Identification of a predictive biomarker for hematologic toxicities of gemcitabine

Junichi Matsubara, Masaya Ono, Ayako Negishi, Hideki Ueno, Takuji Okusaka, Junji Furuse, Koh Furuta, Emiko Sugiyama, Yoshiro Saito, Nahoko Kaniwa, Junichi Sawada, Kazufumi Honda, Tomohiro Sakuma, Tsutomu Chiba, Nagahiro Saijo, Setsuo Hirohashi, Tesshi Yamada

Research output: Contribution to journalArticlepeer-review

41 Scopus citations


Purpose: Gemcitabine monotherapy is the current standard for patients with advanced pancreatic cancer, but the occurrence of severe neutropenia and thrombocytopenia can sometimes be life threatening. This study aimed to discover a new diagnostic method for predicting the hematologic toxicities of gemcitabine. Patients and Methods: Using quantitative mass spectrometry (MS), we compared the baseline plasma proteomes of 25 patients who had developed severe hematologic adverse events (grade 3 to 4 neutropenia and/or grade 2 to 4 thrombocytopenia) within the first two cycles of gemcitabine with those of 22 patients who had not (grade 0). Results: We identified 757 peptide peaks whose intensities were significantly different (P < .001, Welch t test) among a total of 60,888. The MS peak with the highest statistical significance (P = .0000282) was revealed to be derived from haptoglobin by tandem MS. A scoring system (nomogram) based on the values of haptoglobin, haptoglobin phenotype, neutrophil count, platelet count, and body-surface area was constructed to estimate the risk of hematologic adverse events (grade 3 to 4 neutropenia and/or grade 2 to 4 thrombocytopenia) with an area under curve value of 0.782 in a cohort of 166 patients with pancreatic cancer. Predictive ability of the system was confirmed in two independent validation cohorts consisting of 87 and 52 patients with area under the curve values of 0.655 and 0.747, respectively. Conclusion: Although the precise mechanism responsible for the correlation of haptoglobin with the future onset of hematologic toxicities remains to be clarified, our prediction model seems to have high practical utility for tailoring the treatment of patients receiving gemcitabine.

Original languageEnglish
Pages (from-to)2261-2268
Number of pages8
JournalJournal of Clinical Oncology
Issue number13
StatePublished - 1 May 2009
Externally publishedYes


Dive into the research topics of 'Identification of a predictive biomarker for hematologic toxicities of gemcitabine'. Together they form a unique fingerprint.

Cite this