WIXLIB

Int. J. Med. Sci. 2018, Vol. 15An in-person interview and blood samplingwere performed prior to the histopathologicalconfirmation of the diagnosis of MPM. Aquestionnaire was applied by trained interviewers toassess socio-demographic information, a detailedoccupational history, exposure to asbestos, smokinghabits, medical history, and other data. Lifetimeoccupational exposure to asbestos was categorized asever or never according to a previously publishedassessment [21]. In brief, an industrial hygienist, whowas unaware of the case-control status, estimated theexposure to asbestos according to the worker’s jobhistory and a list with information on industriesimporting asbestos or companies that manufacturedasbestos fibers in various forms [22] along withrecognized occupational activities and jobs withexposure to asbestos [23-28].The research protocol was approved by IMSS’National Commission for Scientific Research andEthics with registration number R-2011-785-069, andby INER’s Committee on Science and Bioethics inResearch with registration number C30-12. Prior toinclusion in the study, participants signed a letter ofinformed consent.Measurement of Tumor Markers in PlasmaBlood samples were drawn in three 6.0 ml tubes(EDTA vacutainers) and centrifuged at 2,250 g for 10minutes in a laminar flow hood. Plasma and buffycoat were separated and stored at -80 C until analysis.Mesothelin and thrombomodulin ELISAs wereperformed at CINVESTAV in Mexico City, Mexico.For mesothelin determination, sandwich ELISAs wereperformed according to the manufacturer sinstructions (DY3265, R&D Systems, Minneapolis,MN). Thrombomodulin concentration was determined according to the manufacturer s instructions(DY3947, R&D Systems).Plasma samples were shipped to Germanyunder stringent frozen conditions and calretinindetermination was performed utilizing a sandwichELISA according to Raiko et al. [11] at the IPA inBochum, Germany.Statistical ns were presented by median andinterquartile range (IQR). A relatively large number ofcalretinin concentrations were below the limit ofdetection (LOD). Therefore, affected percentiles weremarked as being less ( ) than the respective LOD(Table 1). For the calculations, we set values belowLOD to two-thirds of LOD (2/3*LOD). Mesothelinand thrombomodulin concentrations between cases885and controls were compared using the Wilcoxonrank-sum test. Because 72% of calretininmeasurements were below LOD, the Peto-Prenticetest was applied to compare cases and controls [29,30].Odds ratios (ORs) and 95% confidence intervals(CI) were calculated using unconditional logisticregression models to estimate the relative risk of adiagnosis of MPM based on the log-transformedconcentrations of the tumor markers. These modelswere also used for estimating receiver operatingcharacteristic (ROC) curves and the area under thecurve (AUC) for the markers’ sensitivity for varyingvalues of specificity. Based on the performance of themarkers we predicted the individual probability of adiagnosis of MPM for a male or female subject withthe corresponding estimates of the true positive rate(TPR) and false positive rate (FPR) by these logisticregression models. Statistical analyses wereperformed using SAS software, version 9.4 (SASInstitute Inc., Cary, NC).ResultsTable 1 depicts the characteristics of the casesand controls. Median age of MPM diagnosis was 64years for males and 62 years for females. More womenthan men had never smoked (women: 66.7% in cases,64.7% in controls; men: 33.3% in cases, 37.8% incontrols). A large fraction of men was classified asever exposed to asbestos at the workplace (93.7% incases, 70.3% in controls). The corresponding figures inwomen were 33.3% and 17.7%, respectively.Epithelioid MPM was diagnosed for all twelve femalecases and for 52 (82.5%) male cases.Table 2 shows the distribution of the markerconcentrations by case-control status in men andwomen. The concentrations of mesothelin andcalretinin but not of thrombomodulin werestatistically significant higher in cases than in controls,with distinct IQRs. In men, mesothelin medianconcentrations were 2.21 nmol/l in cases and 0.58nmol/l in controls (p 0.0001). Similar concentrationswere determined in women (2.00 nmol/l and 0.55nmol/l, respectively, p 0.0001). The same patternwas observed for calretinin (men: 0.93 ng/ml in cases, 0.07 ng/ml in controls, p 0.0001; women: 0.86ng/ml in cases, 0.22 ng/ml in controls, p 0.0037).No differences in thrombomodulin concentrationswere observed between cases and controls (men: 2.57vs. 2.30 ng/ml, p 0.81; women: 2.25 vs. 2.05 ng/ml,p 0.84). In addition, we show the distribution of theconcentrations in subgroups stratified by age andexposure to asbestos.http://www.medsci.org

