| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorials |
Cancer Research UK Clinical Centre, St James’s University Hospital, Beckett Street, Leeds LS9 7TF, UK, Fax +44 113 2429886, E-mail r.banks{at}leeds.ac.uk
The discovery of new diagnostic markers has been limited for many years, and new biomarkers that can be used for treatment selection and monitoring are also needed. Proteomic-based studies offer real hope of new marker discovery (1)(2), but many challenges remain and progress has been much slower than expected (2)(3)(4)(5). Many initially promising studies have not been followed by wider-scale validation and application. Why has the rate of attrition of potential markers been so high? The process of taking validated biomarkers through to clinical use is a major logistical and financial challenge in its own right (2)(3)(4)(5), but the limiting factors are even more fundamental. The 2 reports by McLenna and colleagues (6)(7) provide examples of how preanalytical factors can contribute to the lack of fulfillment of initial promise. Although based on the use of the SELDI platform, the thought-provoking results reported in this issue of Clinical Chemistry apply in principle to all clinical proteomic studies, irrespective of analytical platform and disease area.
So, what do these 2 reports show? In 2005, the National Cancer Institute/Early Detection Research Network sponsored a multicenter evaluation of the utility and robustness of SELDI serum profiling for detection of prostate cancer. In the first-stage study, the use of an algorithm developed from earlier single-site data generated impressive discrimination between cases of prostate cancer and controls (8). Also, given stringent standardization and quality control, reproducibility between laboratories approached that achieved within a single laboratory. The reports published in this issue (6)(7) describe the next stage of the validation process, with the finding that diagnostic performance was compromised by bias. Differences in profiles were found depending on whether serum specimens were collected before or after 1996. Several factors, such as time of storage and number of freeze-thaw cycles, may have contributed to this bias. When analyses were repeated using a new cohort of specimens and eliminating this bias, discrimination between cancer and noncancer was not achieved (7). With the benefit of hindsight, it is easy to be critical of the original report (8). The earlier finding of platform reproducibility remains valid, however. The subsequent reports simply emphasize the crucial role of preanalytical factors that can introduce bias and that may be relevant to interpreting the results of many published studies. These studies do not resolve the previously raised issue (9) of the fundamental utility of SELDI as a profiling tool. Resolving that issue will be achieved only with further studies investigating the validation of initial promising findings in several disease areas.
The issue of bias in biomarker studies has been highlighted previously (10) and, of course, is not restricted to proteomics. Awareness of the importance of preanalytical factors is increasing (3)(11) and has been emphasized in several recent editorials and viewpoint presentations (9)(12)(13). Such influences have been recognized by practicing clinical chemists for many years. Their impact is likely to be much greater in proteomics studies, however, given the simultaneous analysis of many proteins, resolution of multiple forms of proteins, and detection of peptide fragments arising from active cleavage processes. Although this issue has been highlighted by SELDI studies, other analytical platforms can be affected in similar ways. Factors including duration and temperature of storage, type of collection tube, and delays in specimen processing, in addition to biological factors such as diet, age, and sex, may be important (3)(11). No one set of conditions will be ideal for all proteins. In any study, the importance lies in the consistency of the approach and the analysis of data to investigate possible extraneous influences that may lead to bias.
Identification of the importance of preanalytical factors has implications for the use of large sample banks. Such banks enable many prospective studies to be carried out in a timely fashion that otherwise would require years for sample acquisition, but sample banks vary in their adherence to consistent sample processing protocols over time. Many proteins, however, may be unaffected by many of the preanalytical variables. Therefore, any study using or proposing to use such banks should not be instantly condemned, but the way in which the banked samples are used should be examined.
A possible, albeit imperfect, solution to this issue is to restrict the initial phase of biomarker discovery to specimens collected by rigorous adherence to banking protocols. Some years ago, when advocating exactly this approach at a conference, I was questioned as to whether such specimens would be representative of real-world specimens collected routinely in hospitals. Rigorous collection protocols, of course, would differ from those of routine clinical practice. By striving to develop banks of ideal specimens, however, we may decrease background "noise," and newly discovered markers are more likely to be real and not introduced artifacts. After marker discovery, the effects of relaxing sample processing can be explored systematically as part of the test evaluation and validation process, and decisions can be made about suitability of existing large sample banks for validation purposes.
Sometimes the only option is to use banked samples for which the degree of variability in handling and storage is not known. Such banks may provide unique opportunities to access samples representing rare diseases or specific clinical trials. In such cases the main control must be exerted at the level of analysis of results by taking into account as much information as possible regarding factors such as duration of storage, temperature of storage, number of freeze-thaw cycles, and collection center. This process is not foolproof but is nevertheless worthwhile because these unique sample collections may lead to discovery of markers of real diagnostic value for further rigorous validation.
Whenever sample collections are used, a key aspect is the information collected about sample processing and banking. As much information as possible should be collected about sample processing and storage, in exactly the same way as the associated clinical data, thus allowing selection of appropriate samples and rigorous examination of confounding factors. Investigation conducted under these circumstances is an example for which the phrase "knowledge is power" (paraphrased interestingly from the English philosopher Sir Francis Bacon, a proponent of scientific methodology) will be borne out in terms of minimizing the risk that biomarker discovery will be compromised by extraneous influences.
Acknowledgments
Grant/funding Support: The author is in receipt of funding for clinical proteomics research from Cancer Research UK, Medical Research Council, Department of Trade and Industry, Kidney Research UK, Parkinson’s Disease Society, and AstraZeneca.
Financial Disclosures: None declared.
Acknowledgments: The funding support of the above bodies is gratefully acknowledged.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
