Kumar, Yogesh MRCS; Gurusamy, Kurinchi MRCS; Pamecha, Vineet MRCS; Davidson, Brian R. MD, FRCS
Pancreatic cancer accounts for 3% of all reported cases of cancer. It is the 11th most common cancer in the UK and the fifth leading cause of cancer death in Western countries.1,2 The prognosis is extremely poor with 5-year survival rate of less than 5%.3 Surgical resection offers a chance of cure, but only 10% to 20% of patients are eligible for resection.4 Early diagnosis to improve the dismal prognosis remains challenging. Most symptoms related to this malignancy occur only after disease advancement to an unresectable stage.5
Carbohydrate antigen 19-9 (CA19-9) has been proposed as a useful marker for pancreatic cancer.6,7 Elevated levels are associated with advanced disease at presentation and disease progression during follow-up.8 Confident discrimination between benign and malignant disease cannot be made on the basis of a solitary elevated CA19-9 (>33 U/mL) measurement.9 The role of currently used tumor markers in pancreatic cancer is limited. Recently, there has been an interest in tumor metabolism associated markers.10 The glycolytic enzyme pyruvate kinase type M2 isoform is present as a tetrameric form in all proliferating cells such as embryonic and adult stem cells and predominantly as a dimeric form, termed tumor M2-pyruvate kinase (TuM2-PK), in cancer cells.10,11 It has been shown to be elevated in the plasma of pancreatic cancer patients.12-19 The aim of this meta-analysis is to determine the utility of TuM2-PK as a tumor marker in exocrine pancreatic cancer and to compare the sensitivity and specificity of TuM2-PK and CA19-9.
MATERIALS AND METHODS
All studies in English language comparing the sensitivity and specificity of TuM2-PK with CA19-9 were considered for inclusion.
The following databases were searched:
1. PubMed (1951 to September 2006)
2. Embase (1974 to September 2006)
3. Central (Issue 3, 2006)
4. Science Citation Index (SCI) Expanded (1980 to September 2006)
References of identified studies were searched for identification of further references.
The following search was used for Pubmed:
("Pyruvate Kinase" [MeSH] or pyruvate kinase) and(m2 or "m-2" or "m 2") and("Neoplasms" [MeSH] or neoplasm or neoplasms orcancer or tumor or tumors or tumour or tumours andEnglish [lang]) and"humans" [MeSH Terms]
Equivalent search strategy was used for other databases.
Evaluation of Quality of Included Studies
Because the quality of the included studies can overestimate or underestimate the diagnostic odds ratios (DORs),20,21 information about the design (eg, case-control), differential verification (different reference test, eg, imaging, operation, histopathology performed based on the results of diagnostic test in question), blinding of assessors, description of population, description of diagnostic test, description of reference test (however, we did not expect the description of histopathology or imaging methods in any study), data collection (prospective or retrospective) was extracted to evaluate the quality of each study. These criteria can overestimate or underestimate the DOR.20
Because different studies used different cutoff values,12,14 it was intended that a meta-analysis of DOR22 was performed. For this purpose, from each study, the sensitivity, specificity, and number with and without disease were calculated. Because the sensitivity (a/(a + c)), people with pancreatic cancer (a + c), specificity (d/(b + d)), and people without pancreatic cancer (b + d) are reported in the studies, it was possible to calculate the true positivity (TP), false positivity (FP), false negativity (FN), and true negativity (TN) (a, b, c, and d, respectively, in Fig. 1). The DOR for each study was then calculated by using the formula23:
The 95% confidence interval (CI) of the DOR was calculated by finding out the antilog of the expression log DOR ± 1.96 SE(log DOR),23 where SE(log DOR) was calculated using the formula23
Equation (Uncited)Image Tools
The DOR and 95% CI calculated were pooled using the statistical software StatsDirect 126.96.36.199 The fixed effect model25 was used if the statistical heterogeneity measured by I2 as calculated by Higgins and Thompson26 was less than 25; otherwise, the random effects model27 was used. Bias was explored using funnel plot.28 Visual asymmetry was used to explore bias. The linear regression method of Egger et al28 was used for statistical evaluation of bias.
The summary estimate DOR and 95% CI were calculated for each of the 2 tests TuM2-PK and CA19-9 and compared to see whether one test was statistically superior to the other (P < 0.05) by evaluating the overlap of 95% CIs. If there was no significant variation in the diagnostic performance with threshold (ie, DOR is asymmetric or the DOR varies with the threshold used) as estimated by the Littenberg and Moses method,22 the equation for the receiver operator characteristic (ROC) curves was calculated using the relation22
Equation (Uncited)Image Tools
If there was a significant variation in the diagnostic performance with threshold, the equation for the ROC curve was calculated using the relation22
Equation (Uncited)Image Tools
where a and b are the estimates in the linear regression equation of Littenberg and Moses method.22
A subgroup analysis for pancreatic cancer versus healthy volunteers and pancreatic cancer versus benign pancreatic disorders was also performed.
