A gene-based risk score for lung cancer susceptibility in smokers and ex-smokers

Transcription

1 See Editorial, p Department of Mediine, Aukland Hospital, Aukland, New Zealand; 2 Department of Mediine, University of Otago, Christhurh, New Zealand; 3 Department of Mediine, Waikato Hospital, Hamilton, New Zealand; 4 Department of Onology, Aukland Hospital, Aukland, New Zealand Correspondene to: Dr R Young, Department of Mediine, Aukland Hospital, Private Bag 92019, Aukland, New Zealand; roberty@ adhb.govt.nz Reeived 23 November 2008 Aepted 2 May 2009 A gene-based risk sore for lung aner suseptibility in smokers and ex-smokers R P Young, 1 R J Hopkins, 1 B A Hay, 1 M J Epton, 2 G D Mills, 3 P N Blak, 1 H D Gardner, 1 R Sullivan, 4 G D Gamble 1 ABSTRACT Bakground: Epidemiologial and family studies suggest that lung aner results from the ombined effets of age, smoking and geneti fators. Chroni obstrutive pulmonary disease (COPD) is also an independent risk fator for lung aner and oexists in 40 60% of lung aner ases. Methods: In a two-stage ase ontrol assoiation study, geneti markers assoiated with either suseptibility or protetion against lung aner were identified. In a test ohort of 439 Cauasian smokers or ex-smokers, onsisting of healthy smokers and lung aner ases, 157 andidate single nuleotide polymorphisms (SNPs) were sreened. From this, 30 SNPs were identified, the genotypes (odominant or reessive model) of whih were assoiated with either the healthy smokers (protetive) or lung aner (suseptibility) phenotype. After genotyping of this 30-SNP panel in a seond validation ohort of 491 subjets and using the same protetive and suseptibility genotypes from our test ohort, a 20-SNP panel was seleted on the basis of independent univariate analyses. Results: Using multivariate logisti regression, inluding the 20 SNPs, it was also found that age, history of COPD, family history of lung aner and gender were signifiantly and independently assoiated with lung aner. Conlusions: When numeri sores were assigned to both the SNP and demographi data, and sequentially ombined by a simple algorithm in a risk model, the omposite sore was found to be linearly related to lung aner risk with a bimodal distribution. Geneti data may therefore be ombined with other risk variables from smokers or ex-smokers to identify individuals who are most suseptible to developing lung aner. Approximately 90% of people with lung aner have a smoking history, yet only 10 15% of hroni smokers get lung aner, suggesting that fators in addition to smoking exposure must be relevant. 1 Epidemiologial studies have identified age, smoking exposure, impaired lung funtion and family history as key risk fators for lung aner. 2 Geneti fators have been shown to play a modest role in determining suseptibility to lung aner, 3 4 most likely by onferring an inherent suseptibility (exaggerated or maladaptive response) to hroni inflammation from aero-pollutant exposure (most ommonly smoking). 5 6 Like many aners, this provides the initial stimulus to tissue remodelling (eg, small airway disease and/or emphysema), DNA damage and impaired ell yle ontrol. 6 7 Prospetive studies have shown that,20% of smokers develop hroni obstrutive pulmonary disease (COPD), defined by spirometry (fored Original artile expiratory volume in 1 s (FEV 1 )/fored vital apaity,70% and/or FEV 1,80% of predited), whereas the majority have preserved lung funtion at, or lose to, predited values. 8 Both prospetive and retrospetive studies show that spirometri evidene of COPD is found in 40 60% of smokers diagnosed with lung aner In ontrast with smokers who maintain normal lung funtion, those with COPD have a 2 6-fold greater risk of lung aner, independent of age and pak-years These studies suggest that smokers with COPD are at an inherently inreased risk of lung aner (suseptible phenotype), whereas most smokers with normal lung funtion (estimated to be 80%) are at least risk (resistant phenotype) Geneti predisposition to lung aner is likely to be both polygeni and heterogeneous, onferred by a variable ombination of relatively ommon polymorphisms with low penetrane and modest effet sizes. Moreover, it is likely that important smoking-gene interations underlie lung aner, 14 as seen in other smoking-related aners (eg, bladder and stomah). As geneti variants assoiated with both COPD and lung aner have been identified, and inlude the reently reported hromosome 15q25 gene lous, we suggest it is important to measure lung funtion in all partiipants of ase ontrol studies of lung aner. For both linial and biostatistial reasons, sreening the exposed ontrols will inrease the power of the study to identify relevant geneti variants ompared with studies in whih the ontrol group is unsreened. 17 It is well known that non-geneti risk fators suh as age, history of COPD and smoking history are very important and an be ombined to develop risk-based tools for determining lung aner suseptibility. Reently, genotype data from previously impliated prostate aner suseptibility single-nuleotide polymorphisms (SNPs) were ombined with family history to derive risk estimates for prostate aner. 20 The objetive of this study was to use a similar approah to analysing data from a ase ontrol study and show how geneti variants, previously showing small effets on lung aner risk, an be ombined in an algorithm with other known risk fators to derive a gene-based suseptibility sore for lung aner. METHODS Study population This study was a two-stage ase ontrol design onduted in three entres following the same reruitment protool. Only people of Cauasian Postgrad Med J 2009;85: doi: /pgmj

