In comparing 2 methods for assessing the same variable, it is usually the limits of agreement (the spread of the results) that are more important than the mean difference (net bias), which usually follows a regression toward the mean and can in principle be zero, despite obvious differences between measurement techniques. In our paper, the 95% confidence intervals were ±57.6 g and ±46.3 g for M-mode and 2D (Reichek formula), respectively, when compared with 3D CMR measurements, with mean differences of −14.3 g and −2.6 g, respectively. It is unsurprising that different papers may vary in the mean differences between techniques, given the different populations studied and the wide variety of left ventricular shape, but we note the very similar limits of agreement for both M-mode and the Reichek formula between our study and the correspondents’ paper (in press). For serial studies, it is again this spread of values (interstudy reproducibility) that is of prime importance in assessing individual and group changes, and it was noteworthy that the limits of agreement for the serial changes in our study were very similar to those for the baseline measurements.
We did not set out to test all available echocardiographic formulas. The intention was to test the reliability of the principle of calculated formulas with an assumed geometric shape against direct measurement, and we chose a commonly used and simple formula for 2D measurement, as well as the most widely used M-mode formula. To compare many different formulas would have been cumbersome and added little to the study.
Lastly, the exercise program involved 69 fixed training sessions over 10 weeks, with predominantly strength training in the upper limbs, and endurance training for the lower limbs. In addition, much of the army recruits’ training outside these sessions involved physical exertion of a similar nature. Space constraints precluded a more detailed description of the exercise regime, especially as we were not studying the effect of differing types of exercise on left ventricular growth, which has been well studied previously.1–3