Clinica Chimica Acta 438 (2015) 418–419
Contents lists available at ScienceDirect
Clinica Chimica Acta journal homepage: www.elsevier.com/locate/clinchim
Letter to the Editor A comment to “Normalization of urinary pteridines by urine specific gravity for early cancer detection” [Clin. Chim. Acta 435 (2014) 42–47]
Dear Editor, In their recent publication [1], Burton et al. described a comparison of urinary levels of several pteridine derivatives normalized with respect to urine concentration by either urine specific gravity (USG) or by relation to urinary creatinine concentration. They conclude that correction of urinary pteridine concentrations by USG provides better performance of these biomarkers in early detection of cancer than creatinine-based correction. We comment on this study with respect both to analytical/biological content and formal statistical aspects. Of all the pteridine derivatives investigated, only neopterin has been shown by many authors to provide valuable diagnostic and prognostic information in patients with a wide variety of malignant diseases [2]. As neopterin is produced by human monocytes, e.g., macrophages and dendritic cells upon the action of T-cell derived cytokines like interferon gamma, elevated neopterin concentrations in body fluids are a sensitive marker for an activated cellular immune system [3] and hence, they do not constitute a classical tumor marker but rather reflect the interaction of the host's immune system with the disease. Biopterin in urine might stem from unspecific oxidation of the essential cofactor 5,6,7,8tetrahydrobiopterin or its dihydro derivatives, which are produced by cells other than monocytes including tumor cells but also under the control of pro-inflammatory cytokines. Unspecific oxidation processes of reduced pteridine species are likely responsible for the occurrence of xanthopterin, isoxanthopterin and others. While the authors use an oxidizing solution before analysis of their urine samples, unspecific oxidation processes are suspected to have already occurred in the time interval between urine voiding and analysis. So as the authors correctly state, there is no biological explanation at hand which could explain the apparent correlation between presence of aggressive breast tumors and raised xanthopterin and isoxanthopterin. Furthermore, it is well known since many years that breast cancer is among those malignancies which are associated only very rarely with raised neopterin levels (however; even in such patients, neopterin concentrations bear prognostic information) [4]. In the light of these earlier results, the failure to detect raised neopterin in this study is by no means surprising. Besides these analytical/biological limitations, we are strongly concerned about the statistical quality of the study. The numbers of patients with benign (N = 27) and aggressive breast cancers (N = 21) is quite low, and the statistical conclusions drawn appear doubtful: the authors claim significantly higher USG-corrected levels of isoxanthopterin (P = 0.0011) and xanthopterin (P = 0.0090) in patients with aggressive breast cancer using Mann–Whitney U-test, which they argue to prevent erroneous inflation of statistical significance by Student's t
http://dx.doi.org/10.1016/j.cca.2014.08.032 0009-8981/© 2014 Elsevier B.V. All rights reserved.
test. However, one-sided Student's t testing, using the data of their Table 2, in fact would yield less optimistic P values than those reported in their Table 1: we obtain P = 0.0141 for isoxanthopterin and P = 0.0337 for xanthopterin. While this is only a minor technical aspect, we are particularly concerned about the detailed data shown in their figures: from Fig. 1 we conclude that 3 of 21 patients with aggressive cancers had USG-corrected xanthopterin exceeding 15 μM, in contrast to 0 of 27 patients with benign tumors. Fisher's exact test applied to the corresponding 2 × 2-contingency table yields P = 0.0425; however, would only one of the 27 women with benign tumors have exhibited a value above 15 μM, the significance of the then resulting contingency table would collapse to a mere P = 0.19. For isoxanthopterin (their Fig. 2) we would obtain exactly the same result if we used a cut-off value of 2 μM. Thus, the claim of significant elevations of the USG-corrected concentrations of both xanthopterin and isoxanthopterin rests on very slippery grounds. Our strongest objections, however, are due to the most “unconventional” use and interpretation of the P-values shown in their Table 1: none of the studied uncorrected as well as creatinine-corrected pteridine concentrations differed significantly between patients with aggressive versus benign tumors, and as mentioned before, only USG-corrected levels of xanthopterin and isoxanthopterin produced a P-value less than 0.05. The authors then chose to sum up the – mostly insignificant – Pvalues of the three different methods, and used these summed values for a comparison among them. Bearing in mind what a P-value truly does (measuring the probability to obtain the observed data by pure chance, i.e., under the null hypothesis) it becomes very clear that such a usage of P-values is absolutely inappropriate, and any conclusions drawn are meaningless: how should we reasonably interpret a probability (summed P-value) of, say, 2.0067? We propose to view these criticisms in a larger framework: since 2005 there is a growing concern in the scientific community about the frequently inappropriate usage of P values; in particular it was stated that many published “statistically significant” results in fact may be wrong [5] because scientists expect from P values what they simply cannot afford. Very recently, a most readable article on this subject was published [6] which was recommended to be read by every serious scientist [7].