2 anestry were reruited (all four grandparents of Cauasian desent). Lung aner ases were identified through hospital linis between 2004 and 2007 using the following riteria:.40 years of age, minimum 15 pak-years of smoking, diagnosis onfirmed on histologial or ytologial grounds, and limited to the following four histologial subtypes: adenoarinoma, squamous ell aner, small ell aner and non-small ell aner (generally large ell or bronhoalveolar subtypes). The median time interval between diagnosis and reruitment was 3 months. Patients with lung aner underwent blood sampling for DNA extration, an investigator-administered questionnaire and spirometry with a portable spirometer (Easy-One; ndd Medizintehnik AG, Zurih, Switzerland) following Amerian Thorai Soiety (ATS) riteria. For patients with lung aner who had already undergone surgery, results of preoperative lung funtion tests performed by the hospital laboratory (using ATS riteria) were soured from the medial reords. Control subjets were reruited from the same ommunities as the ases using the following riteria: Cauasian anestry (as defined above), age years and past or urrent smoking history of a minimum of 15 pak-years. Controls were volunteers who met the above riteria and were identified through either a ommunity mail-out or while attending a ommunity-based lub for older people loated in the suburbs that were the referral base for the hospital linis from whih the lung aner ases were reruited. All smoking ontrols underwent blood sampling, spirometry and the same investigator-administered questionnaire given to the ases. Controls with spirometry onsistent with COPD were exluded (30% of those who volunteered). Informed written onsent was obtained from both lung aner patients and ommunityaquired healthy smoker ontrols. The study was approved by the loal ethis ommittee. The questionnaire (modified from the ATS respiratory questionnaire) inluded data on demographi variables suh as age, gender, medial history, family history of lung disease, ative and passive tobao exposure, and oupational aero-pollutant exposures. Seletion and genotyping of SNPs After extensive review of both the lung aner and COPD literature, polymorphisms with the following attributes were seleted for initial sreening in the test ohort: (a) SNPs in genes enoding proteins in pathways of ell-yle ontrol, oxidant response, apoptosis and airways inflammation; (b) SNPs known either to have funtional effets on in vitro assays or to be either non-synonymous or in regulatory regions. In a test ohort of 439 smokers (run 1 reruited during : 239 lung aner ases and 200 ontrol smokers), 157 andidate SNPs were sreened (available on request), and those where the differene in genotype frequenies exeeded a 20% magnitude differene and p value,0.20 were identified as part of our model-forming approah. 21 Where the all rate (perentage of samples for whih genotyping failed for tehnial reasons) fell below 95% for any ohort, the reading and/or genotyping of failed samples was repeated for that SNP; after retesting, SNPs with all rates,95% were not inluded in further analysis. SNPs were assigned as protetive when the homozygote and/or heterozygote genotype for either allele was found more often in ontrol smokers than lung aner ases (in a reessive or odominant model). SNPs were assigned as suseptible when the homozygote and/or heterozygote genotype was found signifiantly more often in lung aner ases than ontrol smokers. Genotyping Genomi DNA was extrated from whole blood samples using standard salt-based methods. Purified genomi DNA was aliquoted (10 ng/ml onentration) into 96-well plates and genotyped on a Sequenom system (Sequenom Autoflex Mass Spetrometer and Samsung 24 pin nanodispenser) by the Australian Genome Researh Faility ( using sequenes designed in-house (available on request) and reommended amplifiation and separation methods (iplex; www. sequenom.om). 16 Of the 157 andidate SNPs sreened in our disovery ohort, 30 met the above riteria. These SNPs were genotyped in a seond ohort of 491 smokers using idential reruitment methods (run 2 reruited during : 207 lung aner ases and 284 ontrol smokers). For all SNP assays, again a minimum all rate of 95% was required. This seond validation ohort of lung aner ases and ontrol smokers was idential with the first groups with respet to demographi fators and lung aner harateristis. On the basis of independent univariate analyses in run 1 and run 2 (onsisteny, diretion and signifiane of assoiation), a final panel of the 20 most disriminatory SNPs was seleted (12 suseptibility SNPs and eight protetive SNPs from the test panel of 30). Algorithm The assignment of a protetive or suseptible SNP genotype was made from the test ohort data (run 1) and was stritly applied to the data from run 2. On the basis of an algorithm derived from our work on the genetis of COPD (unpublished data), a soring system was applied to the genotypes for eah of the suseptibility and protetive SNPs. For eah subjet, a numerial value of 21 was assigned for eah of the protetive genotypes present among the protetive SNPs and +1 for eah of the suseptible genotypes present. Where an individual did not have either the protetive or suseptibility genotype for that SNP, the sore was 0 (ie, did not ontribute to the geneti sore). This approah is onsistent with a reently published study in prostate aner. 20 Weighting the presene of speifi suseptible or protetive genotypes aording to their individual odds ratios (ORs; from univariate regression) did not signifiantly improve the disriminatory performane of the raw SNP sore (unpublished data). Lung aner suseptibility sore The approah of deriving an overall suseptibility sore by ombining independent risk fators is omparable to existing risk-soring systems suh as the Prostate Caner Test, Framingham Sore for oronary artery disease risk and the Gail Sore for breast aner. By using multivariate logisti and stepwise regression analysis, the 20-SNP panel was examined in ombination with relevant non-geneti fators. This analysis of run 1 data identified age, family history of lung aner and previous diagnosis of COPD as signifiant ontributors to lung aner suseptibility. In addition, and onsistent with other ase ontrol studies, female gender in our study was also assoiated with a small inreased risk of lung aner (p,0.01). However, we did not inlude gender in the final risk model, as its importane in prospetive studies has been laking. 24 On the basis of a multivariate logisti regression analysis in run 1 (see results for ombined analysis below), a sore was assigned aording to age, history of COPD and family history. These variables have been identified in other risk assessment tools for lung aner suseptibility and improved 516 Postgrad Med J 2009;85: doi: /pgmj

