Figure 2.

Cancer-risk characteristics across the patient clusters. The final diagnosis categories were non-randomly distributed across the five patient clusters identified in Figure 1. The blue and green patient clusters were significantly enriched for patients with 'low cancer-risk' characteristics (bars in yellow) while the red, purple and gold clusters were significantly enriched for patients with 'high cancer-risk' characteristics (bars in dark brown). (A) The patient counts, from left to right, within 'no diagnosis', 'benign pathologies', 'stones and inflammation', 'benign prostate enlargement' (BPE), 'other cancers', 'non-muscle invasive urothelial cancer' (NMI UC) and 'muscle invasive urothelial cancer' (MI UC) are illustrated for each of the five patient clusters. Following agglomerative clustering 30/36 (83%) patients within the 'no diagnosis' category were in either the blue or green patient clusters (yellow bars). (B) The numbers of patients with normal protein levels are shown by yellow bars. Most patients with normal protein levels fell within the blue (54/112 (48%)) and green clusters (43/112 (38%)). (C) The numbers of patients with pTa stage UC are shown by yellow bars. Within the blue and green patient clusters, 18/28 (64%) and 16/18 (89%), respectively, of the patients with UC had pTa disease (yellow bars). In contrast, when the red, purple and gold patient clusters were combined, 16/34 (47%) of the UC patients had high stage disease (dark brown bars). (D) The number of patients with Grade 3 UC is shown by dark brown bars. Within the red, purple and gold patient clusters 5/7 (71%), 7/11 (64%), and 9/15 (60%), respectively, had Grade 3 UC. In comparison, 10/27 (37%) and 4/18 (22%), respectively, in the blue and green patient clusters had grade 3 UC (dark brown bars).

Emmert-Streib et al. BMC Medicine 2013 11:12   doi:10.1186/1741-7015-11-12
Download authors' original image