A total of 258 references were retrieved from PubMed (78), Embase (75), Central (17), and SCI Expanded (88). One hundred eighteen duplicates were removed and 132 references were excluded by reading the titles and abstracts. Full text was obtained for 8 references12-19 of 7 studies. No further references were identified by searching the references of identified studies. All the 7 studies could provide data for meta-analysis.
The characteristics of the included studies are tabulated in Table 1. The quality of the included studies is stated in the Table 2. There were a total 303 pancreatic cancer patient and 1164 healthy controls in the included 7 studies.
The sensitivity, specificity, TP, FP, FN, TN, DOR, and 95% CIs are tabulated in Table 3. All the 7 studies reported the sensitivity and specificity of TuM2-PK, whereas only 3 studies15,17,19 included the specificity of CA19-9. Only one study13 reported the sensitivity and specificity of TuM2-PK at different cutoff levels (15 and 18 U/mL, respectively). Because the DOR was better for the cutoff level of 15 U/mL, the sensitivity and specificity corresponding to this cutoff level were used for the meta-analysis. One study19 reported the sensitivity and specificity of TuM2-PK in distinguishing pancreatic cancer from other gastrointestinal cancers, benign pancreatic disorders, and other benign gastrointestinal diseases. However, this study reported the number of people positive for TuM2-PK for different conditions individually, so it was possible to calculate the sensitivity and specificity of TuM2-PK in distinguishing pancreatic cancer from other benign pancreatic disease.
The meta-analysis of DOR for TuM2-PK is shown inFigure 2. Because I2 was 64.6%, the random effects model was used for the meta-analysis. The summary estimate of the DOR for TuM2-PK was 35.00 (95% CI, 19.66-62.30) and that of CA19-9 was 44.04 (95% CI, 26.53-73.10). There was significant overlap of the 95% CIs in the 2 tumor markers. The funnel plot is shown in Figure 3. Although visual inspection showed some asymmetry, there was no statistically significant bias (P = 0.6944).
There was significant variation in the diagnostic performance with threshold as estimated by the Littenberg and Moses method.22 The appropriate equation for the ROC was used to calculate the ROC. The ROC for TuM2-PK is shown in Figure 4. This corresponds to a sensitivity of 95% for a specificity of 60%.
A subgroup analysis to determine the role of TuM2-PK in identifying people with pancreatic cancer from healthy volunteers was performed. Five studies12-16 reported the sensitivity and specificity of TuM2-PK in pancreatic cancers compared with healthy volunteers. The DOR was 23.40 (95% CI, 15.48-35.36). There was no statistical heterogeneity because the I2 was 0%. Because there was significant variation in the diagnostic performance with threshold, the sensitivity and specificity were calculated by the Littenberg and Moses method. This corresponded to a sensitivity of 94% for a specificity of 55%.
Another subgroup analysis was performed to determine the role of TuM2-PK in identifying pancreatic cancer from benign pancreatic disorders. Only 2 studies17,19 reported this. Thus, a meta-analysis was not performed. The sensitivity and specificity were 85% and 35%, respectively, in the study19 that reported the number of people positive for TuM2-PK for different conditions individually. In the other study,17 the specificity and sensitivity were 66% and 60%, respectively.
Two studies16,19 reported the sensitivity and specificity of TuM2-PK alone and in combination with CA19-9. Of these 2, one study16 reported only the sensitivity without reporting the specificity. The combination of TuM2-PK and CA19-9 identified 96% of the cancers. In the other study,19 the combination identified pancreatic cancer with a sensitivity of 97%, which was higher than TuM2-PK (sensitivity, 85%) or CA19-9 (sensitivity, 75%) when used alone. The specificity was 38%. However, the later study included other malignancies also.
Tumor M2-pyruvate kinase is not an organ-specific cancer marker and is elevated in different type of cancers. Different specificity in the studies used in our meta-analysis is dependent on the type of control populations or the diagnostic threshold used. Because of the different thresholds used in the different studies (Table 1), DOR was calculated. Summary ROC is a useful method of meta-analysis when different studies use different thresholds for calculating the sensitivity and specificity of a diagnostic test.22 Because the ROC curve was asymmetric (identified by Littenberg and Moses method22), the appropriate formula was used to calculate the summary ROC curve. Most of the studies were of satisfactory quality with no significant bias in the evaluation of diagnostic test, although there was some asymmetry in the funnel plot, but it was not statistically significant. Histological confirmation was the reference test used in most of the studies based on which the sensitivity of TuM2-PK or CA19-9 was determined.