References [1] Burton C, Shi H, Ma Y. Normalization of urinary pteridines by urine specific gravity for early cancer detection. Clin Chim Acta 2014;435:42–7. [2] Reibnegger G, Fuchs D, Fuith LC, et al. Neopterin as a marker for activated cellmediated immunity: application in malignant disease. Cancer Detect Prev 1991;15: 483–90. [3] Fuchs D, Hausen A, Reibnegger G, Werner ER, Dierich MP, Wachter H. Neopterin as a marker for activated cell-mediated immunity: application in HIV infection. Immunol Today 1988;9:150–5. [4] Murr C, Bergant A, Widschwendter M, Heim K, Schröcksnadel H, Fuchs D. Neopterin is an independent prognostic variable in females with breast cancer. Clin Chem 1999; 45:1998–2004. [5] Joannidis JPA. Why most published research findings are false. PLoS Med 2005;2: e124.
Letter to the Editor [6] Nuzzo R. Statistical errors. P values, the ‘gold standard’ of statistical validity, are not as reliable as many scientists assume. Nature 2014;506:150–2. [7] Boyd JP, Annesley TM. To P or not to P: that is the question. Clin Chem 2014;60: 909–10.
Gilbert Reibnegger Institute of Physiological Chemistry, Center of Physiological Medicine, Medical University of Graz, Harrachgasse 21/II, A-8010 Graz, Austria Corresponding author.
419
Dietmar Fuchs Division of Biological Chemistry, Biocenter, Innsbruck Medical University, Center for Chemistry and Biomedicine, Innrain 80, 4th Floor, Room M04-313, A-6020 Innsbruck, Austria 28 July 2014
Contents lists available at ScienceDirect
Clinica Chimica Acta journal homepage: www.elsevier.com/locate/clinchim
Letter to the Editor A comment to “Normalization of urinary pteridines by urine specific gravity for early cancer detection” [Clin. Chim. Acta 435 (2014) 42–47]
Dear Editor, In their recent publication [1], Burton et al. described a comparison of urinary levels of several pteridine derivatives normalized with respect to urine concentration by either urine specific gravity (USG) or by relation to urinary creatinine concentration. They conclude that correction of urinary pteridine concentrations by USG provides better performance of these biomarkers in early detection of cancer than creatinine-based correction. We comment on this study with respect both to analytical/biological content and formal statistical aspects. Of all the pteridine derivatives investigated, only neopterin has been shown by many authors to provide valuable diagnostic and prognostic information in patients with a wide variety of malignant diseases [2]. As neopterin is produced by human monocytes, e.g., macrophages and dendritic cells upon the action of T-cell derived cytokines like interferon gamma, elevated neopterin concentrations in body fluids are a sensitive marker for an activated cellular immune system [3] and hence, they do not constitute a classical tumor marker but rather reflect the interaction of the host's immune system with the disease. Biopterin in urine might stem from unspecific oxidation of the essential cofactor 5,6,7,8tetrahydrobiopterin or its dihydro derivatives, which are produced by cells other than monocytes including tumor cells but also under the control of pro-inflammatory cytokines. Unspecific oxidation processes of reduced pteridine species are likely responsible for the occurrence of xanthopterin, isoxanthopterin and others. While the authors use an oxidizing solution before analysis of their urine samples, unspecific oxidation processes are suspected to have already occurred in the time interval between urine voiding and analysis. So as the authors correctly state, there is no biological explanation at hand which could explain the apparent correlation between presence of aggressive breast tumors and raised xanthopterin and isoxanthopterin. Furthermore, it is well known since many years that breast cancer is among those malignancies which are associated only very rarely with raised neopterin levels (however; even in such patients, neopterin concentrations bear prognostic information) [4]. In the light of these earlier results, the failure to detect raised neopterin in this study is by no means surprising. Besides these analytical/biological limitations, we are strongly concerned about the statistical quality of the study. The numbers of patients with benign (N = 27) and aggressive breast cancers (N = 21) is quite low, and the statistical conclusions drawn appear doubtful: the authors claim significantly higher USG-corrected levels of isoxanthopterin (P = 0.0011) and xanthopterin (P = 0.0090) in patients with aggressive breast cancer using Mann–Whitney U-test, which they argue to prevent erroneous inflation of statistical significance by Student's t
http://dx.doi.org/10.1016/j.cca.2014.08.032 0009-8981/© 2014 Elsevier B.V. All rights reserved.