Int. J. Med. Sci. 2018, Vol. 15886Table 1. Description of the study population of cases with MPM and controls from MexicoCharacteristicsTotalAge (years)Median (IQR)Histologic diagnosisEpithelioidSarcomatoidOtherNot specifiedSmoking statusNever smokerFormer/current smokerOccupational exposure to asbestosNeverEveraMaleCases N (%)63Controls N (%)172P-valueaFemaleCases N (%)12Controls N (%)68P-valuea64 (56-72)62 (55-71)0.5962 (53-68)61 (53-70)0.8252 (82.5)3 (4.8)7 (11.1)1 (1.6)12 (100)0.000321 (33.3)42 (66.7)65 (37.8)107 (62.2)4 (6.3)59 (93.7)51 (29.7)121 (70.3)8 (66.7)4 (33.3)44 (64.7)24 (35.3)8 (66.7)4 (33.3)56 (82.4)12 (17.7) 0.00010.24Continuous variables were compared using the Wilcoxon rank-sum test, categorical variables were compared using Fisher’s exact test. IQR: interquartile rangeTable 2. Distribution of biomarker concentrations in plasma samples from cases with MPM and controls from 0.580.520.760.40-0.870.38-0.760.48-1.02 .790.45-0.87MalesAge 70 years 70 yearsOccupational exposure to asbestosNeverEver6240220.931.000.85 0.45-2.18 0.31-2.500.58-1.9817212052 0.07 0.08 0.06 0.01- 0.22 0.01- 0.23 0.01- 0.224581.410.86 0.74-2.16 0.45-2.1851121 0.06 0.08 0.01- 0.20 0.01- 0.24FemalesAge 70 years 70 yearsOccupational exposure to asbestosNeverEver12930.86 0.353.09 0.25-1.33 0.24-1.120.97-4.98685018 0.22 0.22 0.18 0.10-0.48 0.12-0.48 0.03- 0.46841.04 0.25 0.55-2.16 0.13- 0.855612 0.22 0.27 0.10-0.48 0.08- 0.47MalesAge 70 years 70 yearsOccupational exposure to .70-2.871.86-2.83FemalesAge 70 years 70 yearsOccupational exposure to 2.721.26-2.66Mesothelin (nmol/L)MalesAge 70 years 70 yearsOccupational exposure to asbestosNeverEverFemalesAge 70 years 70 yearsOccupational exposure to asbestosNeverEver 0.0001Calretinin (ng/mL) 0.00010.0037Thrombomodulin (ng/mL)0.810.84aMesothelin and thrombomodulin concentrations were compared using the Wilcoxon rank-sum test, calretinin concentrations were compared using the Peto-Prentice test.IQR: interquartile range; ‘ ’ indicates percentiles below the limit of detection (LOD)http://www.medsci.org

Int. J. Med. Sci. 2018, Vol. 15Table 3 presents the ORs with 95% CI asestimates of the relative risk of an MPM per unitincrease of the log-transformed concentrations of themarkers. Whereas thrombomodulin was notassociated with the development of MPM in men(OR 1.05, 95% CI 0.44-2.48) and women (OR 1.02,95% CI 0.19-5.47), both mesothelin and calretinin weresignificant factors. The MPM risks per unit increase ofthe mesothelin and calretinin concentrations were14.51 (95% CI 6.96-30.28) and 3.03 (95% CI 2.20-4.18) inmen and 28.57 (95% CI 4.09-199.4) and 2.69 (95% CI1.31-5.56) in women, respectively. In the multivariateanalysis mesothelin appears to be more clearlyassociated with MPM risk than calretinin for malesand females.887Using ROC analyses, the calculated AUCs for themesothelin concentrations were 0.90 (95% CI0.85-0.95) in men and 0.92 (95% CI 0.79-1.00) inwomen. The AUCs for calretinin were slightlyweaker, with 0.88 (95% CI 0.82-0.94) in men and 0.77(95% CI 0.61-0.93) in women (Figure 1). Due to thelow AUC, thrombomodulin was excluded fromfurther analysis. Based on the plasma concentrationsof mesothelin and calretinin we predicted theindividual probability of an MPM in a male or femalesubject using the prediction formulas presented inTables 4 and 5, respectively, resulting in estimatedTPRs and FPRs for the whole study populationstratified to probabilities ranging from 10% to 90%.Table 3. Odds ratios from logistic regression analyses as estimates of the relative risk for an MPM based on the plasma concentrations incases and controlsStudy n) )[ng/mL]InterceptCoefficientOR (95% CI)-1.150.55-0.872.681.110.0414.51 (6.96-30.28)3.03 (2.20-4.18)1.05 (0.44-2.48)-0.172.110.598.26 (3.77-18.10)1.80 (1.31-2.47)-1.77-0.72-1.863.350.990.0228.57 (4.09-199.4)2.69 (1.31-5.56)1.02 (0.19-5.47)-1.363.080.3321.86 (3.13-152.51)1.39 (0.67-2.88)OR: odds ratio; CI: confidence interval; ln: natural logarithmFigure 1. ROC curves of MPM biomarkers in incident cases and controls by gender (all CIs were 95%). A, Males: mesothelin (AUC 0.90, CI:0.85-0.95), calretinin (AUC 0.88,CI:0.82-0.94), and thrombomodulin (AUC 0.51, CI:0.41-0.61) B, Females: mesothelin (AUC 0.92, CI:0.79-1.00), calretinin (AUC 0.77, CI:0.61-0.93), and thrombomodulin(AUC 0.52, CI: 0.32-0.72).http://www.medsci.org