3 the disriminatory power of the SNP sore data alone. As smoking exposure (pak-years) was a reruitment riterion for this study and omparable between ases and ontrols, it was not surprising to find that it made little ontribution to this soring system derived from our ohorts. The lung aner suseptibility sore was plotted with (a) the frequeny of lung aner and (b) the floating absolute risk (equivalent to OR) aross the ombined smoker/ex-smoker ohort. Statistial analysis Patient harateristis in the ases and ontrols were ompared by unpaired t tests for ontinuous variables and x 2 test for disrete variables. Genotype and allele frequenies were heked for eah SNP by Hardy Weinberg equilibrium (tests that genotype frequenies were as expeted from the allele frequenies). Population admixture was exluded by the population struture analysis on genotyping data from 40 unrelated SNPs. 27 Distortions in the genotype frequenies were identified between ases and ontrols using ontingeny tables. Genotype data (20-SNP panel) and the most relevant non-geneti variables were ombined in a stepwise fashion to assess their ombined effets on disriminating low and high risk (by OR and reeiver operating harateristi (ROC) urve) by sore quintile. The frequeny distribution of the optimised lung aner suseptibility sore was ompared aross the ases and ontrols. Its linial utility was assessed using ROC analysis, whih assesses how well the model predits risk aross the sore (ie, linial performane of the sore with respet to sensitivity, speifiity and false positive rate). To assess the stability of the optimised risk model, a sensitivity analysis was performed in whih age, gender and smoking dose were more stringently mathed between ases and ontrols. The effet on sensitivity and performane of the lung aner suseptibility sore to the addition of non-geneti variables was also assessed by omparing ORs and ROC analyses. RESULTS Demographi variables and genotyping Table 1 summarises the harateristis of the lung aner ases and healthy ontrol smokers. The 446 lung aner ases from run 1 (n = 239) and run 2 (n = 207) were omparable (with respet to demographi harateristis, histology and staging) and similar to a large published series. 28 Given the small differene in age, the 482 healthy ontrol smokers (200 in run 1, 282 in run 2) were omparably exposed with respet to smoking and other aero-pollutants. The lower frequeny of urrent smokers in the lung aner group probably reflets oexisting COPD (higher quit rates), and longer duration of smoking in lung aner ases reflets an older age. In a gene-bysmoking interation model suh as this, differenes in smoking exposure are more likely to obsure effets (bias to the null) than generate effets. Consistent with the findings of others, the lung aner ohort had higher rates of a family history of lung aner (19% vs 9%) and history of COPD (29% vs 5%). The latter (5%) probably reflets a linial diagnosis of COPD, based on symptoms but not spirometry, in smokers with asthma and/ or hroni bronhitis. As expeted, lung funtion was worse in the lung aner ohort than the healthy smoker ontrols. Testing lung funtion in the lung aner ases (performed within 3 months of diagnosis, in the absene of pleural effusions and before surgery) allows us to test for onfounding by COPD (see below). Table 1 smokers Variable Summary of harateristis for the lung aner and resistant Lung aner patients (n = 446) Original artile Healthy ontrol smokers (n = 484) p Value Charateristi Male 53% 60% Age (years) 69 (10) 60 (10),0.001 Height (m) 167 (0.08) 170 (0.09),0.001 Weight (kg) 69 (15) 79 (15),0.001 History of COPD 29% 5%,0.001 Smoking history Non-smokers (%) 7% 0 Current smoking (%) 35% 48%,0.001 Age started (years) 18 (4) 17 (3) 0.18 Years smoked 41 (12) 35 (11),0.001 Pak-years 41 (25) 40 (19) 0.28 Cigarettes/day 20 (10) 24 (11),0.001 History of other exposures In utero smoke exposure 18% 17% 0.45 Mother smoked in 37% 41% 0.03 hildhood Home ETS exposure as 79% 58%,0.001 adult Work ETS exposure 86% 63%,0.001 Work dust exposure 63% 47%,0.001 Work fume exposure 41% 38% 0.16 Asbestos exposure 23% 16% 0.02 Family history COPD 33% 28% 0.12 Lung aner 19% 9%,0.001 Lung funtion FEV 1 (litres) 1.86 (0.48) 2.86 (0.68),0.001 FEV 1 % predited 73% 99%,0.001 FEV 1 /FVC 64 (13) 78 (7),0.001 Spirometri COPD* 51% 0%,0.001 Values are frequeny (given as %) or mean (SD). *Aording to GOLD 2+ riteria (FEV 1 /FVC, 70% and FEV 1 % predited ( 80%). COPD, hroni obstrutive pulmonary disease; ETS, environmental tobao smoke; FEV 1, fored expiratory volume in 1 s; FVC, fored vital apaity. The observed genotypes for the 20 SNPs in this study were in Hardy Weinberg equilibrium (table 2), thereby exluding signifiant genotyping error. The genotype frequenies for the ontrols were omparable to those from the International Hapmap Projet ( The development of the lung aner suseptibility sore is desribed in the Methods setion, and a summary of the 20-SNP panel univariate analysis is presented in table 3. Although six of the top 20 SNPs do not reah traditional levels of signifiane, they have been inluded in the panel beause (a) in previous studies they have been shown to have funtional effets, (b) they have been assoiated with COPD and/or lung aner (see Disussion), () in ombination they make a ontribution to the performane of the suseptibility sore, and (d) their inlusion reognises the likely geneti heterogeneity that exists in lung aner ase ontrols studies. A SAS maro was used to estimate the false disovery rate (FDR) (Osborne JA, North Carolina State University; pdf) and produe a q statisti as the smallest p value that would be said to be statistially signifiant while preserving an overall 5% signifiane level. Postgrad Med J 2009;85: doi: /pgmj

