Night shift work, chronotype and prostate cancer risk in the MCC-Spain case-control study

Night shift work, chronotype and prostate cancer risk in theMCC-Spain case-control study

Kyriaki Papantoniou1,2,3,4, Gemma Casta~no-Vinyals1,2,3,4, Ana Espinosa1,2,3,4, Nuria Aragon�es4,5,6,

Beatriz P�erez-G�omez4,5,6, Javier Burgos4,7,8,9, In�es G�omez-Acebo4,10,11, Javier Llorca4,10,11, Rosana Peir�o4,12,

Jose Juan Jimenez-Mole�on4,13, Francisco Arredondo4,14,15, Adonina Tard�on4,16, Marina Pollan4,5,6

and Manolis Kogevinas1,2,3,4,17

1 Centre for Research in Environmental Epidemiology (CREAL), Barcelona, Spain2 Bioanalysis Research Group, IMIM (Hospital Del Mar Medical Research Institute), Barcelona, Spain3 Universitat Pompeu Fabra (UPF), Barcelona, Spain4 CIBER Epidemiolog�ıa Y Salud P�ublica (CIBERESP), Madrid, Spain5 Environmental and Cancer Epidemiology Area, National Center of Epidemiology, Carlos III Health Institute, Madrid, Spain6 Cancer Epidemiology Research Group, Oncology and Hematology Area, IIS Puerta De Hierro, Madrid, Spain7 Servicio De Urolog�ıa, Hospital Ram�on Y Cajal, Madrid, Spain8 Instituto Ram�on Y Cajal De Investigaci�on Sanitaria (IRYCIS), Madrid, Spain9 Universidad De Alcal�a De Henares, Madrid, Spain10 University of Cantabria, Santander, Spain11 IDIVAL, Santander, Spain12 Fundaci�on Para El Fomento De La Investigaci�on Sanitaria Y Biom�edica De La Comunidad Valenciana (FISABIO), Valencia, Spain13 Department of Preventive Medicine and Public Health, Instituto De Investigaci�on Biosanitaria Ibs.GRANADA, Hospitales Universitarios De Granada/Univer-

sidad De Granada, Granada, Spain14 Hospital Infanta Elena, Huelva, Spain15 Centro De Investigaci�on En Salud Y Medio Ambiente (CYSMA), Universidad De Huelva, Spain16 IUOPA, Universidad De Oviedo, Asturias, Spain17 National School of Public Health, Athens, Greece

Night shift work has been classified as a probable human carcinogen based on experimental studies and limited human evi-

dence on breast cancer. Evidence on other common cancers, such as prostate cancer, is scarce. Chronotype is an individual

characteristic that may relate to night work adaptation. We evaluated night shift work with relation to prostate cancer, taking

into account chronotype and disease severity in a population based case-control study in Spain. We included 1,095 prostate

cancer cases and 1,388 randomly selected population controls. We collected detailed information on shift schedules (perma-

nent vs. rotating, time schedules, duration, frequency), using lifetime occupational history. Sociodemographic and lifestyle fac-

tors were assessed by face-to-face interviews and chronotype through a validated questionnaire. We used unconditional

logistic regression analysis adjusting for potential confounders. Subjects who had worked at least for one year in night shift

work had a slightly higher prostate cancer risk [Odds Ratio (OR) 1.14; 95%CI 0.94, 1.37] compared with never night workers;

this risk increased with longer duration of exposure (�28 years: OR 1.37; 95%CI 1.05, 1.81; p-trend 5 0.047). Risks were

more pronounced for high risk tumors [D’Amico classification, Relative Risk Ratio (RRR) 1.40; 95%CI 1.05, 1.86], particularly

among subjects with longer duration of exposure (�28 years: RRR 1.63; 95%CI 1.08, 2.45; p-trend 5 0.027). Overall risk was

higher among subjects with an evening chronotype, but also increased in morning chronotypes after long-term night work. In

this large population based study, we found an association between night shift work and prostate cancer particularly for

tumors with worse prognosis.

Key words: night shift work, chronotype, prostate cancer, prognosis, survival

Additional Supporting Information may be found in the online version of this article.

