There has been considerable discussion regarding the potential role for proteomics for cancer screening and detection (Reymond,2007; Lin,2009; Dziadziuszko,2008; Lomnystska, 2007; Unwin,2007); however, there are inadequate data in the peer-reviewed literature to permit scientific conclusions regarding ovarian, prostate, or other malignancies.

 

Petricoin and colleagues (2002) reported on the technical feasibility of proteomic screening in a test series of serum from 50 patients with and 50 patients without ovarian cancer.  The spectra of proteins were analyzed by an iterative searching algorithm that identified a cluster pattern that segregated the patients with cancer from those without. This discovered pattern was then used to classify an independent set of 116 masked serum samples; 50 from women with ovarian cancer and 66 from unaffected women or those with nonmalignant conditions. Patients without cancer were considered at high risk, due either to familial breast or cancer syndrome or positivity of BRCA 1 or BRCA 2 mutations. All 50 with ovarian cancer were correctly identified, including the 18 with Stage I cancer. Of the 66 benign cases, 63 were identified as not cancer, yielding a sensitivity of 100% and a positive predictive value of 94%. The authors note that while a positive predictive value of 94% may be acceptable for those high-risk patients, in the larger population of average-risk patients, the positive predictive value must be close to 100% to avoid a high number of false-positive results, which, in turn, would generate additional workup. One of the key outcomes of an ovarian cancer screening test is the ability to identify Stage I ovarian cancer that is potentially curable with surgery. The above study only included 18 patients with Stage I ovarian cancer. The authors state that an important future goal is the confirmation of the diagnostic performance of proteomic screening for the prospective detection of Stage I ovarian cancer in trials of both high- and low-risk women.

 

It should also be noted that the technology used in the Petricoin study is not the same as the technology proposed for the OvaCheck® test. According to the National Cancer Institute, 2005, “The two techniques use different mass spectrometry instrumentation and detection methods, as well as different sample handling and processing methods. Therefore the class of molecules analyzed by these two approaches, and thus the molecules that constitute the diagnostic patterns would be expected to be entirely different.”  Other comments and correspondence in the literature (Lancet, 2002) also question the statistical analysis used by Petricoin and other technical issues. (Diamandis,2004) The results of the Petricoin study have not been reproduced elsewhere. (Unwin,2007)

 

Ornstein and colleagues (2004) reported the results of serum proteomic profiling in 154 men with serum PSA ranging from 2.5 to 15.0 ng/mL.  A total of 63 samples (30 malignant, 33 benign) were used as the training set to identify a proteomic pattern that could distinguish benign from malignant disease. The results of the training set were then applied to the remaining 91 samples (i.e., the “testing” set) in a blinded fashion. In this testing set of 63 with negative biopsies and 28 with positive biopsies, there was 100% sensitivity and 67% specificity. These data imply that if the results of proteomic profiling were used to deselect patients for biopsy, 42 of 63 (67%) patients without prostate cancer could have avoided biopsy. The authors note that using a training set of only 63 samples may be inadequate, and that “before this new technology can be applied in clinical practice, much larger and diverse training and testing sets will be needed.”

 

McLerran and colleagues (2008) selected serum samples from biorepositories from patients with 1) prostate cancer with Gleason score 7 or higher; 2) prostate cancer with Gleason score less than 7; or 3) negative prostate biopsies with PSA 10 mcg/L or less and no history of cancer of any kind, a normal digital rectal exam, and no inflammatory disease. They also selected two control groups: one with a history of inflammatory disease but no cancer and one with no history of prostate cancer but a history of another type of cancer.  Four hundred specimens were analyzed by mass spectrometry after random selection from the 5 groups of patients, with 125 from the group with high Gleason grade, 125 with low Gleason grade, 125 from the biopsy-negative group and 50 from each of the control groups. The investigators sought to derive a decision algorithm for classification of prostate cancer from the mass spectrometry data, but found that they were unable to separate the patients with prostate cancer from biopsy-negative controls, nor were they able to separate patients with high and low Gleason scores. The conclusion was made that in the validation process, this protein-expression profiling approach did not perform well enough to advance to the prospective study stage.

 

A number of preliminary proteomic studies are available for many cancers including lung, colorectal, gastric, pancreatic, liver, cervical, endometrial, bladder, lymphoma/leukemia, melanoma and astrocytomas. (Reymond,2007; Garrisi,2008; Li, 2006; Leman,2007; Belluco, 2007; Bast,2007).

 

As of May 2009, there are 8 ongoing phase III trials for a variety of malignancies, which are secondarily analyzing proteomic profiles in serum as predictors of survival, risk of disease progression and response to treatment: non-small cell lung cancer (NCT00300729 and NCT00738881), melanoma (NCT00389571), ovarian cancer (NCT00426257), breast cancer (NCT00516425), prostate cancer (NCT00567580), hepatoblastoma (NCT00652132), and colon cancer (NCT00647530).

 

 

 

 

 

 

 

 

A literature search conducted through September 2016 did not reveal any new information that would prompt a change in the coverage statement.