Int. J. Med. Sci. 2018, Vol. 15888Table 4. Performance of mesothelin and calretinin in men and the probability of a diagnosis of MPM, conditional on the observedbiomarker concentrationsProbability Males%Model 1Cut off Mesothelin 0FPR0.010.010.010.030.030.060.110.220.40Model 2Cut off Calretinin .100.180.340.49Model 3Cut off Mesothelin [nmol/L]3.662.151.081.460.740.971.010.651.31Cut off Calretinin 050.070.110.140.33Probability was used to estimate TPR and FPR: Probability 1/(1 e (-φ)). Logistic regression models: (1) with log-mesothelin as predictor, φ exp [-1.15 2.68 *ln(mesothelin)]; (2) with log-calretinin as predictor, φ exp [0.55 1.11 * ln(calretinin)]; (3) with log-mesothelin and log-calretinin, φ exp [-0.17 2.11 * ln(mesothelin) 0.59 *ln(calretinin)]. TPR: true positive rate; FPR: false positive rateTable 5. Performance of mesothelin and calretinin in women and the probability of a diagnosis of MPM, conditional on the observedbiomarker concentrationsProbability Females%Model 1Cut off Mesothelin 020.020.030.030.120.20Model 2Cut off Calretinin 250.500.580.92FPR0000000.060.240.49Model 3Cut off Mesothelin [nmol/L]2.131.351.211.050.90Cut off Calretinin robability was used to estimate TPR and FPR: Probability 1/(1 e (-φ)). Because of the small number of female cases (12), not for all set probabilities corresponding markerconcentrations were available. Logistic regression models: (1) with log-mesothelin as predictor, φ exp [-1.77 3.35 * ln(mesothelin)]; (2) with log-calretinin as predictor, φ exp [-0.72 0.99 * ln(calretinin)]; (3) with log-mesothelin and log-calretinin, φ exp [-1.36 3.08 * ln(mesothelin) 0.33 * ln(calretinin)]. TPR: true positive rate; FPR: falsepositive rateUtilizing mesothelin alone a pre-defined 30%probability for the diagnosis of MPM in an individualwas associated with high TPRs in the studypopulation (0.79 in men and 0.75 in women) andmoderate FPRs (0.11 in men and 0.03 in women). Forcomparison, a high individual probability of 80% wasassociated with a moderate TPR of 0.43 and a low FPRof 0.01 in men and similar rates in women. Anincrease of the TPR, with stable FPR, could beachieved in men by including calretinin into thedecision model (30% probability: TPR 0.84, 80%probability: TPR 0.53). Both markers are moderatelycorrelated in plasma samples of cases (Kendall’s taucorrelation coefficient: 0.37, p 0.0001). The cut offvalues in Tables 4 and 5 can be used to apply thebiomarkers to different tasks/issues, e.g., screening oroptimizing histopathological diagnostic workup.DiscussionMPM is an aggressive cancer and difficult todiagnose, particularly at early stages. Reliable tumormarkers could improve the diagnosis of MPM,including early stages [11, 31, 32]. In some countrieslike Mexico, the resources for histopathologicaldiagnosis of suspected cases are limited. MPM is alsoa rare disease, with a very low probability of anindividual suffering from this type of cancer.Additional information on the blood concentrations oftumor markers can enhance an individual’sprobability. Here, we provided a prediction model forthe probability of a diagnosis of MPM ranging from10% to 90% in order to optimize the histopathologicaldiagnostic workup of tissue samples.Formulas given in Tables 4 and 5 can be used tocalculate the probability of a subject for a diagnosis ofMPM by gender for given plasma concentrations ofmesothelin and calretinin. A high individualprobability of a subject such as 90% is associated witha high tumor marker level. This high tumor markerconcentration implies that the rates of true positivesand of false positives at group level are lower. Bycontrast, a low individual probability of a subject suchas 10% can be achieved at a lower tumor marker level.Hence, the corresponding rates of true positives andof false positives are higher.For men, we observed a good individualperformance of mesothelin as well as calretinin. Acombination of both markers showed moderateimprovement of the performance. For screening, onewould want to achieve a high probability in order tokeep the FPR low. For example, at 80% probability forhttp://www.medsci.org

Int. J. Med. Sci. 2018, Vol. 15a diagnosis of MPM mesothelin and calretininshowed a low rate of 1% and 0% false-positivedecisions corresponding to 43% and 26% true-positivedecisions, respectively. The combination of bothmarkers resulted in 1% false-positive and 53%true-positive decisions. However, for diagnosticworkup a lower probability can be accepted. Using aprobability of 30%, mesothelin and calretinin showed11% and 18% false-positive decisions withcorresponding 79% and 81% true-positive decisions,respectively. The combination of both markersimproved the performance to an acceptable 11%false-positive decisions and a relatively high rate of84% true-positive decisions.MPM is a very rare disease, especially in womenwhere markers used to diagnose MPM may be alsoelevated in cases with cancer of the