Grant sponsor: “Accion Transversal del Cancer,” Spanish Ministry Council; Grant sponsor: Instituto de Salud Carlos III-FEDER; Grant

numbers: PI08/1770, PS09/00773, PS09/01286, PS09/01903, PS09/02078, PS09/01662, AP_061/10, PI11/01889 and PI12/01270; Grant

sponsor: predoctoral grant PFIS; Grant number: FI09/00385

DOI: 10.1002/ijc.29400

History: Received 8 Aug 2014; Accepted 2 Dec 2014; Online 20 Dec 2014

Correspondence to: Manolis Kogevinas, Centre for Research in Environmental Epidemiology (CREAL), Barcelona, Doctor Aiguader, 88

08003 Barcelona, Spain, E-mail: [email protected]

Epidemiology

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

International Journal of Cancer

IJC

Prostate cancer is the most common cancer in males with anincreasing incidence worldwide.1,2 The etiology of this tumorremains largely unknown and the only well-known risk fac-tors include age, race and family history of prostate cancer.3

Hereditary factors are important and numerous genes associ-ated with prostate cancer have been identified.4 Physical inac-tivity, smoking, sleep deprivation, vitamin D deficiency dueto low sun exposure and diet related factors, such as highmeat and alcohol and low vegetable consumption are possiblerisk factors for prostate cancer, for which evidence is stillinconclusive.3 Androgens also play a key role in the prostatecell’s growth, function and proliferation. Lifetime exposure toincreased androgen levels may increase risk for prostate can-cer and antiandrogenic therapies are commonly used againstprostate cancer.5

A potential effect of circadian disruption on prostate can-cer incidence has been recently suggested (evidence reviewedin Sigurdardottir et al. 2012).6 Night shift work, the maincause of circadian disruption, is one of the most widespreadoccupational exposures in the industrialized part of theworld, with about 15–20% of total active population workingpartly or entirely during the night. Night shift work has beenclassified by the International Agency for Research on Cancer(IARC) as a probable carcinogen (group 2A) for humans.7

Most epidemiological studies have focused on female breastcancer8 and evidence on other tumor sites is scarce. Anincreased prostate cancer risk has been observed among nightworkers in a few studies, however some reports were basedon small number of cases or limited information on shiftwork.9–12 A consensus report stressed the need for a morerefined exposure assessment to capture aspects of shift workschedules (permanent vs. rotating shifts, number of nightshifts, years of night work) that might be more relevant forcancer risk and preventive actions.13

Night shift work and subsequent exposure to light at nightdisrupts human biologic rhythms resulting in circadian dis-ruption, melatonin suppression, sleep deprivation and circa-dian gene deregulation, possibly involved in prostate cancerrelated pathways.14,15 A small number of circadian genes areresponsible for maintaining the 24-hr circadian rhythms butalso control about 10% of the whole genome.16 Chronic dis-ruption of the circadian genes may lead to cell cycle deregu-lation and could increase DNA damage replication errorsand resulting mutations, thus lead to tumor initiation.17 Fur-thermore some experimental evidence suggests that circadiandisruption may as well accelerate tumor development and

growth.18,19 A recent study showed an association betweenshift work and elevated prostate-specific antigen (PSA) lev-els,20 a marker of higher prostate cancer risk, but also worsedisease prognosis and increased mortality.21 Therefore it ishypothesized that shift work that involves circadian disrup-tion may be associated not only with increased cancer riskbut also with decreased survival.22

Chronotype is an individual characteristic that describesthe circadian phase and correlates with diurnal preference,the individual preference for morning or evening activity.23,24

It has been suggested that chronotype may modify the riskfor cancer since it has been associated to the capacity ofnight workers to adapt to nonday work schedules.25 A fewrecent studies have evaluated chronotype in conjunction withnight shift work and breast cancer risk,26–28 and no previousstudy has evaluated this hypothesis for prostate cancer.

We examined the association between night shift workand prostate cancer among males enrolled in the MCC-Spainstudy, a population based case-control study. We assessedcircadian disruption in shift workers using detailed exposuremetrics taking into account individual chronotype and sever-ity of prostate cancer.

MethodsStudy population

The MCC-Spain study is a population based multicase-control study on frequent tumors in Spain that includes 23hospitals in 12 regions and assesses 5 types of cancer (breast,colorectal, prostate, stomach and chronic lymphocytic leuke-mia) using the same series of population controls for allcases. The main aim of this study is to investigate environ-mental and genetic factors related to the above cancer typesin Spain (www.mccspain.org). The MCC-Spain study beganin the year 2008 and the recruitment of incident cancer casesand population controls took place until the end of 2013.