4 Table 2 SNP No Expeted genotype frequenies and Hardy Weinberg equilibrium (HWE) Allele frequenies Allele frequenies (study total = 930) HWE observed Study ontrols (n = 484) Hapmap Cauasian* genotypes SNP name (rs) Major Minor (p Value) Major Minor Major Minor 1 A5 nachr (rs ) CYP 2E1 (rs ) Interleukin-18 (rs ) Interleukin-8 (rs 4073) Interleukin 1B (rs 16944) ITGA11 (rs ) NAT 2 (rs ) a1-antihymotrypsin (rs 4934) Cerberus 1 (rs ) DAT1 (rs ) TNFR1 (rs ) TLR9 (rs ) P73 (TP73) (rs ) SOD3 (rs ) ITGB3 (rs ) DRD2 (rs ) BCL2 (rs ) XPD (ERCC2) (rs 13181) REV1 (REV1L) (rs ) FasL (TNFSF6) (rs ) *Allele frequenies for Cauasians from SNP, single nuleotide polymorphism. Risk model development In a multivariate logisti regression analysis that inluded the SNPs (individually), age (.60 years), family history of lung aner (first-degree relative), gender and history of COPD, the OR for the suseptibility and protetive SNPs was and , respetively (the ombined SNP sore is independently related to lung aner, p,0.001). The OR for age.60 years, family history of lung aner and history of COPD was 3.5 (95% CI 2.5 to 4.9, p,0.001), 2.5 (95% CI 1.6 to 4.0, p,0.001) and 7.5 (95% CI 4.5 to 12.4, p,0.001), respetively (total area under the urve (AUC) = 0.80 where SNPs were inluded individually with adjustment for all variables). History of COPD in this model onfers a high risk, in part due to differenes in lung funtion derived from the study design. On the basis of these findings, and those from previously published studies, we derived an optimised sore by assigning sores to non-geneti variables as follows; +4 for those aged.60 years old, +3 for those with a family history of lung aner and +4 for those with a diagnosis of COPD (ie, age and diagnosis of COPD equally weighted). Lung aner sore = (number of suseptible genotypes) (number of protetive genotypes) + 3 (for positive family history for lung aner) + 4 (for past diagnosis of COPD) + 4 (for age.60 years old). Suh an approah is onsistent with existing risk sores and plaes the SNP data in an appropriate linial ontext Gender was not inluded in the finalised risk model for the reasons desribed above (and its inlusion did not alter the AUC). Model performane In the optimised model, the lung aner suseptibility sore was ompared with frequeny of lung aner, and a linear relationship was found aross the lung aner suseptibility sores (1 to 8+, with lung aner frequeny spanning 17 86% (fig 1). The magnitude of this effet was also sequentially examined using the floating absolute risk plotted on a log sale (equivalent to an OR), whih referenes the lowest frequeny group as OR = 1 (referent group) and ompares the lung aner sore with the referent group. The OR for SNPs alone (fig 2a), SNPs, family history and age (fig 2b) and SNPs, family history, age and COPD (fig 2) spanned from 1 to 10 (p,0.001), 1 to 19 (p,0.001) and 1 to 28 (p,0.001), respetively, aross the lung aner sores when subjets were grouped approximately as heptiles or quintiles. Subgrouping by age band or histology did not alter this linear relationship between sore and OR (data not shown).the lung aner suseptibility sore for lung aner ases and ontrols shows a bimodal distribution on frequeny distribution, indiating potential utility as sreening test. 29 Analysis of model sensitivity To orret for the small differenes in age, smoking status, COPD and gender mix between ases and ontrols, a subgroup (sensitivity) analysis was performed (a) limited to those.60 years of age (age weighting equally applied to all), (b) removing COPD from the model, and () where mean age, pak-years and gender were losely mathed between ases and ontrols (n = 450: 72 vs 69 years, 45 vs 43 pak-years and 70% vs 70% male respetively). A linear inrease in OR aross quintiles of the lung aner suseptibility sore (range 1 58, p,0.01) was still evident, with onfidene intervals onsistent (ie, overlapping) with those derived with the full dataset (fig 2). ROC analysis In a ROC analysis (n = 930) of the optimised model, we found that the AUC or statisti for run 1, run 2 and run 1+2 was 0.82, 0.75 and 0.79, respetively. The AUC in the total ohort for the 20-SNP panel, age, family history of lung aner and history of COPD on their own were 0.68, 0.70, 0.55 and 0.62, respetively. When just geneti fators are used in the risk model (SNPs + family history of lung aner), as seen in the Prostate Caner Study, 20 the ORs span 1 10 aross septiles and the AUC = 0.70 (with no ontribution from age and COPD). 518 Postgrad Med J 2009;85: doi: /pgmj

5 Table 3 Genotypes and results of univariate analysis SNP (rs No)* Genotype Lung aner On stepwise analysis, age and the SNP panel make the greatest ontribution to the AUC (SNPs = 0.68, age + SNPs = 0.76, age + SNPs + family history = 0.77), with history of COPD making a small additional ontribution (total AUC = 0.79) (fig 3). Using an FDR analysis, 12 SNPs were identified as being signifiantly assoiated with lung aner and, when ombined with age and family history, derived an AUC = When gender was inluded in the model, the AUC was not improved. Inlusion of COPD As smokers with normal lung funtion were seleted as ontrols (resistant or lowest risk phenotype), and history of COPD was inluded in the model, it is neessary to examine the effet of inluding COPD in the model. In the ROC analysis, history of COPD alone was a modest disriminator (AUC = 0.62) and added little to the other variables in the ombined model Smoking ontrols Call rate (%) Univariate OR (95% CI) p Value Phenotype a5-nachr (rs ) AA 68 (16%) 45 (9%) 1.8 (1.2 to 2.7) Suseptibility AG/GG 361 (84%) 426 (91%) CYP 2E1 (rs ) TT/TC 24 (6%) 14 (3%) (1.0 to 4.3) 0.03 Suseptibility CC 379 (94%) 463 (97%) Interleukin-18 (rs ) CC 237 (54%) 208 (45%) (1.1 to 1.9) Suseptibility CG/GG 201 (46%) 250 (55%) Interleukin-8 (rs 4073) TT 129 (31%) 109 (23%) (1.1 to 2.1) Suseptibility AT/AA 284 (69%) 367 (77%) Interleukin 1B (rs 16944) GG 215 (49%) 212 (44%) (0.9 to 1.6) 0.14 Suseptibility AA/AG 224 (51%) 269 (56%) ITGA11 (rs ) AA 14 (3%) 6 (1%) (0.9 to 7.6) 0.04 Suseptibility GA/GG 422 (97%) 470 (99%) N-Aetylysteine transferase 2 GG 239 (56%) 222 (47%) (1.1 to 1.9) Suseptibility (rs ) AA/AG 189 (44%) 253 (53%) a1-antihymotrypsin GG 123 (28%) 96 (20%) Suseptibility (rs 4934) AG/AA 312 (72%) 383 (80%) (1.2 to 2.2) Cerberus 1 (rs ) AA/AG 71 (16%) 59 (12%) (0.9 to 2.0) 0.10 Suseptibility GG 363 (84%) 413 (88%) DAT1 (rs ) GT/TT 64 (15%) 50 (10%) (1.0 to 2.3) 0.04 Suseptibility GG 367 (85%) 431 (90%) TNFR1 (TNFRSF1A) (rs AA 148 (36%) 142 (30%) (1.0 to 1.8) 0.05 Suseptibility ) AG/GG 258 (64%) 329 (70%) TLR9 (rs ) CC 12 (3%) 6 (1%) (0.8 to 6.6) 0.12 Suseptibility CT/TT 419 (97%) 455 (99%) P73 (TP73) (rs ) CC 219 (52%) 292 (62%) (0.49 to 0.85) Protetive TC/TT 206 (48%) 178 (38%) SOD3 (rs ) GG/GC 4 (1%) 15 (3%) (0.10 to 0.90) 0.02 Protetive CC 425 (99%) 451 (97%) ITGB3 (rs ) GG/GA 44 (10%) 77 (16%) (0.39 to 0.89) Protetive AA 391 (90%) 403 (84%) DRD2 (rs ) CDel/Del.Del 70 (16%) 107 (22%) (0.48 to 0.96) 0.02 Protetive CC 359 (84%) 372 (78%) BCL2 (rs ) AA 103 (24%) 145 (31%) (0.53 to 0.97) 0.03 Protetive AC/CC 328 (76%) 330 (69%) XPD (ERCC2) (rs 13181) GG 60 (14%) 81 (18%) (0.51 to 1.10) 0.11 Protetive GT/TT 376 (86%) 377 (82%) REV1 (REV1L) (rs ) CC 128 (29%) 163 (34%) (0.59 to 1.10) 0.10 Protetive TC/TT 310 (71%) 312 (66%) FasL (TNFSF6) (rs ) TT 53 (12%) 78 (16%) (0.49 to 1.10) 0.09 Protetive TC/CC 379 (88%) 403 (84%) *OMIM nomenlature. SNP, single nuleotide polymorphism. (inreased AUC from 0.77 to 0.79). When history of COPD was removed from the model, the ORs span 1 19 aross quintiles (p,0.001) and performane harateristis are minimally affeted (AUC = 0.77). The model was also tested in young smokers ((60 years old), in whom COPD prevalene was only 3% and the ORs spanned 1 16 (p,0.001). Most importantly, the model was assessed by omparing the smoking ontrols and lung aner ases subgrouped aording to lung funtion (fig 4a,b). This shows that (a) the distribution of the suseptibility sore was omparable between lung aner ases divided by those with high or low lung funtion (fig 4a) and (b) when people with COPD (based on spirometry) are removed from the analysis (leaving smoking ontrols ompared with lung aner ases with normal lung funtion), the bimodal distribution is not affeted (fig 4b). We onlude that the risk model is not signifiantly affeted after adjustment for differenes in COPD prevalene between ases and ontrols. Postgrad Med J 2009;85: doi: /pgmj

6 Figure 1 Frequeny of lung aner aording to the lung aner suseptibility (risk) sore modelled with single-nuleotide polymorphisms, age, family history and hroni obstrutive pulmonary disease. DISCUSSION This study has used a two-stage ase ontrol andidate gene approah and identified a panel of protetive and suseptibility SNPs that individually onfer only small effets (OR ranging from 0.3 to 2.6). This is very muh in keeping with the experiene from ase ontrol assoiation studies to date Consistent with existing risk models, relevant fators were ombined using an algorithm (in this study inluding SNP data) to derive a suseptibility sore on a simple linear sale. This study design, and the algorithmi approah that underlies our lung aner suseptibility sore, takes into aount important epidemiologial observations relevant to geneti predisposition to lung aner. Firstly, that, although smoking exposure is for the majority a prerequisite to developing lung aner, inreasing age, smoking dose and poor lung funtion have important independent effets on lung aner suseptibility. Seondly, the geneti fators underlying lung aner risk are likely to be both polygeni and heterogeneous, onferred by a variable ombination of geneti variants (ie, SNPs with low penetrane and small effet sizes). Thirdly, geneti fators may onfer either a protetive or suseptibility 16 phenotype to lung aner. Here we report a 20-SNP panel whih, ombined with family history, 20 define risk (OR) aross quintiles ranging from 1 to 10 with an AUC of A risk tool with greater linial utility an be derived by inluding age and presene of COPD to identify those at greatest suseptibility to lung aner (OR range 1 28 and AUC = 0.79). Several other important fators relevant to the geneti epidemiology of lung aner have been onsidered in the design of this study. We sought to minimise false-positive results in a number of ways. The most important of these was to internally validate our findings using a two-stage design with an initial test ohort (run 1) to identify SNPs of potential interest. We then tested only those SNPs in a seond ohort of ases and ontrols (run 2) using univariate and multivariate analysis to rank the SNPs under both onditions. Seondly, population stratifiation was exluded, and, thirdly, the presene of genotyping error was minimised through Hardy Weinberg equilibrium analysis (see Methods) and by the exlusion of SNPs with,95% all rate (fails on genotyping are invariably Figure 2 Odds ratio of lung aner aording to the lung aner suseptibility (risk) sore using (a) single-nuleotide polymorphisms (SNPs) only, (b) SNPs, family history (FHx) and age, and () SNPs, family history, age and history of hroni obstrutive pulmonary disease (COPD). genotype speifi, thus generating false-positive assoiations). With respet to important onfounding fators, our lung aner ases and healthy smoking ontrols were mathed for smoking exposure (pak-years). They were also similar with respet to gender and age mix. When the ombined ohort was subgrouped by age band, the lung aner suseptibility sore maintained its disriminating utility aross all groups. It was onluded that lung funtion was not onfounding the results of this study, as the distribution of the lung aner suseptibility sore aross the lung aner ases, subdivided by normal and low lung funtion, showed no signifiant differene. The same ould not be said of previously published ase ontrol studies in whih lung funtion was not measured. However, weaknesses in this study inlude the modest size of the ohorts, borderline signifiane of some SNPs in the absene 520 Postgrad Med J 2009;85: doi: /pgmj

7 Figure 3 Distribution of the lung aner suseptibility (risk) sore between ases and ontrols. (a) Single-nuleotide polymorphisms (SNPs) only, (b) SNPs, family history and age, and () SNPs, family history, age and history of hroni obstrutive pulmonary disease. Ca, aner. of orretion, ross-setional design, and reruitment limited to Cauasians. Moreover, it is aepted that, by seleting a ontrol population with normal lung funtion (but omparable exposure) and inluding COPD (history) in the sore, we will inrease the sore in those with lung aner ompared with ontrols (5% vs 29%). However, although this inreases the magnitude of the differene between ases and ontrols (refleted in the ORs), it ontributes little to the performane of the sore (adds 0.02 to the AUC; see the Results setion). Moreover, when subjets with COPD are exluded from the analysis (fig 4b, smoking ontrols versus lung aner ases with normal lung funtion), the disriminating utility of the sore is unaffeted. In addition, in the youngest age band (onfined to ases and ontrols (60 years old), the prevalene of COPD (history) was only 3% (little effet from COPD weighting), there was no age weighting, and the suseptibility sore was still a good disriminator (OR spans 1 16, p,0.001). We argue that sreening individuals with COPD (based on spirometry) out of the ontrols has the following advantages: (1) best reflets the majority of smokers with no COPD estimated at 80% 8 ; (2) best reflets the majority of smokers who will not develop lung aner (resistant phenotype) estimated to be 80 90% (thereby minimising the dilutional effets of inluding patients with COPD, ie, mislassifiation) 1 17 ; (3) best suited to identifying protetive SNPs by omparing exposed individuals at either end of the risk spetrum. That said, repliation Postgrad Med J 2009;85: doi: /pgmj