test. However, one-sided Student's t testing, using the data of their Table 2, in fact would yield less optimistic P values than those reported in their Table 1: we obtain P = 0.0141 for isoxanthopterin and P = 0.0337 for xanthopterin. While this is only a minor technical aspect, we are particularly concerned about the detailed data shown in their figures: from Fig. 1 we conclude that 3 of 21 patients with aggressive cancers had USG-corrected xanthopterin exceeding 15 μM, in contrast to 0 of 27 patients with benign tumors. Fisher's exact test applied to the corresponding 2 × 2-contingency table yields P = 0.0425; however, would only one of the 27 women with benign tumors have exhibited a value above 15 μM, the significance of the then resulting contingency table would collapse to a mere P = 0.19. For isoxanthopterin (their Fig. 2) we would obtain exactly the same result if we used a cut-off value of 2 μM. Thus, the claim of significant elevations of the USG-corrected concentrations of both xanthopterin and isoxanthopterin rests on very slippery grounds. Our strongest objections, however, are due to the most “unconventional” use and interpretation of the P-values shown in their Table 1: none of the studied uncorrected as well as creatinine-corrected pteridine concentrations differed significantly between patients with aggressive versus benign tumors, and as mentioned before, only USG-corrected levels of xanthopterin and isoxanthopterin produced a P-value less than 0.05. The authors then chose to sum up the – mostly insignificant – Pvalues of the three different methods, and used these summed values for a comparison among them. Bearing in mind what a P-value truly does (measuring the probability to obtain the observed data by pure chance, i.e., under the null hypothesis) it becomes very clear that such a usage of P-values is absolutely inappropriate, and any conclusions drawn are meaningless: how should we reasonably interpret a probability (summed P-value) of, say, 2.0067? We propose to view these criticisms in a larger framework: since 2005 there is a growing concern in the scientific community about the frequently inappropriate usage of P values; in particular it was stated that many published “statistically significant” results in fact may be wrong [5] because scientists expect from P values what they simply cannot afford. Very recently, a most readable article on this subject was published [6] which was recommended to be read by every serious scientist [7].
References [1] Burton C, Shi H, Ma Y. Normalization of urinary pteridines by urine specific gravity for early cancer detection. Clin Chim Acta 2014;435:42–7. [2] Reibnegger G, Fuchs D, Fuith LC, et al. Neopterin as a marker for activated cellmediated immunity: application in malignant disease. Cancer Detect Prev 1991;15: 483–90. [3] Fuchs D, Hausen A, Reibnegger G, Werner ER, Dierich MP, Wachter H. Neopterin as a marker for activated cell-mediated immunity: application in HIV infection. Immunol Today 1988;9:150–5. [4] Murr C, Bergant A, Widschwendter M, Heim K, Schröcksnadel H, Fuchs D. Neopterin is an independent prognostic variable in females with breast cancer. Clin Chem 1999; 45:1998–2004. [5] Joannidis JPA. Why most published research findings are false. PLoS Med 2005;2: e124.
Letter to the Editor [6] Nuzzo R. Statistical errors. P values, the ‘gold standard’ of statistical validity, are not as reliable as many scientists assume. Nature 2014;506:150–2. [7] Boyd JP, Annesley TM. To P or not to P: that is the question. Clin Chem 2014;60: 909–10.
Gilbert Reibnegger Institute of Physiological Chemistry, Center of Physiological Medicine, Medical University of Graz, Harrachgasse 21/II, A-8010 Graz, Austria Corresponding author.
419
Dietmar Fuchs Division of Biological Chemistry, Biocenter, Innsbruck Medical University, Center for Chemistry and Biomedicine, Innrain 80, 4th Floor, Room M04-313, A-6020 Innsbruck, Austria 28 July 2014