Prostate cancer cases, aged 27–85, were recruited in 11hospitals in 7 Spanish regions (Barcelona, Madrid, Cantabria,Valencia, Granada, Huelva and Asturias). All cases had anew histologically confirmed diagnosis of prostate cancerfrom September 2008 through December 2012 and lived inthe catchment area of each hospital for at least 6 monthsprior to diagnosis. Control subjects were men free of prostatecancer history, living in the same catchment area as cases.Controls were selected randomly from the rosters of GeneralPractitioners at the primary health centers (PHC) involved inthe study and were frequency matched to cases by age in 5-

What’s new?

Up to 20% of workers do night-shift work, which may increase the risk of some cancers. In this study, the authors found that

long-term night-shift work was associated with an increased risk of prostate cancer and decreased survival. Overall risk was

higher among workers with an evening chronotype (i.e., a preference for working in the evening vs. in the morning), but risk

also increased for morning chronotypes if the duration of night-shift work increased. These results may improve our under-

standing of prostate cancer etiology and potential prevention strategies.

Epidemiology

2 MCC-Spain case-control study

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

year age groups and study area. They were contacted onbehalf of their doctor and invited to participate to the study.Subjects incapable to participate in the interview due to com-munication difficulties or excess impairment of physical abil-ity were excluded. In total, 1,115 cases and 1,562 malecontrols recruited were eligible for participation and com-pleted the interview. In this analysis, we included 1,095 pros-tate cancer cases and 1,388 population controls, with shiftwork information. Response rates varied between centers andwere on average 74% among cases (range 47–94%) and 54%(range 30–94%) among controls with valid telephone num-bers in the PHC rosters.

Data collection

Information was obtained through face-to-face interviewsperformed by trained personnel. Lifetime occupational his-tory was assessed for all jobs held for more than a year.Detailed questions were used to ascertain information onshift work for each job, including shift work type (permanentvs. rotating), beginning and ending year, time schedules,hours worked per day, job title and worke�rs activity. Infor-mation was also collected on other potential risk factors suchas age, educational level, family socioeconomic level, race,body mass index (BMI), family history of prostate cancer,smoking status and leisure time physical activity. Food con-sumption was reported for all cases and controls through aself-administered diet questionnaire, completed by 85% ofparticipants. Participants were contacted again and a phoneinterview was performed to collect missing information onjobs with rotating night shift work (exact time schedules foreach shift, nights performed per month) and assess individualchronotype with the use of the Munich Chronotype Ques-tionnaire (MCTQ).29 In total, 2,104 subjects (1,208 controlsand 896 cases) agreed to participate in this follow-up andcompleted the chronotype questionnaire.

Clinical information was collected for most cases(n5 1,068) from medical records including anatomopathologi-cal and clinical stage, PSA levels and Gleason score. We usedthe D’Amico classification of prostate cancer that classifiespatients in groups of low (cT1-cT2a, Gleason score< 7 andPSA<= 10 ng/ml), medium (cT2b, Gleason score5 7 orPSA> 10 and PSA<= 20 ng/ml) and high (cT2c or PSA> 20ng/ml or Gleason> 7) risk. We also used Gleason score atdiagnosis alone (Gleason score< 7, Gleason score5 7 andGleason score> 7) to classify prostate cancer patients.

The MCC-Spain Study followed the national and internationaldirectives on ethics and data protection [declaration of Helsinkiand Spanish law on confidentiality of data (Ley Organica 15/1999de 13 Diciembre de Proteccion de Datos de car�acter personal-LOPD)]. All subjects that agreed to participate and fulfilled the eli-gibility criteria signed an informed consent form.

Night shift work assessment

Night shift work was defined as a working schedule thatinvolved working partly or entirely between 00:00 and 06:00

hr, at least three times per month. This definition includedlate evening shifts, overnight shifts and early morning shifts.Information on the type of night shift work was collectedand permanent vs. rotating schedules were compared. Therotating shift without nights was a shift schedule thatincluded rotation between the morning and afternoon shifts.The reference group consisted of men who were everemployed but had never performed night shift work for morethan a year. Rotating shift workers with no night shifts (55controls and 35 cases) were included in the reference group.Cumulative duration of night shift work was calculated as thetotal number of years worked at night [

Pjobs(years night

work/job)] and cumulative frequency was the total numberof nights worked [

Pjobs(night shifts/month/job * months

worked/job)] over working life. Cumulative duration and fre-quency of night shift work was assessed using tertiles ofexposure among controls that had ever performed night shiftwork and the same cut-offs were applied in the analysis ofpermanent and rotating night shift work.