8 Figure 4 (a) Frequeny distribution of the lung aner sore among ontrols and lung aner ases divided aording to low and normal lung funtion. (b) Frequeny distribution of lung aner sore among ontrols and lung aner ases with normal lung funtion. Ca, aner; FEV 1, fored expiratory volume in 1s. using an unseleted ontrol group (in whih COPD prevalene would be 10% or more) might better reflet an unseleted at-risk population and, as expeted, redue differenes in the suseptibility sore between ases and ontrols (dilutional effet). Although population stratifiation was formally tested, and our population is onfined to Cauasians (where population admixture is less of an issue), it is possible that this remains a problem. A further limitation of the study is that, although the ases and ontrols were arguably representative, not all variables were preisely mathed (eg, age, gender and smoking patterns). We reanalysed our data in a losely mathed ohort (n = 450: 72 vs 69 years, 45 vs 43 pak-years and 70% vs 70% male for ases and ontrols, respetively) and found the performane of the suseptibility sore aross quintiles was unhanged (OR range 1 58, p,0.01). Further studies will need to be arried out to address these issues. It is likely that geneti suseptibility to lung aner results from a variable ombination of several geneti variants in genes enoding proteins involved in several pathways ativated by hroni smoke exposure and the inflammatory response that follows. A andidate gene (ie, hypothesis-driven) approah was used to identify potentially funtional SNPs assoiated with the development of lung aner. Although the SNPs identified in this study may only reflet linkage disequilibrium with funtional variants nearby, these SNPs are likely to have funtional effets and involvement diretly with suseptibility to lung aner. Two SNPs are from genes involved in the metabolism of smoking-derived arinogens (N-aetyltransferase 2 and ytohrome P450 2E1) and previously linked to smoking-related aners of the aerodigestive system. Five SNPs are from genes enoding inflammatory ytokines impliated in arinogenesis or lung matrix remodelling (COPD), the latter strongly impliated in lung aner development (interleukins 1, 8 and 18, tissue nerosis fator reeptor, Toll-like reeptor 9) Two SNPs are from genes that have been impliated in smoking addition and lung aner (dopamine D reeptor and dopamine transporter 1). Two SNPs are funtional and found in genes involved in the antioxidant response to aero-pollutants suh as smoking (a1-antihymotrypsin and extraellular superoxide dismutase) Both of these have been assoiated with COPD, and one is upregulated in lung aner. Six of the SNPs are found in genes involved in proesses suh as ell-yle ontrol, DNA repair and apoptosis, and assoiated with lung aner in previously published studies (xeroderma pigmentosum omplementary group D, p73, Bl-2, FasL, Cerb1 and REV1) Two of the SNPs are from genes enoding integrins also impliated in apoptosis, aner suseptibility and, for one, upregulation in lung aner ells One of the SNPs (a5 nachr) has reently been assoiated with both lung aner and COPD in genome-wide assoiation studies. This reeptor appears to de diretly related to niotine effets on airway inflammation. 63 As an be seen, the SNP panel (table 3) is made up of a variety of SNPs from genes impliated in metabolism of smoke-derived arinogens, oxidant response, ell-yle ontrol and inflammation. Twelve of these SNPs have been assoiated with lung aner in other ohorts. It is likely that other SNPs from as yet unidentified genes will be identified in the future. To assess further the utility of the lung aner suseptibility sore, a prospetive study is in progress. To date, the lung aner ases (n = 43) have the same mean and distribution as the lung aner ases reported in this study (unpublished data). Further ase ontrol and funtional studies will be needed to further explore the role of these SNPs in lung aner suseptibility. We propose that linial utility of genotype data requires that many SNPs are analysed and their effets ombined with other epidemiologial fators of relevane. 20 The algorithm approah used in this study is omparable to that reently published for prostate aner 20 and involves minimal assumptions (not hierarhial or path analysis based). The patient s sore an be ompared with the sores in smokers with least suseptibility to lung aner (lowest quintiles) in a simple linear fashion. Suh an approah is omparable to the risk tools developed by others and similar in approah to reently published studies on risk in diabetes, where SNP data were ombined with non-geneti risk variables to refine existing risk models The linial utility of the lung aner suseptibility sore was assessed by ROC analysis. This showed the statisti to be 0.79 and, at a ut-off of >3, an estimated sensitivity of 89% and orresponding speifiity of 45%. After FDR analysis, 12 signifiantly assoiated SNPs were inluded in the model, with little derease in the AUC (0.75 vs 0.77). These findings are omparable to the ROC performane of the Framingham Sore ( statisti = 0.74), 22 although other methods of assessing model performane have been advoated (eg, relassifiation table approah 66 ). The statisti for the 20-SNP panel on its own was 0.68 (and 0.70 when ombined with family history), indiating its utility in the urrent ohort. In ontrast with the models for diabetes and prostate aner, in our risk model for lung aner it has been possible to aount for the important environmental risk fator of smoking. There is evidene, although limited, that 522 Postgrad Med J 2009;85: doi: /pgmj

