Writing the Discussion Section
5- 25
Example of a Well-Written Discussion (Basic Science Study)
Our finding of molecular genetic differences between ovarian cancers from BRCA1 and BRCA2 mutation carriers, familial non – BRCA1/2-related ovarian cancers, and sporadic ovarian cancers suggests that BRCA1/2 mutation status and family history of ovarian cancer influence the somatic genetic pathway of ovarian cancer progression. [Conclusion and supporting finding] Some of the data from this study are consistent with previous metaphase comparative genomic hybridization (CGH) and loss of heterozygosity (LOH) studies in ovarian cancer (18 – 20); for example, regions of common deletion on chromosomes 4p, 6q, 9p, 13q, 18q, and Xp have frequently been identified using LOH analysis (18). Similarly, a previous study in which metaphase CGH was used to analyze 100 sporadic ovarian tumors identified multiple regions of loss and gain that are consistent with our findings (19). [Studies that agree] However, [A good transition word; it links the ideas in this paragraph to those in the previous paragraph and indicates that this paragraph contains ideas that differ from those in the previous paragraph] some of our data differ from those in previous studies (18 – 20); for example, there are notable differences in the frequency with which alterations on chromosomes 3q, 5q, 6p, 12q, 17, 19p, 22q, and Xq were detected between this and other studies. [Studies that disagree] Some of the disparity between CGH and LOH data may be explained by differences between the 2 methods in their ability to resolve genetic alterations. Metaphase CGH paints a picture of gross genomic alterations, including changes in chromosome copy number, whereas LOH analysis produces better, locus-specific resolution. However, LOH analysis is limited as a genome-wide screen because it requires high-density microsatellite mapping, which is both time- consuming to perform and a considerably greater drain on DNA resources than is CGH. Another reason for some of the differences between this and other metaphase CGH studies could be that whereas most previously published CGH data are from sporadic ovarian cancers only, approximately half of all tumors in this study were from BRCA1/2 mutation carriers and only a fifth from sporadic cases. Previous studies in breast cancer suggested that the presence of a germ-line BRCA1/2 mutation can influence the pattern of somatic genetic alterations during tumor development (21, 22). [Gives explanations for the disagreements, and interprets current findings] We found evidence of a similar [A transition word that links the ideas in this paragraph to those in the previous one and indicates that the ideas in this paragraph are similar to those in the last paragraph] influence when the frequencies with which somatic genetic changes in tumors from BRCA1, BRCA2, familial non-BRCA1/2, and sporadic cases were compared. We identified multiple differences between the 4 tumor groups, which suggests that they differ in some aspects of tumor development. However, we carried out a large number of significance tests, and it is likely that some of these differences are chance occurrences. [Possible limitation of the study] There were 200 individual comparisons between the 4 tumor groups, and 32 alterations with a P value of <0.2 were selected for additional analyses in which an additional 320 pair- wise comparisons were performed. Applying the Bonferroni correction to the results of these analyses would require a P value of <0.00016 to achieve a conventional level of significance of P < 0.05. The sample size of this study was not large enough to generate such a small P value, and indeed the smallest observed P value was 0.001. However, we observed 41 pair-wise comparisons with significant differences at the 0.05 level, compared with 16 expected if there were no true differences in frequency of alteration between tumor types; and 8 significant differences at the 0.01 level, compared with 3 expected. This suggests that a substantial proportion of these differences are real. [Reason why the possible limitation might not be a true limitation, and interpretation of findings]
Made with FlippingBook Digital Publishing Software