Chronotype and other risk factors

Chronotype (MSFcorr) was estimated as the mid-sleep timeon free days [MSF5 (sleep onset on free day1 sleep durationon free day)/2)], corrected for oversleep on free days com-pared to working days [MSFcorr5MSF – (sleep duration onfree day-sleep duration on a working day)/2]. For example, ifan individual slept at 00:00 hr and woke up at 08:00 hr bothon free and working days, his chronotype (MSFcorr) scorewould be at 04:00 hr [MSFcorr 5MSF5 00:001 (8/2hrs)5 00:00 14 hrs5 04:00 hr]. We assessed chronotypeusing the continuous MSF score expressed in local time, butalso categorical variables with seven, five and three categories(morning type: MSF< 04:00 hr, neither type: MSF5 04:01–05:00 hr, evening type: MSF> 05:00 hr). Actual sleep dura-tion was assessed and the answers included only integers.Sleep problems (waking up during the night, problems fallingasleep, use of sleep medication) were assessed over lifetimefor problems that persisted for at least 1 year. Educationalstatus, a proxy for socioeconomic status, reflects the highestgrade completed. Daily consumption of meat, vegetable, fruitand total energy intake were calculated in g/day based on thedifferent food items and using food composition tables. Alco-hol intake (g ethanol/day) was calculated for both presentand past (at 30–40 years of age) consumption. Leisure timephysical activity information (type of activity, frequency: daysper week and hours per day, duration: age beginning and ageending) was available for all activities held over lifetime.Total and average annual estimations (METS-hr) were per-formed over lifetime, adulthood and the last 10 years of life.The annual mean of physical activity (METS-hr/week) overthe last 10 years was finally used in the models categorizingsubjects as inactive: 0 METS-hr/week, slightly active: from0.0001 to 8 METS-hr/week, moderately active: 8.0001 to 16METS-hr/week and very active: greater than 16 METS-hr/week. Family history of prostate cancer was available for all

Epidemiology

Papantoniou et al. 3

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

Table 1. Potential prostate cancer risk factors among participants in the MCC-Spain study (numbers may differ due to missing values; SD:standard deviation)

Controls (N 5 1388) Cases (N 5 1095) p values

Factor N (%) N (%)

Age; mean (SD) 66 (0.2) 66 (0.2) 0.492

Participating centers

Barcelona 478 (34.4) 396 (36.2) <0.0001

Madrid 333 (24.1) 314 (28.7)

Cantabria 188 (13.5) 175 (16.0)

Valencia 79 (5.7) 87 (7.9)

Granada 116 (8.4) 63 (5.8)

Huelva 93 (6.7) 44 (4.0)

Asturias 101 (7.3) 16 (1.5)

Family history of prostate cancer

No 1291 (93.4) 862 (79.1) <0.0001

Yes 91 (6.6) 228 (20.9)

Education

Less than primary 243 (17.5) 257 (23.5) <0.0001

Primary 451 (32.5) 428 (39.1)

High school 398 (28.7) 240 (21.9)

University 296 (21.3) 170 (15.5)

BMI (kg/cm2)

<22.5 93 (6.7) 70 (6.4) 0.996

22.5–25 265 (19.1) 214 (19.5)

25–30 701 (50.5) 557 (50.9)

�30 329 (23.7) 254 (23.2)

Smoking status

Never 388 (27.8) 326 (29.8) 0.401

Ex-smoker 719 (51.8) 568 (51.9)

Smoker 281 (20.2) 201 (18.4)

Past sun exposure

Only weekends/vacations 90 (6.5) 68 (6.2)

<1 hr/day 309 (24.0) 190 (17.4)

1–1.9 hr/day 336 (24.3) 306 (28.1) 0.003

2–3.9 hr/day 316 (22.9) 222 (20.4)

>4 hr/day 331 (24.0) 305 (28.0)

Sleep duration (hr/day)

�6 454 (33.2) 359 (33.2) 0.578

7–8 776 (56.7) 602 (55.6)

�9 139 (10.2) 122 (11.3)

