Klin Onkol 2010; 23(1): 25-33.
Summary
Background: With respect to diagnostic and therapeutic progress, it may occur that the statistical sets of patients evaluated and treated with uniform methods are small. As a consequence, it is meaningful to check a greater number of statistical approaches. It is suitable to verify whether, for instance, the differences between the results (+) and (++) for estrogen and progesterone receptors (ER and PR) in breast cancer have an effect on the length of survival. This question could be answered with the use of several Kaplan‑ Meier survival curves. However, it is also profitable to judge the simple graph of survival in dependence on receptor concentration. Nevertheless, traditional regression brings too great an error to this method of assessment. Therefore, the use of orthogonal regression is much more precise. Since it can be assumed that no non‑revealable micro‑metastases were present at the time of operation in some patients with N0, it is possible to achieve healing ad integrum of them using only simple surgery. Consequently, we concluded that it was necessary to exclude from the evaluation the group of patients in N0 surviving 10 years (in the search for evidence of the post‑operative impact of age‑based reduction of blood estrogen on survival). Design and Subjects: We verified these considerations
when monitoring the ER and PR influence on overall survival. We performed this analysis in an approximately 2‑year sample of 74 female patients who received the described treatment in Pardubice hospital. At the time of operation, 56 were postmenopausal and 21 of these postmenopausal
patients were in stage N1. Methods and Results: ER and PR in breast tumours were examined in the cytosol of operational biopsies. Adjuvant radiological treatment was used in addition to the surgical treatment of primary tumours and their original and post‑operative metastases.
In the case of premenopausal patients with ER, (+) therapeutic sterilization was performed. The finding of higher ER in postmenopausal surviving patients (in comparison to dead ones) was below the boundary of statistical significance. Also, longer survival in cases of higher ER concentrations in the group of dead N1 patients was below the boundary of statistical significance in the use of traditional regression. Therefore, we put together evidence from the group of surviving patients with evidence from the group of dead patients. In the case of N1 patients surviving 10 years, we rounded their survival period to 15 years for inclusion in the graph of survival dependence on ER. In the case of the combined (premenopausal with postmenopausal) group, statistical reliability appeared for longer survival of higher ER already in traditional regression. However, for the postmenopausal alone, the difference was statistically insignificant. Nevertheless, if we used orthogonal regression (similar to Deming regression) instead of traditional regression, then the reliability of the dependence of the length of survival on ER increased (in the last cited graph) to such a degree that it was statistically highly significant (at the level of 0.001) even in case of just postmenopausal patients. The same level of statistical reliability was achieved in the Kaplan‑ Meier analysis. Also in the case of PR – the higher concentrations of this receptor in survivors compared to dead patients were not statistically significant. But (in contrast to ER) in the case of PR, we observed a statistically significant increase in survival time depending on the receptor concentration within the group of only the dead patients – hence without putting them together with the surviving patients).
Conclusions: The graph of the Kaplan‑ Meier analysis is more frequently used when solving these problems but the graph of simple dependence of survival on receptor concentration should not be neglected either because, for example, it better shows the difference in survival between ER(+)
and (++). Nevertheless, it is necessary to use orthogonal regression in it. The greater suitability of PR and ER for short‑term and long‑term prognosis, respectively, which we identified in our statistical set, is in concordance with the literature.
