Regression to the mean

Subject variability High intraindividual variability in QTc values (circadian and seasonal variation law of regression to the mean) High interindividual variability in QTc values (males versus females) Unknown prevalence in the general population of subjects carrying silent mutations in the ion channels responsible for cardiac repolarization (these subjects have normal QTc value but reduced repolarization reserve) Variability in the individual metabolic capacity for a given drug... 

This second assumption is equivalent to an assumption that there is no regression-to-the-mean. Regression-to-the-mean is a phenomenon originally reported in 1885 by Galton who showed that the children of tall parents tend to be shorter than their parents, and conversely children of shorter parents tend to be taller than their parents. This is of importance in the context of clinical research because patients are chosen to participate in a clinical trial because they have... 

Changes seen after treatment may not be due to the treatment alone because untreated patients will generally improve because of regression-to-the mean. There is therefore a need to disentangle the treatment effect from regression-to-the-mean. In order to this we need concurrent randomised controls. 

In Equations 4 and 5, r is the multiple correlation coefficient, r2 is the percent correlation, SE is the standard error of the equation (i.e the error in the calculated error squares removed by regression to the mean sum of squares of the error residuals not removed by regression. The F-values were routinely used in statistical tests to determine the goodness of fit of the above and following equations. The numbers in parentheses beneath the fit parameters in each equation denote the standard error in the respective pa-... 

The mechanism of placebo response in blood pressure is not well understood. In a study of 1292 stage I and stage II hypertensive females, 30% responded to placebo erapy. Response varied with age and ethnicity. The response in elderly Caucasian females was greater than in other age and race subgroups, 38% versus 23-27%, respect vely. Factors such as natural history of the disease and regression to the mean have been suggested as the basis for this placebo response. 

The effect claimed cannot be explained by other phenomena such as regression to the mean, time trends, or bias. This highlights the need for appropriate study design and data acquisition. 

The third problem concerns the construction of the yardstick itself. We need to know what would have happened to the patients had we treated them otherwise, but whatever approach we use will have its difficulties. For example, we may use the method of historical controls by which we compare our results to results obtained previously with different patients using the alternative treatment (which may be another treatment or no treatment at all), but the problem is that not only are the treatments different so are the patients and they may differ in important ways we cannot even measure. Another alternative that is possible with some diseases is the method of baseline comparisons, whereby patients are compared with their own baseline values. This method is subject to very many possible biases, among which regression to the mean (see below) and time trends (see discussion about multiple sclerosis above) are important. 

Chuang-Stein C, Tong DM (1997) The impact and implication of regression to the mean on the design and analysis of medical investigations. Statistical Methods in Medical Research 6 115-128. 

A second difficulty arises from circadian rhythms. Measurements may fluctuate during the day according to some pattern quite apart from that induced by a dosing schedule. A third problem is that measurement conditions may vary during the course of a trial. A fourth, which Elliot and Meredith do not consider, is regression to the mean (see Chapter 3). All of these phenomena show the value of measuring the effect of a treatment with respect to a control rather than with respect to baseline. 

Regression to the mean. A statistical phenomenon and an important (and often unrecognized) source of bias in uncontrolled studies which may be explained by example. If a large group of stable individuals chosen at random have their systolic blood pressure measured at two different times, then the mean of the blood pressure on the two occasions may well be similar and so may the variances. Because of variability within individuals, however, the correlation between repeat measures will not be perfect. The only way in which all these three statistical results (regarding mean, variance and correlation) can be satisfied will be if on average there is an increase for... 

---