Sleep problems

Never 1003 (72.6) 820 (75.2) 0.076

Ever 378 (27.4) 271 (25.0)

Chronotype

Morning type 541 (49.2) 452 (51.1) 0.412

Neither type 413 (37.6) 307 (34.7)

Evening type 145 (13.2) 125 (14.1)

Epidemiology

4 MCC-Spain case-control study

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

family members. The variable used in the model includedany family member with prostate cancer history (there wasonly a small number of non-first degree relatives). Informa-tion on habits of sun exposure was collected for the previoussummer and a summer 10 years ago (question: For how longdid you use to stay in the sun taking into account your workand everyday activities? answer: rarely <1 hr/day, a fewtimes: 1–1.9 hr/day, regularly: 2–3.9 hr/day, very frequently�4 hr/day, with frequency in weekends and vacations).

Statistical analysis

We compared the distribution of well-known or potentialrisk factors for prostate cancer between cases and controls,using the independent t test and the Pearson’s v2 test. Wealso evaluated the differences in sociodemographic and life-style characteristics between night and day workers. Weinvestigated the association of shift work and prostate cancerusing unconditional logistic regression analysis. We calculatedadjusted Odds Ratios (OR) with 95% confidence intervals(CI) for having ever done night shift work, using differentmetrics including types of night work and lifetime cumulativeexposure (duration and frequency). We assessed the directassociation of chronotype and prostate cancer risk and eval-uated chronotype as an effect modifier for the night shiftwork–prostate cancer association. We tested possible interac-tions, using the Wald test, between night shift work andchronotype, obesity and age at diagnosis (<60 years, >60years). We also analyzed night shift work by different clinicalstages of prostate cancer (D’Amico classification) and byincreasing Gleason score using polytomous logistic regressionmodels. Relative Risk Ratios (RRR) were estimated for eachclinical stage of cancer compared to controls. (e.g., outcomevariable: 0: controls 1: low risk cancer, 2: medium risk cancer3: high risk cancer). A DAG (directed acyclic graph) wasbuilt including potential confounders of the night shift work-prostate cancer association that were considered for inclusion

in multivariate analysis (Supporting Information Fig. S1).Each of these variables was added one at the time in thebasic model that included age, center and educational level,to assess whether they would modify the association betweennight shift work and prostate cancer. No variable changedthe OR for night shift work by [mt]10%. In the final model,we kept as potential confounders variables for which therespective risk estimates in the model reached a significancelevel of p< 0.25. Family history of prostate cancer, leisuretime physical activity, smoking status, past sun exposure andmeat consumption were included in the final model, in addi-tion to age, center and educational level. Most potential con-founders had a small number of missing information (<1%)but chronotype had 20% and diet related variables 15% ofmissing data. In addition, cumulative frequency was missingin 29% of night workers while cumulative duration in <1%.To increase efficiency and minimize selection bias, we per-formed multiple imputation of missing values using chainedequations.30 Multiple imputations were done separately forcases and controls. Distributions in imputed datasets weresimilar to the distributions in the observed data, with this weassumed the data was missing at random. We generated 30complete data sets, which were analyzed individually toobtain a set of parameters and then combined into overallestimates using Rubin’s rule.

ResultsTable 1 shows the characteristics of the 1,095 incident pros-tate cancer cases and 1,388 population controls with com-plete shift work information. Cases reported morefrequently family history of prostate cancer (20.9 vs. 6.6%)and lower educational level (less than primary school 23.5vs. 17.5%), compared with controls. Total energy consump-tion, past alcohol consumption (at age of 30–40) and meatconsumption were higher among cases (p< 0.05). No signif-icant differences were observed for age, BMI, smoking

Table 1. Potential prostate cancer risk factors among participants in the MCC-Spain study (numbers may differ due to missing values; SD:standard deviation) (Continued)

Controls (N 5 1388) Cases (N 5 1095) p values

Factor N (%) N (%)

Physical activity (METS-hr/year)1

Inactive 543 (39.1) 436 (39.8) 0.756

Slightly active 174 (12.5) 148 (13.5)

Moderately active 163 (11.7) 130 (11.9)

Very active 508 (36.6) 381 (34.8)

Diet habits; mean (SD)

Total energy consumption (kcal/day) 2029 (20.5) 2083 (22.6) 0.074

Past alcohol consumption (g ethanol/day)2 30.0 (0.8) 32 (1.2) 0.144

Meat consumption (g/day) 72 (0.9) 76 (1.5) 0.0352

Vegetables consumption (g/day) 170 (2.9) 175 (3.8) 0.311

1Leisure time physical activity assessed over the last decade.2Assesed at 30–40 years.

