The Link between Biological Features and Clinical Outcome - Essay Example

To analyse the link between biological features and clinical outcome is the main purpose of the article analysed. The used very easy to interpret methods to show their data (contingency tables, frequencies), but also used different to analyse the same samples ( and Fisher's exact test), they further used specially developed statistical tools (Kaplan-Meier estimation methods and Cox proportional hazard regression models) that are particularly developed for the study of survival statistics. The authors show that ILC patient are not different from IDC patients when it comes to survival but can be treated different with better outcome given their different biological signature. Introduction: In this paper, the analysis of clinical outcome from histological findings has been analysed to show correlation between biological factors and patient survival. The main hypothesis tested is to show the existence or absence of any differences between two histological features to the patients' clinical treatment and outcome: Invasive lobular carcinoma (ILC) and Invasive ductal carcinoma (IDC). The aim was first to establish risk factors and use them as variables for patient survival through the time of the test. This approach has the potential of producing quantifiable risk factors that can help clinicians to design different clinical management strategies based on easy and fast diagnostic procedures. In this article, we will try to elucidate the merit of this paper to achieving the aims it sets. The Body: In order to achieve some statistical significance the authors analysed a big study population of 49309 patients, and split the population into two groups depending on the histological features of the breast tumour so that 4140 fell under the Invasive lobular carcinoma (ILC) requirements, and the other 45, 169 under the Invasive ductal carcinoma (IDC). The authors, however, excluded the mixed population that might have shown the nature of the relation between the two features examined (e.g. additive effects). The criteria used for the assignment of particular sample to each group were very varying as they were classified using different histochemical, and biochemical methods from 1970 to 1984, this has an effect on the heterogeneity of the populations tested. Have these considerations taken in account; it would have established further subgroups, exposing the whole study to similar drawbacks of earlier studies where the size of the groups was limited. The factors analysed to test the hypothesis were selected from the literature, both laboratory and clinic: Estrogen receptor (ER) levels, progesterone receptor (PgR) levels, DNA ploidy, S-phase fraction, HER-2 status, Epidermal growth factor receptor (EGFR) levels and p53 status. Other factors, if chosen, may have affected the outcome of the results such as the expression of cyclin D1, a known regulator of cell cycle, this regulator is a known prognostic factor for breast cancer. Having examined the main statistical samples and variables used in this study, we now have to concentrate on the statistical methods used in the study to achieve its conclusions. To describe, the populations used, the authors used frequencies and medians; the comparison between the characteristics for correlation was achieved through contingency analysis, square and Fisher's exact tests. Contingency tables are by far the best analytical tool suited for the comparison of two variables, in this article the authors compare as in table one between ILC and IDC and how dependent are they on the characteristics they tested. Contingency coefficient usage has the added the added benefit of readable interpretation over a simple square as it shows independence at value 0, and show increasing dependence as the values approach 1. The square is most appropriate for the analysis of large population as it is the case in this study; however, Fisher's exact test is a robust tool for analysing the dependence of two variables independent of the population size as it uses exact probabilities from the hypergeometric distribution. In most of the analysis reported the p values were significant, showing that the studied dependence, correlation or regression is unlikely to occur by chance. For testing the survival hypothesis, the authors chose clear cut definitions for overall survival (OS), and Disease Free Survival (DFS). The curves drawn for OS and DFS were achieved by using the Kaplan-Meier estimation methods, whereby the vertical axis represents hypothetical probability estimates of survival, and the precision of estimates is dependent on the number of scored events. Though these curves allow an accurate estimate of survival irrespective of censoring or drop out, they tend to show higher survival rates at the beginning of the curve when the number of patients is considerably high. For The comparison between the two variables, the authors used Log-rank which is suited for when some of the observations are censored; otherwise it can be reduced to Mann-Whitney test as seen in Figure 1. The authors also tested the importance of multiple variables and how they are associated with the two histological features. In order to achieve some significant statistical outcome the authors, used Cox proportional hazard regression models. In these models several survival risk factors are analyzed, the hazard is the probability of the final endpoint (e.g. Death in table 4 and 5) and is shown as a relative risk. Using this myriad of tools the authors have been able to report some conclusions about the histological importance of breast cancer and its effect on survival. ILCs have been found to be larger in size (53.8% compared top 48.6% larger than 2 cm with P < 0.0001), Nuclear receptor positive (e.g. ER 92.7% as compared to 81.2% with P < 0.0001), diploid but with low S phase (69.8% versus 43.6%; P < 0.0001, and 87.8% of versus 68.6%; P < 0.0001, respectively), and similar significant outcomes for the status of p53, EGFR and HER2. These new findings correlated and fitted well the literature on the subject, and were consistent with the difference in age between ILC and IDC patients, as these biological findings would predict an older population. As the patients age, low proliferation is noticed and DNA damage genes are induced which is the case with the size, cell cycle analysis, nuclear receptors expression, and p53 status. These also pointed to the role of these characteristics in the invasiveness of ILC as compared to IDC. From tables 2 and 3, ILC has shown strong metastasis potential to sites where nuclear receptors are important (e.g. ovary), or prone to DNA damage due to high cell turnover (e.g. Gastrointestinal tracts). It is also important to note that the involvement of both breasts in ILC tumorgenesis is almost double that of IDC confirming that the biological findings of this study are very accurate and reflect not only statistical significant but very indicative therapeutic potential. The next logical question answered by the study was: given the differences in risk factors mentioned between ILC and IDC, is there any difference in the overall survival of the patients From the first set of results it could postulated that there is a correlation between one or some of the factors reported. Using regression analysis, the authors failed to show any difference between the two different histological features of breast cancer. Conclusion: In this article(Arpino et al., 2004), the authors analyzed some biological factors and their role in the prediction of patients survival and clinical outcome. They compared the role of these factors in Invasive lobular carcinoma (ILC) and Invasive ductal carcinoma, and found that the former occurrence is dependent on age and microenvironment sensors, and that these can be used in designing better therapy for this group, but failed to establish any survival advantage of having either of these features. They use in this paper, tables and curves that are easily interpretable, and show the results in an easy format. They used other more robust techniques to validate their data and show any relation between variables, and chose to use different tools to test the same set of events. Arpino, G., Bardou, V. J., Clark, G. M., and Elledge, R. M. (2004). Infiltrating lobular carcinoma of the breast: tumour characteristics and clinical outcome. Breast Cancer Res 6, R149-156. Read More