Proteomics: Promise and ProblemsBy Jennifer H. Miller and Barnett S. Kramer
Early-detection programs have great intuitive appeal in oncology: Identifying a cancer in its infancy and "nipping it in the bud" appears a self-evident method for reducing morbidity and mortality....
However, we must not confuse early diagnosis with improved outcome. It has been observed that screen-detected cancers do tend, on average, to be earlier-stage malignancies, partly because more aggressive lesions are more likely to produce symptoms within a shorter period of time (Figure 1). Additionally, 5-year survival rates for stage I cancers are typically greater than for stage III or IV. This simply means the staging system works; survival rates are not equivalent to mortality rates.
Five-year survival is the proportion of individuals alive five years after a diagnosis of cancer, out of all individuals with the disease. Mortality is the proportion of individuals dead due to a given cancer in an entire at-risk population. Furthermore, although early detec-tion, by its very purpose, shifts the date of diagnosis, it does not necessarily follow that the date of demise shifts as well (Figure 2). A reduction in mortality as a result of a screening exam must be directly demonstrated; it cannot merely be inferred by stage at detection.
The concepts of overdiagnosis and pseudodisease are pertinent here: A peptidomic screening tool may largely diagnose abnormalities with no true potential for harm, either because the cancer would take longer to mature than the person's remaining lifespan, or because the abnormality did not possess the capability to develop into a dangerous lesion. The ultimate purpose of screening is therefore not to diag-nose a disease earlier, but to help the person live longer.
Evidence suggests that cancer screening can disengage any association between cancer survival and cancer mortality. Consider the national screening program for neuroblastoma that Japan instituted in 1985. Urinary VMA testing was effective in identifying more cases than was clinical diagnosis alone: Cases of early-stage (I or II) neuroblastoma rose several-fold. However, the number of late-stage cases identified did not decline, and mortality due to neuroblastoma also did not decrease.2 It has been hypothesized that the asymptomatic tumors identified by screening tended to be indolent, spontaneously regressing abnormalities - fundamentally different in nature from the more aggressive (and familiar) symptomatic variety.3 Furthermore, it has been shown that in the United States, no cor-relation at all exists between trends in 5-year cancer survival rates and cancer mortality from 1950 to 1996.4
Setting aside ongoing issues concerning the reproducibility of previously published results,5-8 while high specificity and sensitivity are of course necessary elements, again, they are not the true endpoint of interest. Validating the specificity of a peptidomic test at 99.99%, while admirable, does not prove the worth of the test; only proving a reduction in mortality from the cancer in question can do that.
The utility of proteomics in cancer is likely to remain a theoretical construct and a hope for now: Peptidomes have a host of other obstacles to overcome before their impact on cancer mortality can even be tested. Petricoin and Liotta have presented a logical expla-nation of why peptidomic screening should work. The logic still begs empiric proof (or disproof). The field of proteomics is in its infancy and is a work in progress, albeit promising. While intuitively appealing, validation of reproducibility, and ultimately of benefit, still lies ahead. Early detection is not automatically synonymous with better care
1. E.F. Petricoin, L.A. Liotta, "Proteomic analysis at the bedside: early detection of cancer," Trends Biotechnol, 20:S30-S34, 2002.[Pubmed]
2. S. Honjo et al., "Neuroblastoma trends in Osaka, Japan, and Great Britain, 1970-1994, in relation to screening," Int J Cancer, 103:538-43, 2003.[Pubmed]
3. J.A. Ross, S.M. Davies, "Screening for neuroblastoma: progress and pitfalls," Cancer Epidemiol Biomarkers Prev, 8:189-94, 1999.[Pubmed]
4. H.G. Welch et al., "Are increasing 5-year survival rates evidence of success against cancer?" J Am Med Assoc, 283:2975-8, 2003.[Pubmed]
5. E.P. Diamandis, "Analysis of serum proteomic patterns for early cancer diagnosis: drawing attention to potential problems," J Nat Cancer Inst, 96:353-6, 2004.[Pubmed]
6. K.A. Baggerly et al., "Signal in noise: evaluating reported reproducibility of serum proteomic tests for ovarian cancer," J Nat Cancer Inst, 97:307-9, 2005.[Pubmed]
7. D.F. Ransohoff, "Lessons from controversy: ovarian cancer screening and serum proteomics," J Nat Cancer Inst, 97:315-9, 2005.[Pubmed]
8. D.F. Ransohoff, "Bias as a threat to the validity of cancer molecular-marker research," Nat Rev Cancer, 5:142-9, 2005.[Pubmed]