Epidemiology

Papantoniou et al. 5

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

status, sleep habits or leisure time physical activity. The dif-ferent characteristics of subjects with and without nightshift work history, as well as between permanent and rotat-ing night shift workers are shown in Supporting Informa-tion Table S1.

Twenty-nine percent of controls and 33% of cases hadever worked at night for more than 1 year (Table 2). Rotatingnight work (19% in cases and 17% in controls) was more fre-quent compared with permanent night work (14% in casesand 12% in controls). Night work was found in a wide rangeof occupational settings such as restaurant services (10.4%),transportation (8.0%), health (4.3%) and industry (industrialoperators: 4.2%) such as food, tobacco and beverages industry(3.6%) (Supporting Information Table S2). Having everworked at night was associated with an increased but not sig-nificant risk for prostate cancer [OR 1.14; 95% CI (95% CI)0.94, 1.37], compared with subjects that had never worked atnight (Table 2). ORs were slightly higher for rotating night(OR 1.16; 95% CI 0.92, 1.46) than for permanent night shiftwork (OR 1.10; 95% CI 0.85, 1.43), after adjusting for poten-

tial confounders. ORs for night shift work were similarbetween the less and more adjusted models. Inclusion ofimputed confounders in the models did not change the riskestimates.

The association between lifetime cumulative duration andfrequency of night work and prostate cancer risk are pre-sented in Table 3. We observed an adjusted OR of 1.38 (1.05,1.81) for subjects that had been engaged in night shift workfor �28 years, the top tertile of exposure and a statisticallysignificant trend (p-trend5 0.047). The respective risk esti-mates were 1.40 (0.83, 2.37) for permanent and 1.37 (0.97,1.94) for rotating night shift work. Lifetime cumulative num-ber of night shifts (�2,857 nights: OR 1.30; 0.97, 1.74) wasalso associated with an increased prostate cancer risk forboth types of night work (p-trend5 0.084). Analysis usingthe nonimputed duration and frequency variables gave verysimilar results to the imputed ones that are presented here.

In stratified analysis by chronotype (Table 4), OR of havingever performed night shift was highest among evening types(OR 1.50; 95% CI 0.85, 2.66), followed by morning types (1.14;

Table 2. Association of night shift work and prostate cancer risk in the MCC-Spain case-control study (OR: odds ratio; 95% CI: 95% confi-dence interval)

Controls (N 5 1388) Cases (N 5 1095) OR (95% CI)1 OR (95% CI)2

N (%) N (%)

Never night work 984 (70.9) 733 (66.9) 1 (Ref) 1 (Ref)

Ever night work 404 (29.1) 362 (33.1) 1.14 (0.95, 1.36) 1.14 (0.94, 1.37)

Types of night work

Permanent night work 165 (11.9) 156 (14.3) 1.10 (0.86, 1.41) 1.10 (0.85, 1.43)

Rotating night work 239 (17.2) 206 (18.8) 1.17 (0.93, 1.46) 1.16 (0.92, 1.46)

1Adjusted for age, center and educational level.2Adjusted for age, center, educational level, family history of prostate cancer, physical activity over the past decade, smoking status, past sun expo-sure and daily meat consumption.

Table 3. Association of different types of night shift work and prostate cancer risk by metrics of duration and frequency (numbers may differdue to missing values; OR: odds ratio, 95% CI: 95% confidence interval)

Ever night work Permanent night work Rotating night work

Never night work Controls/Cases (N) OR (95% CI)1 Controls/Cases (N) OR (95% CI)1 Controls/Cases (N) OR (95% CI)1

984/733 1 (Ref) 984/733 1 (Ref) 984/733 1 (Ref)

Lifetime cumulative duration of night work (years)

�10 Years 145/128 1.10 (0.83, 1.45) 83/75 1.07 (0.75, 1.51) 77/73 1.21 (0.85, 1.74)

11–27 Years 130/92 0.94 (0.69, 1.27) 52/41 1.01 (0.65, 1.56) 78/47 0.84 (0.56, 1.26)

