April Core Spotlight: Biostatistics
In this DF/HCC News Spotlight Edition, we present the Biostatistics Core, which provides statistical expertise for the planning, conduct, analysis, and reporting of clinical and translational research studies, prevention and genetic investigations, and basic research in the biology of cancer. The Core also supports DF/HCC members in preparing grant proposals, and plays a key role in protocol review and monitoring.
The Core is housed in 5 of the DF/HCC institutions. Investigators should go to www.dfhcc.harvard.edu/research/core-facilities/biostatistics/contact-information/ to find a list of statisticians and their contact information, alphabetized by cancer research areas.
As of the CCSG renewal, we now have a new bioinformatics navigator service, headed by Dr. Aedin Culhane (email@example.com, 617-632-2468). Dr. Culhane is available to help researchers identify appropriate bioinformatics resources throughout Harvard and its affiliated institutions. Researchers should contact Aedin directly to learn more about this valuable new resource within the core.
- Design, monitoring and analysis of PI-initiated clinical trials, retrospective studies and observational cohorts of anti-cancer therapy, cancer risk reduction and cancer prevention, toxicity amelioration and prophylaxis, and patient and clinician decision making
- Assistance with population and family studies of germline and somatic genetics of cancer, environmental and behavioral cancer risk factors, effects of screening, and medical care outcomes
- Consultation on studies of prognostic, predictive, and surrogate markers in cancer treatment or risk reduction or prevention
- Collaboration on studies of cancer biology, immunology, and virology experiments on cancer initiation, promotion, progression, and metastases in cell lines, animals, or specimens from tissue banks
- Education and training on statistical methods, and data collection design and implementation.
Circulating exosomal miRNAs are critical prognostic markers independent of cytogenetics and International Staging System in Multiple Myeloma
PI I. M. Ghobrial, MD (DFCI), statisticians E. Weller (DFCI) and L. Trippa (DFCI)
This project’s aim was to establish the prognostic significance of circulating exosomal microRNAs (miRNAs) in multiple myeloma (MM). [A] Specimens from 10 MM patients and 5 healthy controls were analyzed by small RNA sequencing using quantile normalization and average of two replicates after normalization; this resulted in 22 exosomal miRNAs found to consistently distinguish between MM samples and control samples.
[B] Serum samples from the time of diagnosis for 156 MM patients after treatment with bortezomib and dexamethasone were assessed using a quantitative RT-PCR array for the 22 miRNAs, each being evaluated both as a continuous measure and when classified as “high” or “low”. Using a stratified algorithm, the patients were randomly divided into a training set and a validation set.
A proportional hazards model for progression-free survival (PFS) which included as covariates the MM International Staging System (ISS) and poor risk cytogenetics (t[4;14] and del17p) was then done in the training dataset to identify “significant” miRNA signatures; these signatures were then tested in the validation dataset. This process, including the random division into two data sets was repeated 500 times, and the proportion of times that a signature found in the training set was validated in the validation set was noted.[C] The signature based on miR106b, miR18a, and let7e was selected for further validation, since the three represent different functions. Compared to patients with a low risk score for this signature, patients with a high risk score had a shorter PFS in the validation set with an average hazard ratio of 2.6 and p=0.0005 (Fig. 4). In ROC analysis, a combination of this signature together with the ISS and cytogenetics had a significantly better prognostic value than the combination of only ISS and cytogenetics (Fig.5).
Contribution of the Core: The Biostatistics Core provided statistical design and analysis for studies A, B, and C above and helped draft the manuscript.
Publication: Blood, 2017 Feb 17; Epub ahead of print, PMID: 28213378
Manier S, Liu CJ, vet-Loiseau H, Park J, Shi J, Campigotto F, …,Weller E, Trippa L, Ghobrial I, Prognostic role of circulating exosomal miRNAs in multiple myeloma, Blood 2017 Feb 17, doi:10:1182/Blood-2016-09-742296 [Epub ahead of print].
Protective Effect of Colonoscopy for Rectal Cancer but not for Colon Cancer
PI: E. Giovannucci, MD, ScD (HSPH), statistician B. Rosner (BWH)
The aim of this project was to examine whether the protective effects of colonoscopy screening are different for colon vs. rectal cancer. The study was based on 1718 colorectal cancer cases in the Nurses’ Health Study from 1980-2010, (1342 colon and 376 rectal cancers); competing risk Cox regression models were used. A unique aspect of the analysis was the construction of exposure variables where risk factors are updated every 2-4 years over a 30 year period. It was assumed that cancer risk is based on long-term exposure levels and that cumulative sums of exposure over time are the most accurate way to characterize the risk induced by a risk factor.
In general, in this data set the risk factors for colon and rectal cancer are similar. After adjusting for smoking history, BMI, physical activity, red or processed meat servings per day, folate levels, and calcium levels, protective effects of colonoscopy screening are stronger for rectal vs. colon cancer. Plots of the estimated cumulative incidence of colon and rectal cancer for some subgroups of women are shown below. For colon cancer (Fig. 8), the effect of screening is small compared with the effect of other risk factors (the solid green line and dotted purple line for high risk women are close together and both are very far from the dashed lines for medium risk and low risk individuals). However, for rectal cancer (Fig. 9) the effect of screening is proportionally large (difference between solid green line and dotted purple line), and about the same magnitude as the overall difference between high risk and medium risk individuals (average of solid green and dotted purple lines, as compared to dashed blue line). This work could not have been done without DF/HCC support.
Contribution of the Core: The Core provided an analysis plan (including construction of exposure variables changing over time) and then did the analysis.
A manuscript is being drafted.
For more information or to contact the Medicinal Chemistry Core, visit the core website here.