9 Main messages Lung aner results from the ombined effets of smoking and geneti suseptibility. Chroni obstrutive pulmonary disease is a ommon preexisting and independent risk fator for lung aner. Geneti suseptibility for lung aner inludes geneti variants (single nuleotide polymorphims (SNPs)) onferring redued risk ( protetive ) best identified using a healthy smoking ohort. Geneti suseptibility for lung aner results from the ombined effets of geneti variants (SNPs) onferring either suseptibility or protetive predisposition. Geneti and non-geneti variables an be ombined to give a global risk sore for suseptibility to lung aner. Current researh questions Can the lung aner suseptibility sore be validated in smokers and ex-smokers in prospetive studies and other populations? Will use of the lung aner suseptibility sore improve patient outomes to redue risk of lung aner and/or detet lung aner at a treatable stage? geneti testing may positively alter the behaviour of smokers in the ontext of smoking essation (inrease intent and possibly improve quit rate ) or by lowering smoking prevalene. 69 The lung aner suseptibility sore may also have utility in early diagnosis of lung aner where delays in diagnosis may affet survival. 70 Although further validation studies are required, this study suggests that geneti data may be ombined with other risk variables from smokers or ex-smokers to identify individuals most suseptible to developing lung aner. Aknowledgements: We gratefully aknowledge the partiipation of subjets in this study, in partiular the patients with lung aner. Funding: This study was in part funded by the Health Researh Counil of New Zealand (Grant ), the Aukland Medial Researh Foundation of New Zealand and the University of Aukland (Staff Researh Fund), New Zealand. Competing interests: This study was part funded by Synergenz BioSiene Ltd. RY is an advisor to this ompany. Provenane and peer review: Not ommissioned; externally peer reviewed. REFERENCES 1. Mattson ME, Pollak ES, Cullen JW. What are the odds that smoking will kill you? Am J Pub Health 1987;77: Alberg AJ, Samet JM. Epidemiology of lung aner. Chest 2003;123: Jonsson S, Thorsteinsdottir U, Gudbjartsson DF, et al. Familial risk of lung arinoma in the Ielandi population. JAMA 2004;292: Lihtenstein P, Holm NV, Verkasalo PK, et al. Environmental and heritable fators in the ausation of aner: analyses of ohorts of twins from Sweden, Denmark and Finland. N Eng J Med 2000;343: Young RP, Hopkins R, Eaton TE. Fored expiratory volume in one seond: not just a test of lung funtion but a biomarker of premature death from all auses. Eur Respir J 2007;30(4): Brody JS, Spira A. Chroni obstrutive pulmonary disease, inflammation, and lung aner. Pro Am Thora So 2006;3: Mannino DM, Watt G, Hole D, et al. The natural history of hroni obstrutive pulmonary disease. Eur Respir J 2006;27: Lokke A, Lang P, Sharling H, et al. Developing COPD: a 25 year follow up study of the general population. Thorax 2006;61: Mannino DM, Aguayo SM, Petty TL, et al. Low lung funtion and inident lung aner in the United States: data from the first NHANES follow-up. Arh Int Med 2003;163: Original artile 10. Young RP, Hopkins RJ, Christmas T, et al. COPD prevalene is inreased in lung aner independent of age, gender and smoking history. Eur Respir J 2009; 5 th Feb on-line. 11. Anthonisen NR. Prognosis in hroni obstrutive pulmonary disease: results from multienter linial trials. Am Rev Respir Dis 1989;140:S95 S Xu H, Spitz MR, Amos CI, et al. Complex segregation analysis reveals a multigene model for lung aner. Hum Genet 2005;116: Shields P, Harris C. Caner risk and low-penetrane suseptibility genes in geneenvironment interations. J Clin Onol 2000;18: Zhou W, Liu G, Park S, et al. Gene-smoking interation assoiations for the ERRC1 polymorphisms in the risk of lung aner. Caner Epidemiol Biomarkers Prev 2005;14: Shwartz AG, Prysak GM, Bok CH, et al. The moleular epidemiology of lung aner. Carinogenesis 2007;28: Young RP, Hopkins RJ, Hay BA, et al. Lung aner gene assoiated with COPD: triple whammy or onfounding effet? Eur Respir J 2008;32: Moskvina V, Holmans P, Shmidt KM, et al. Design of ase-ontrols studies with unsreened ontrols. Ann Hum Genet 2005;69: Cassidy A, Duffy SW, Myles JP, et al. Lung aner risk predition: a tool for early detetion. Int J Caner 2006;120: Spitz MR, Hong WK, Amos CI, et al. A risk model for predition of lung aner. J Natl Caner Inst 2007;99(9): Zheng SL, Sun J, Wiklund F, et al. Cumulative assoiation of five geneti variants with prostate aner. NEJM 2008;358: Lee K, Koval JJ. Determinations of the best signifiane level in forward stepwise regression. Commun Stats 1997;26: Grundy SM, Balady GJ, Criqui MH, et al. Primary prevention of oronary heart disease: guidane from Framingham. Cirulation 1998;97: Gail MH, Brinton LA, Byar DP, et al. Projeting individual probabilities of developing breast aner for white females who are being examined annually. J Natl Caner Inst 1989;81: Neugut AI, Jaobson JS. Women and lung aner: Gender equality at a rossroad? JAMA 2006;296: Easton DF, Peto J, Babiker AG, et al. Floating absolute risk: an alternative to relative risk in survival and ase-ontrol analysis avoiding an arbitrary referene group. Statistis in Mediine 1991;10: Plummer M. Improved estimates of floating absolute risk. Statistis in Mediine 2004;23: Prithard JK, Stephens M, Donnelly P. Inferene of population struture from multilous genotype data. Genetis 2000;155: Yang P, Allen MS, Aubry MC, et al. Clinial features of 5,628 primary lung aner patients; experiene at Mayo lini from 1997 to Chest 2005;128: Wald NJ, Hakshaw AK, Frost CD. When an a risk fator be used as a worthwhile sreening test? Brit Med J 1999;319: Young RP, Hopkins RJ, Blak PN, et al. Funtional variants of anti-oxidant genes in smokers with COPD and in those with normal lung funtion. Thorax 2006;30: Juul K, Tybjærg-Hansen A, Marklund S, et al. Genetially inreased antioxidative protetion and dereased COPD. Am J Respir Crit Care Med 2006;173: WikmanH, Thiel S, Jäger B, et al. Relevane of N-aetyltransferase 1 and 2 (NAT1,NAT2) geneti polymorphisms in non-small ell lung aner suseptibility. Pharmaogenetis 2001;11: Wu X, Amos CI, Kemp BL, et al. Cytohrome P450 2E1 Dra I polymorphisms in lung aner in minority populations. Caner Epidemiol Biomarkers Prev 1998;7: Engels EA, Wu X, Gu J, et al. Systemati evaluation of geneti variants in the inflammation pathway and risk of lung aner. Caner Res 