ovaries [33]. Ourprediction for female candidates with tissue sampleswaiting for histopathological diagnostic workupresulted in 3% false-positive decisions and 75%true-positive decisions for a probability of 30% formesothelin alone. A combination with calretinin didnot improve the performance of mesothelin.However, our study is of limited statistical power toestimate sound results for women, based on 12 casesonly. Another, more general, limitation is thecase-control design, where we cannot estimate theeffect of age and gender as risk factors of developingan MPM due to the matching procedure.Despite that a large number of candidatemarkers has been reported for MPM in the literature,mesothelin is still the best blood-based tumor markeravailable and the only one with FDA approval. It hasa good performance to detect MPM with AUCsranging between 0.72 and 0.93 [34]. Our results formesothelin are in line with similar findings that MPMpatients exhibit higher mesothelin concentrations inblood than healthy controls.Calretinin is one of the best immunohistochemical markers for MPM diagnosis [35]. This is thefirst study determining calretinin in plasma ations in cases and controls were similar tothose reported by Raiko et al. in samples from Franceand Germany and recently by Johnen et al. in samplesfrom Australia and Germany [5, 11].A few conditions may result in elevated markerconcentrations and can thus lead to false-positivedecisions. The association of pre-analytical variationswith mesothelin were analyzed in several studies[36-38]. Particularly renal failure was shown to have astrong effect on the mesothelin concentrations,resulting in false-positive tests [36]. Within theframework of the MoMar study, we analyzed suchpre-analytical variations in archived plasma samples889from a prospective cohort of elderly subjects withoutmalignant diseases [39]. Knowledge of possibleinfluencing factors could improve the performance ofthe markers.It is common opinion that a panel of tumormarkers may improve the performance of a singlemarker [31, 40, 41]. However, many markers maychange in parallel during carcinogenesis as could beshown in a statistical analysis of various markersapplied to lung tissue [42]. Depending on the chosenprobability, the inclusion of calretinin into theprediction model with mesothelin led to animprovement, albeit moderate, of the markerperformance. A major reason is the tight correlationbetween the plasma concentrations of the twoproteins in cases. However, more sophisticateddecision algorithms are needed to take advantage ofthe combination of both markers. A more promisingapproach to benefit from a panel may be to combinemarkers from different molecular levels [31, 41]. Weare currently investigating combinations of proteinswith epigenetic as well as RNA markers. Initial resultswith mesothelin and the microRNA miR-103-3p lookpromising [43].Many new MPM markers are emerging fromsmall cross-sectional comparisons of cases andcontrols. Such a design suffers from methodologicalshortcomings [44]. The most critical problem forapplication in symptom-free cohorts for the earlydetection of MPM is a potential overestimation of thesensitivity in cases with late-stage cancer. However,this is of minor concern in our current analysis. Wetook advantage of marker concentrations frompatients with symptoms to predict an MPM diagnosisin order to accelerate the histopathological workup,facing limited resources in Mexico. In the future,however, it would also be desirable to providemarkers to be used in surveillance programs forhigh-risk groups of asbestos-exposed workers toallow early detection of MPM. A timely treatment atearly stages of tumor development should improvethe prospects of survival. Before these markers can beused in symptom-free subjects, however, theirperformance to detect MPM earlier has to be validatedusing a prospective study design [41, 45]. BecauseMPM is a rare disease, a sufficiently large cohort willbe necessary [32, 46]. Our study was conducted withinthe framework of MoMar, which is an internationalinitiative to identify and validate markers for MPM.Within MoMar, an at-risk cohort of more than 2,700former asbestos-exposed workers will be available formarker validation. Mesothelin and calretinin appearto be promising candidates for such a validationstud

minutes in a laminar flow hood. Plasma and buffy coat were separated and stored at -80 C until analysis. Mesothelin and thrombomodulin ELISAs were performed at CINVESTAV in Mexico City, Mexico. For mesothelin determination, sandwich ELISAs were performed according to the manufacturer s instructions (DY3265, R&D Systems, Minneapolis, MN).