• 12/17/2006
  • Vancouver, BC, Canada
  • Miriam P. Rosin et al.
  • Fifth AACR International Conference on Frontiers in Cancer Prevention Research, Nov 12-15, 2006

The genomic era has fueled a rapid emergence of new technology, with the potential for developing innovative approaches to detection, risk assessment and management of premalignant disease. A key missing link in the development of novel screening and intervention strategies has been our limited understanding of the natural history of the disease.

Not all early disease will progress to cancer. To be effective in reducing cancer risk, molecular (and other) technologies need to target change in early lesions that is strongly associated with outcome – in other words, the likelihood of progression to cancer.

This presentation will describe early results of an on-going Oral Cancer Prediction Longitudinal (OCPL) study located in British Columbia, funded by NIDCR for 8 years (1999 – 2008). This study is evaluating a set of innovative technologies alone and in combination to best correlate with outcome for oral premalignant lesions (OPLs). The plan is to use these devices to guide key clinicopathological decisions on patient risk and treatment. The long-term goal is to create a province-wide screening network in which these devices would act as a series of overlapping sieves that will in a step-by-step fashion progressively filter out patients in the community with high-risk OPLs and triage them to dysplasia clinics where higher-cost molecular tools will guide intervention.

The design of the OCPL study is as follows. Approximately 500 patients are being followed over time (patients are continuing to accrue): half with an oral cancer history (at risk for recurrence) and the other half with low-grade dysplasia (no oral cancer history, at risk for progression to cancer). Patients are seen at 6-month intervals with a rigorous collection of clinical, pathological and demographic data at each visit and repeated sample collection from oral premalignant lesions (OPLs) and high-risk sites (a combination of exfoliated cell brushings and biopsies).

The devices/approaches chosen for assessment in the OCPL study are positioned at key decision points that represent major barriers to screening activities. The first revolves around the clinical assessment of the oral cavity and detection of an oral lesion that requires follow-up. OPLs vary considerably in clinical appearance and the ability to differentiate abnormalities requiring biopsy from reactive lesions, associated with other causes such as infection or trauma, can be difficult. Visualization devices that facilitate the decision to biopsy (the next step in patient evaluation) could have a profound affect on outcome. We are assessing 2 approaches: a hand-held visualization device which makes use of tissue autofluorescence to detect and delineate abnormal lesions and fields requiring follow-up and the use of optical contrast agents such as toluidine blue. Preliminary data using each of these approaches has been promising. We have recently shown that OPLs with retention of toluidine blue have a >6-fold elevation in cancer risk. Fluorescence visualization appears to detect clinically non-apparent disease that is histologically high-risk and may play a role in the identification of surgical margins in the future. A second barrier revolves around risk prediction for OPLs.

At present, the gold standard for prediction involves the determination of the presence and degree of dysplasia in a biopsy. Histology is a good predictor of risk for severe dysplasia or carcinoma in situ (CIS), grouped as high-grade premalignant lesions, which are characterized by persistence, recurrence, and high risk of eventual progression to invasive cancer. Unfortunately, the majority of OPLs have little (mild and moderate dysplasia) or no dysplasia and histology alone does not clearly differentiate between those that will progress and those that will not.

The OCPL study is confirming several retrospectively-observed loss of heterozygosity (LOH) risk patterns previously associated with progression for OPLs. Data from the first 100 oral dysplasia in follow-up (with a limited follow-up time of 44 months) is promising. However, LOH analysis is comparatively time-consuming and labor-intensive. We are currently evaluating a further two semi-automated computer microscopy systems as high throughput filters that could be used for population-based studies and serve as upstream sieves for risk assessment prior to LOH analysis. These computer imaging devices measure specific phenotypic characteristics that make up the appearance of dysplastic cells in a quantitative fashion, with ~120 such features being assessed for each nuclei in a sample.

We have recently completed a pilot study in which a small set of tissue biopsies of OPLs with known outcome (N = 44) were assessed for algorithm-derived nuclear phenotype scores. Strikingly, there was a 9-fold increase in relative risk of progression to cancer for cases with high versus low scores. In a separate study, we used a parallel system, this time with exfoliated cell samples, to generate some equally interesting data. The sample set included a total of 196 cytological samples from areas collected just prior to biopsy, 108 normals and 60 abnormals (patients with squamous cell carcinoma, carcinoma in situ and severe dysplasia). Using a crude algorithm with just 2 features, the system was able to correctly identify 86% of the abnormal cases and 86% of the normal cases. Of equal importance was the fact that the system also correctly identified 92% of samples from sites with inflammation/infection (N = 28) as non-OPLs, supporting a potential use of the system to assist clinicians in differentiating reactive lesions from OPLs requiring biopsy.

In summary, early data from the OCPL study suggests that positioning devices at critical decision points in patient evaluation might greatly facilitate screens for high-risk lesions. It is important to note that this approach of stepwise “sieving” for risk carries with it an additional equally important benefit: the creation of a cohort of patients with OPLs of known outcome upon which genomic studies can be conducted. The OCPL study has such a strategy in place. We are currently using a whole genome bacterial artificial chromosome tiling set (BAC) array (~32,000 clones) to create a database of high-resolution genomic profiles of tumors and early OPLs, with known outcome, to ultimately identify a novel set of predictive markers of progression that may guide intervention strategies.

Finally, we are beginning the process of knowledge transfer to community clinics. A screening clinic has been established in the Downtown Eastside of Vancouver, one of the poorest neighborhoods in Canada with a population characterized by multiple risk factors including heavy tobacco/alcohol usage, poverty, poor nutrition and chronic infections/inflammations. We have already identified a very high risk of disease in this community using our screen: of 250 residents, 2 had cancers and 9 had precancerous lesions. We have also begun the process of knowledge transfer to community health practitioners, beginning with 10 dental offices in the Vancouver mainland. The plan is to begin the process of seeding the use of the various devices into the 2900 dentists of British Columbia, thus creating a broad network that will identify cases for triage to clinics for assessment and intervention.

Authors:
Miriam P. Rosin, Catherine F. Poh, Martial Guillard, Michele Williams, Wan Lam, Calum MacAulay and Lewei Zhang

Authors’ affiliation:
BC Cancer Agency, Vancouver, BC, Canada