2007;67: Zienolddiny S, Ryberg D, Maggini V, et al. Polymorphisms of the interleukin-1 B gene are assoiated with inreased risk of non-small ell lung aner. Int J Caner 2004;109: Campa D, Zienolddiny S, Maggini V, et al. Assoiation of a ommon polymorphism in the ylooxygenase 2 gene with risk of non-small ell lung aner. Carinogenesis 2004;25: Hoshino T, Kato S, Oka N, et al. Pulmonary inflammation and emphysema: role of the ytokines IL-18 and IL-13. Am J Respir Crit Care Med 2007;176: Hodge SJ, Hodge GL, Reynolds PN, et al. Inreased prodution of TGFB and apoptosis of t lymphoytes isolated from peripheral blood in COPD. Am J Physiol Lung Cell Mol Physiol 2003;285:L492 L Bhavsar TM, Cerreta JM, Cantor JO, et al. Short term igarette smoke exposure predisposes the lung to seondary injury. Lung 2007;185: Droemann D, Albreht D, Gerdes J, et al. Human lung aner sells express funtionally ative Toll-like Reeptor 9. Respiratory Researh 2005;6: Noakes PS, Hale J, Thomas R, et al. Maternal smoking is assoiated with impaired neonatal toll-like reeptor mediated immune responses. Eur Respir J 2006;28: Lazarus R, Klimeki WT, Raby BA, et al. Single nuleotide polymorphisms in the tolllike reeptor 9 gene: frequenies, pairwise linkage disequilibrium and haplotypes in three US ethni groups and exploratory ase-ontrol disease assoiation studies. Genomis 2003;81: Campa D, Zienolddiny S, Lind H, et al. Polymorphisms of dopamine reeptor/transporter genes and risk of non-small ell lung aner. Lung Caner 2006;56(1): Wu X, Hudmon KS, Detry MA, et al. D2 dopamine reeptor gene polymorphisms among Afrian-Amerians and Mexian-Amerians: a lung aner ase-ontrol study. Caner Epidemiol Bio Preven.t 2000;9: Postgrad Med J 2009;85: doi: /pgmj

10 45. Ishii T, Matsuse T, Teramoto S, et al. Assoiation between alpha1-antihymotrypsin polymorphisms and suseptibility to hroni obstrutive pulmonary disease. Eur J Clin Invest 2000;30: Zelvyte I, Wallmark A, Piitulainen E, et al. Inreased plasma levels of serine proteinase inhibitors in lung aner patients. Antianer Researh 2004;24: Hu Z, Wei Q, Wang X, et al. DNA repair gene XPD polymorphism and lung aner risk: a meta-analysis. Lung Caner 2004;46: Yin JY, Vogel U, Ma Y, et al. A haplotype enompassing the variant allele of DNA repair gene polymorphism ERCC2/XPD Lys 751Gln but not the variant allele of Asp312Asn is assoiated with risk of lung aner in a northeastern Chinese population. Can Genet Cytogen 2007;175: Spitz MR, Wu X, Wang Y, et al. Modulation of nuleotide Exision repair apaity by XPD polymorphisms in lung aner patients. Caner Res 2001;61: Goujun L, Wang LE, Chamberlain RM, et al. p73 G4C14-to-A4T14 polymorphism and risk of lung aner. Can Res 2004;64: Sata M, Takabatake N, Inoue S, et al. Introni single nuleotide polymorphisms in Bl-2 are assoiated with hroni obstrutive pulmonary disease severity. Respirology 2007;12: Jain M, Kumar S, Lal P, et al. Role of Bl-2 (ala43thr), CCND1 (G870A) and FAS (A- 670G) polymorphisms in modulating the risk of developing esophageal aner. Caner Dete Prev 2007;31(3): Zhang X, Miao X, Sun T, et al. Funtional polymorphisms in ell death pathway genes FAS and FASL ontribute to risk of lung aner. J Med Genet 2005;42: Sakiyama T, Kohno T, Mimaki S, et al. Assoiation of amino aid substitution polymorphisms in DNA repair genes TP53, POLI, REV1, and LIG4 with lung aner. Int J Caner 2005;114: Rudd MF, Webb EL, Matakidou A, et al. Variants in the GH-IGF axis onfer suseptibility to the lung aner. Genome Res 2007; August Jakubowska A, Gronwald J, Menkiszak J, et al. Integrin beta3 Leu 33Pro polymorphism inreases BRCA1-assoiated ovarian aner risk. J Med Genet 2007;44: Zhu CQ, Popova SN, Brown ER, et al. Integrin alpha11 regulates IGF2 expression in fibroblasts to enhane tumorigeniity of human non-small-ell lung aner ells. Pro Natl Aad Si 2007;104: Save your favourite artiles and useful searhes 58. Chong IW, Chang MY, Chang HC, et al. Great potential of a panel of multiple hmth1, SPD, ITGA11, and COL11A1 markers for the diagnosis of patients with non-small ell lung aner. Onol Rep 2006;16: Thorgeirsson TE, Geller F, Sulem P, et al. A variant assoiated with niotine dependene, lung aner and peripheral arterial disease. Nature 2008;452: Hung RJ, mkay JD, Gaborieau V, et al. A suseptibility lous for lung aner maps to niotini aetylholine reeptor subunit genes on 15q25. Nature 2008;452: Amos CI, Wu X, Broderik P, et al. Genome-wide assoiation san of tag SNPs identifies a suseptibility lous for lung aner at 15q25.1. Nat Genetis 2008; April 2 nd, Pillai SG, Shianna K, Ge D, et al. Genome-wide assoiation study of hroni obstrutive pulmonary disease (COPD) in a ase ontrol population from Norway [abstrat]. Amerian Thorai Soiety International Conferene: May 2008: Toronto pg A Gwilt CR, Donnelly LE, Rogers DF. The non-neuronal holinergi system in the airways: an unappreiated regulatory role in pulmonary inflammation? Pharmaol Therapeut 2007;115: Lyssenko V, Jonsson A, Almgren P, et al. Clinial risk fators, DNA variants, and the development of type 2 diabetes. NEJM 2008;359: Meigs JB, Shrader P, Sullivan LM, et al. Genotype sore in addition to ommon risk fators for predition of type 2 diabetes. NEJM 2008;359: Cook NR. Use and misuse of the reeiver operator harateristi urve in risk predition. Cirulation 2007;115: MBride CM, Bepler G, Lipkus IM, et al. Inorporating geneti suseptibility feedbak into a smoking essation program for Afrian-Amerian smokers with low inome. Caner Epidemiol Biomarkers Prev 2002;11: Lerman C, Gold K, Audrain J, et al. Inorporating biomarkers of exposure and geneti suseptibility into smoking essation treatment: effets on smoking-related ognitions, emotions and behavior hange. Health Psyhology 1997;16: Strange C, Dikson R, Carter C, et al. Geneti testing for alpha1-antitrypsin defiieny. Genet Med 2004;6: Hamilton W, Peters TJ, Round A, et al. What are the linial features of lung aner before the diagnosis is made? A population based ase-ontrol study. Thorax 2005;60: Use the My folders feature to save and organise artiles you want to return to quikly saving spae on your hard drive. You an also save searhes, whih will save you time. You will only need to register one for this servie, whih an be used for this journal or all BMJ Journals, inluding the BMJ. 524 Postgrad Med J 2009;85: doi: /pgmj