�28 Years 128/138 1.38 (1.05, 1.81) 30/36 1.40 (0.83, 2.37) 83/85 1.37 (0.97, 1.94)

p-Trend 0.047 0.251 0.158

Lifetime cumulative frequency of night work (night shifts)

�1152 Nights 99/85 1.03 (0.75, 1.42) 31/28 0.97 (0.59, 1.59) 85/71 1.08 (0.78, 1.50)

1153–2856 nights 90/71 1.09 (0.78, 1.52) 17/17 1.06 (0.63, 1.77) 56/48 1.18 (0.78, 1.77)

�2857 Nights 96/100 1.30 (0.97, 1.74) 37/38 1.27 (0.85, 1.91) 41/42 1.26 (0.80, 1.99)

p-Trend 0.084 0.247 0.254

1ORs are based on the imputed variables for duration and frequency and are adjusted for age, center, educational level, family history of prostatecancer, physical activity over the past decade, smoking status, past sun exposure and daily meat consumption.

Epidemiology

6 MCC-Spain case-control study

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

0.87, 1.51) and neither types (1.02; 0.72, 1.44), compared withnever shift workers with the same chronotype. However, theinteraction term was not statistically significant for permanentor rotating night shift work, after adjusting for potential con-founders. Results were similar using the five- and seven-category chronotype variable (data not shown). Morning chro-notypes with long-term exposure (�28 years) had an OR of1.79 (1.16, 2.76) and p-trend5 0.017, while the estimates were1.00 (0.62, 1.64) and 1.33 (0.56, 3.16) for neither and eveningtypes, respectively. Evening types had a slightly higher basalrisk for prostate cancer (OR 1.10; 95% CI 0.82, 1.46) afteradjusting for shift-work, while morning types had no risk (0.98;0.79, 1.21), compared with neither chronotypes, although differ-ences were not statistically significant. Similar results wereobserved when limiting the analysis to never shift workers.

Table 5 shows the association of night work with severityof prostate cancer. We found wide differences between low,medium and high risk tumors across all shift work exposuremetrics. A positive and statistically significant association wasfound between night shift work and high risk prostatetumors (RRR 1.40; 1.05, 1.86) according to the D’Amico clas-sification. After excluding controls differences remained sig-nificant (p5 0.013) between the low and high risk group,that can be interpreted as an interaction between night workand disease severity. Statistically significant trends were foundacross tertiles of increasing cumulative duration (p-trend=0.027) and frequency (p-trend5 0.007) for high risktumors. Subjects in the top tertiles of cumulative durationand frequency had RRRs of 1.63 (1.08, 2.45) and 1.78 (1.17,2.69), respectively, for high risk prostate cancer, compared

with never exposed subjects. Similar results were found whenpatients were classified using Gleason score alone (SupportingInformation Table S3). Subjects with history of night shiftwork and Gleason score> 7 at diagnosis had a higher risk(RRR 1.43; 0.99, 2.07), compared to subjects with lower score(Gleason score5 7: RRR 1.10; 0.86, 1.41 and Gleason score-< 7: RRR 1.09; 0.85, 1.38).

We did not find statistically significant interactions betweennight shift work and age groups (<60, �60) or obesity(BMI<30, �30). We repeated analyses removing the smallnumber of subjects that had reported rotating shift work butno night shifts from the control group and results were verysimilar. A sensitivity analysis was also performed excluding alljobs with shift durations shorter than 4 hr or longer than 12and jobs with shift work characterization but missing informa-tion on the time schedules and results remained unchanged.

DiscussionIn this large population based study, we found an associationbetween night shift work and prostate cancer particularly fortumors with worse prognosis. Risk was higher in subjects withlonger and more frequent exposure to night shift work. Wefound some indication of an effect modification by chronotype.

A higher although not statistically significant risk of pros-tate cancer was found for night shift work. These findingsare in line with most, but not all, existing cohort and case-control studies on shift work and prostate cancer.9–12,31–33

Slightly higher risks were observed among rotating nightworkers, compared to permanent night shift workersalthough this difference was not consistent in all exposure

Table 4. Association of night shift work and prostate cancer stratified by chronotype (numbers may differ due to missing values; OR: oddsratio, 95% CI: 95% confidence interval)

Morning chronotype (N 5 993) Neither chronotype (N=720) Evening chronotype (N=270)

Controls/cases (N) OR (95% CI)1 Controls/cases (N) OR (95% CI)1 Controls/cases (N) OR (95% CI)1

