Statistically significant but clinically irrelevant correlation?
- PDF / 408,113 Bytes
- 2 Pages / 595 x 842 pts (A4) Page_size
- 77 Downloads / 169 Views
Wien Klin Wochenschr https://doi.org/10.1007/s00508-020-01634-9
Statistically significant but clinically irrelevant correlation? Lin Yang · Thomas Waldhoer
Received: 3 February 2020 / Accepted: 4 March 2020 © Springer-Verlag GmbH Austria, part of Springer Nature 2020
Dear Editor, With great interest we read the article of Pekar et al. [1] describing the correlation between spermidine, age and memory performance measured by the minimental state examination. The statement that “spermidine plays a major role in the occurrence of senile dementia” in the discussion section appears to be of concern. Our concerns pertain to the strength of the correlation between spermidine and mini-mental state examination described by the Kendall’s rank correlation (r = 0.15, p = 0.025). It corresponds to an unadjusted R2-value of around 1% in a linear regression model (not shown in their manuscript) compared to 26% in the regression model on spermidine and age (Fig. 3, [1]). By definition, the R2-value is interpreted as the proportion of explained variation. Hence, spermidine only explains 1% of the observed variation in the mini-mental state examination, in sharp contrast to 26% of age for spermidine. In addition, the low R2values of 1% (adjusted R2~ = 0.0%!) is accompanied by very wide prediction intervals in Fig. 1 showing that spermidine certainly has no prognostic ability. That is to say, spermidine does not provide any statistically or clinically meaningful prediction for the value of the mini-mental state examination. Considering that
age has a strong effect on spermidine as well as on the mini-mental state examination but that age is not adjusted for when calculating the spermidine and mini-mental state examination correlation, there is no reason to presume “spermidine plays a major role in the occurrence of senile dementia” based on these analyses. Furthermore, taking into account the number of statistical tests (n = 5), the significance of the pvalue of 0.025 for the association spermidine and the mini-mental state examination may be questionable due to multiple testing. In conclusion, even when there is evidence for an association between spermidine and memory performance in animal studies [2], the findings of the current study do not support the hypothesis of any relevant effect of spermidine on neurocognitive changes in humans. We think carrying out an expensive clinical trial in humans requires more solid preliminary data to support the hypothesis.
L. Yang, PhD Cancer Epidemiology and Prevention Research, Alberta Health Services, Calgary, Canada Departments of Oncology & Community Health Sciences, Cumming School of Medicine, University of Calgary, Calgary, Canada Prof. T. Waldhoer, PhD () Department of Epidemiology, Center for Public Health, Medical University of Vienna, Kinderspitalgasse 15, Vienna, Austria [email protected]
K
Fig. 1
Association spermidine and memory performance
Statistically significant but clinically irrelevant correlation?
letter to the editors
Conflict of interest L. Yang
Data Loading...