Different authors used different cutoff points to calculate the specificity and sensitivity. Hence, the Littenberg and Moses method22 was used to perform the meta-analysis. The main limitation of this meta-analysis is that the information about sensitivity and specificity is lost because of calculation of the DOR. Furthermore, the overall cutoff value for the test cannot be recommended. However, the main strength of this meta-analysis is that it has summarized the evidence in literature regarding the diagnostic utility of TuM2-PK. The Forest plot results are reproducible in different centers, with the DOR varying between 13 and 35 in 6 of the 7 studies.
Tumor M2-pyruvate kinase may not be useful for screening of general population but may be of more value in patients at high risk for pancreatic cancer. There are few high-risk groups that would be considered for screening, for example, Peutz-Jeghers syndrome, relatives of familial pancreatic cancer, hereditary pancreatitis, intraductal papillary mucinous neoplasms, and chronic pancreatitis.29 Familial pancreatic cancer groups are as rare30 as familial chronic pancreatitis, which has a high incidence of malignant transformation.31 Chronic pancreatitis has a small but significant high risk,32 but the role of TuM2-PK in distinguishing pancreatic exocrine cancers from benign pancreatic diseases is limited with only 2 studies17,19 comparing TuM2-PK levels in pancreatic exocrine cancers and benign pancreatic diseases.
The DOR was higher for CA19-9 than TuM2-PK, although the difference was not statistically significant. This meta-analysis, therefore, suggests that TuM2-PK has equivalent diagnostic utility to CA19-9 in distinguishing patients with pancreatic cancer from healthy individuals. A potential role for TuM2-PK would be in combination with CA19-9 in patients with jaundice, where elevated CA19-9 levels are elevated in the presence of biliary obstruction independent of the etiology. In such cases measurement of TuM2-PK would be of utility because its level in plasma is not influenced by elevated bilirubin levels.18,19 Because of the limited number of studies involving combination of markers, it was difficult to calculate the DOR for TuM2-PK and CA19-9 combination. However, these studies indicate that combining these 2 markers identify more patients with pancreatic cancer than when used alone.
Recently, several protein markers that are the product of altered gene expression in pancreatic cancer have been identified.33 These include macrophage inhibitory cytokine 1, synuclein-γ, mesothelin, osteopontin, and S100A4, which have been investigated as potential markers for pancreatic cancer, but their efficacy as serum markers remain undetermined.34-38 Tumor M2-pyruvate kinase, on the contrary, has a better reproducibility of diagnostic efficacy in pancreatic cancer either alone or in combination at different centers. This meta-analysis would suggest further trials comparing levels of TuM2-PK, either alone or in combination with CA19-9, in patients with exocrine pancreatic cancer and patients with high risk for pancreatic cancer.
Tumor M2-pyruvate kinase has a potential role as a marker of pancreatic cancer. It can reliably distinguish healthy individuals from patients with pancreatic cancer. Its efficacy as a tumor marker is similar to that of CA19-9. Further trials are needed on TuM2-PK, used alone or in combination with CA19-9, in patients with suspected pancreatic cancer.
1. Rosewicz S, Wiedenmann B. Pancreatic carcinoma. Lancet
3. Wood HE, Gupta S, Kang JY, et al. Pancreatic cancer in England and Wales 1975-2000: patterns and trends in incidence, survival and mortality. Aliment Pharmacol Ther
4. Vogt DP. Pancreatic cancer: a current overview. Curr Surg
5. Kalser MH, Barkin J, MacIntyre JM. Pancreatic cancer. Assessment of prognosis by clinical presentation. Cancer
6. Kau SY, Shyr YM, Su CH, et al. Diagnostic and prognostic values of CA 19-9 and CEA in periampullary cancers. J Am Coll Surg
7. Osswald BR, Klee FE, Wysocki S. The reliability of highly elevated CA 19-9 levels. Dis Markers
8. McLaughlin R, O'Hanlon D, Kerin M, et al. Are elevated levels of the tumour marker CA19-9 of any clinical significance?-An evaluation. Ir J Med Sci
9. Mann DV, Edwards R, Ho S, et al. Elevated tumour marker CA19-9: clinical interpretation and influence of obstructive jaundice. Eur J Surg Oncol
10. Mazurek S, Boschek CB, Hugo F, et al. Pyruvate kinase type M2 and its role in tumor growth and spreading. Semin Cancer Biol
. 2005;15: 300-308.
11. Mazurek S, Grimm H, Boschek CB, et al. Pyruvate kinase type M2: a crossroad in the tumor metabolome. Br J Nutr
. 2002;87(suppl 1):S23-S29.
