Comment on: “clinical relevance of insulin-like growth factor-1 to cardiovascular risk markers”

We read with great interest the article by Harada et al. [1] entitled “Clinical relevance of insulin-like growth factor-1 to cardiovascular risk markers”. Although its original concept, we want to address some points that merit more attention.

The empirical analysis consists of comparison of groups. The authors state in the first sentence under the subtitle ‘study protocol’ that patients are divided into three groups and the levels of insulin-like growth factor-1 (IGF-1) were compared between groups. The data collection for this purpose is meticulous. Not only the patients are categorized into three groups depending on pituitary disease, the non-pituitary cases are also divided into subgroups. Furthermore, data on IGF-1, hemoglobin, hematocrit, C-reactive protein, procalcitonin, D-dimer, fibrin/fibrinogen degradation products, as well as data on age and gender is collected [1].

Following such careful and conscientious data gathering, the authors express in the first sentence under the subtitle ‘statistical analysis’ that analysis of variance (ANOVA) and linear regression analyses were used to determine differences between means [1]. It is our humble opinion that it is this part in which there is the most room for improvement. First reason is empirical findings regarding the ANOVA and the linear regression is not reported at all. The only exception is the expression of the negative relation between IGF-1 levels with advancing age in all as well as non-pituitary cases and across genders as well in case of linear regression analysis. Linear regression could have easily depicted how IGF-1 levels changes across the three groups and the relationship between age of patients and gender. Moreover ANOVA could formally show whether the mean of the three groups differ from each other. However authors limit what they report to correlations (and the relevant p-value) and means (and a confidence intervals of unknown significance levels) of the groups. Unfortunately, what is reported is not sufficient to reach the conclusions the authors delivered in the article. The authors should have reported their findings especially the regression (in more detail) and the ANOVA results. That would certainly increase the credibility of their conclusions.

One final point is that in ANOVA Tukey-Kramer post-hoc test, which finds the means that are significantly different from each other, assumes the groups are normally distributed and within group variances are equal [1]. Since the distributional properties of the data is not reported sufficiently and we see no reason for equal variances and normal distributions in groups, Tukey-Kramer post-hoc test may not be a good choice for investigation of whether means are statistically different from each other.

Disclosure statement

No potential conflict of interest was reported by the authors.