12. Cerwenka H, Aigner R, Bacher H, et al. TUM2-PK (pyruvate kinase type tumor M2), CA19-9 and CEA in patients with benign, malignant and metastasizing pancreatic lesions. Anticancer Res
13. Hardt PD, Ngoumou BK, Rupp J, et al. Tumor M2-pyruvate kinase: a promising tumor marker in the diagnosis of gastro-intestinal cancer. Anticancer Res
14. Oremek GM, Eigenbrodt E, Radle J, et al. Value of the serum levels of the tumor marker TUM2-PK in pancreatic cancer. Anticancer Res
15. Schneider J, Schulze G. Comparison of tumor M2-pyruvate kinase (tumor M2-PK), carcinoembryonic antigen (CEA), carbohydrate antigens CA 19-9 and CA 72-4 in the diagnosis of gastrointestinal cancer. Anticancer Res
16. Schulze G. The tumor marker tumor M2-PK: an application in the diagnosis of gastrointestinal cancer. Anticancer Res
. 2000;20: 4961-4964.
17. Siriwardana HPP, King NKK, France MW, et al. Plasma tumour-M2-pyruvate kinase and serum carbohydrate antigen for the preoperative diagnosis of pancreatic cancer. Br J Surg
. 2005;92 (suppl 1):100. [abstract].
18. Siriwardana P, King N, France M, et al. Prospective evaluation of the diagnostic accuracy of plasma tumour M2-pyruvate kinase (TuM2-PK) and serum carbohydrate antigen (CA19-9) in periampullary cancer. Gastroenterology
. 2005;128(suppl 2):A472-A473. [abstract].
19. Ventrucci M, Cipolla A, Racchini C, et al. Tumor M2-pyruvate kinase, a new metabolic marker for pancreatic cancer. Dig Dis Sci
20. Lijmer JG, Mol BW, Heisterkamp S, et al. Empirical evidence of design-related bias in studies of diagnostic tests. JAMA
21. Rutjes AW, Reitsma JB, Di NM, et al. Evidence of bias and variation in diagnostic accuracy studies. CMAJ
22. Deeks J. Systematic reviews of evaluations of diagnostic and screening tests. In: Egger M, Smith GD, Altman DG, eds. Systematic Reviews in Health Care: Meta-analysis in context
. 2nd ed. London: BMJ Publishing Group; 2001:248-282.
23. Glas AS, Lijmer JG, Prins MH, et al. The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol
. 2003;56: 1129-1135.
24. StatsDirect Statistical Software Version 2.5.6
. Cheshire, UK: StatsDirect; 2005.
25. Demets DL. Methods for combining randomized clinical trials: strengths and limitations. Stat Med
26. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med
27. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials
28. Egger M, Davey SG, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ
29. Canto MI, Goggins M, Yeo CJ, et al. Screening for pancreatic neoplasia in high-risk individuals: an EUS-based approach. Clin Gastroenterol Hepatol
30. Habbe N, Langer P, Sina-Frey M, et al. Familial pancreatic cancer syndromes. Endocrinol Metab Clin North Am
31. Lowenfels AB, Maisonneuve P, Dimagno EP, et al. Hereditary pancreatitis and the risk of pancreatic cancer. International Hereditary Pancreatitis Study Group. J Natl Cancer Inst
32. Lowenfels AB, Maisonneuve P, Cavallini G, et al. Pancreatitis and the risk of pancreatic cancer. International Pancreatitis Study Group. N Engl J Med
33. Okusaka T, Yamada T, Maekawa M. Serum tumor markers for pancreatic cancer: the dawn of new era? JOP
34. Argani P, Iacobuzio-Donahue C, Ryu B, et al. Mesothelin is overexpressed in the vast majority of ductal adenocarcinomas of the pancreas: identification of a new pancreatic cancer marker by serial analysis of gene expression (SAGE). Clin Cancer Res
35. Koopmann J, Buckhaults P, Brown DA, et al. Serum macrophage inhibitory cytokine 1 as a marker of pancreatic and other periampullary cancers. Clin Cancer Res
36. Koopmann J, Rosenzweig CN, Zhang Z, et al. Serum markers in patients with resectable pancreatic adenocarcinoma: macrophage inhibitory cytokine 1 versus CA19-9. Clin Cancer Res
37. Li Z, Sclabas GM, Peng B, et al. Overexpression of synuclein-gamma in pancreatic adenocarcinoma. Cancer
38. Rosty C, Ueki T, Argani P, et al. Overexpression of S100A4 in pancreatic ductal adenocarcinomas is associated with poor differentiation and DNA hypomethylation. Am J Pathol
© 2007 Lippincott Williams & Wilkins, Inc.