Never night work 381/300 1 (Ref) 303/217 1 (Ref) 102/76 1 (Ref)

Ever night work 160/152 1.14 (0.87, 1.51) 110/90 1.02 (0.72, 1.44) 43/49 1.50 (0.85, 2.66)

Types of night work

Permanent night work 60/67 1.19 (0.80, 1.76) 49/36 0.90 (0.55, 1.48) 20/24 1.57 (0.76, 3.27)

Rotating night work 100/85 1.12 (0.80, 1.56) 61/54 1.11 (0.73, 1.70) 23/25 1.44 (0.70, 2.93)

Lifetime cumulative duration of night work (years)

�10 Years 65/51 0.95 (0.63, 1.43) 38/33 1.16 (0.68, 1.97) 14/19 1.92 (0.80, 4.54)

11–27 Years 52/39 0.90 (0.57, 1.40) 30/20 0.87 (0.46, 1.63) 13/14 1.30 (0.55, 3.07)

�28 Years 43/61 1.79 (1.16, 2.76) 42/36 1.00 (0.62, 1.64) 15/16 1.33 (0.56, 3.16)

p-Trend 0.017 0.876 0.619

Lifetime cumulative frequency of night work (night shifts)

�1152 Nights 45/37 0.93 (0.59, 1.46) 28/21 1.03 (0.59, 1.81) 5/11 1.86 (0.70, 4.95)

1153–2856 Nights 35/33 1.22 (0.73, 2.02) 22/14 0.92 (0.46, 1.85) 13/13 1.22 (0.51, 2.92)

�2857 Nights 36/42 1.37 (0.88, 2.13) 30/26 1.08 (0.65, 1.80) 10/12 1.55 (0.66, 3.66)

p-Trend 0.114 0.840 0.431

1Adjusted for age, center, educational level, family history of prostate cancer, physical activity, smoking status, past sun exposure and daily meatconsumption.

Epidemiology

Papantoniou et al. 7

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

classifications. One previous study compared types of nightwork with results similar to ours.11 For both night shift worktypes, higher risks for prostate cancer were found after longerduration and frequency of exposure over lifetime, suggestinga dose-response association. Lower sleep duration, one of theproposed mechanisms for the association between night shiftwork and cancer, has been also recently associated with anincreased prostate cancer risk.34,35 Finally, an ecological studyshowed that only prostate cancer among all tumors exam-ined, was significantly correlated with light at night levelsprovided by satellites at the country level.36

We found increased prostate cancer risk among nightshift workers, particularly for high risk tumors, suggesting apossible association between night shift work and types ofcancer with worse prognosis and higher mortality.37 Resultsbased on the D’Amico classification were confirmed in anal-ysis by patient�s Gleason score, a better marker of diseaseaggressivity.38 None of the previous studies reported on fac-tors related to disease prognosis. A recent study showed anassociation between current shift work and elevated PSAlevels.20 Increased PSA levels have been shown to be predic-tive not only of a higher prostate cancer risk, but also worseprognosis and increased mortality.21 Chronic disruption ofthe circadian genes may lead to cell cycle deregulation andcould increase DNA damage replication errors, mutations

and cancer initiation.17 Circadian disruption also acceleratestumor development and growth in experimental animalmodels.18,19 Our data suggest that night shift work thatinvolves circadian disruption may be associated with cancerpromotion and therefore decreased survival. One recentconference report has provided some evidence for decreas-ing survival of breast cancer among both fixed and rotatingnight shift workers.22 Given the lack of well-known factorsrelated to prostate cancer severity, a better understandingand confirmation of the possible contribution of night shiftwork might be valuable.

We found a higher risk of prostate cancer among subjectswith night shift work history and an evening chronotype.Evening types might be genetically more susceptible to circa-dian disruption or cancer.39,40 An increased breast cancerrisk has been reported among women with evening prefer-ence,27 while neither preference was identified as more riskyin another study.28 However, evening types have been sug-gested to better tolerate night shift work due to their greatercapacity for adaptation.25,27,41 Adaptation is a progressiveshift, mostly phase delay, of the circadian rhythms thatrequires several consecutive nights of work and is hardlyachieved by real world night shift workers.42 Partial shiftadaptation, experienced by the vast majority of night shiftworkers, improves nighttime sleepiness, daytime sleep quality

Table 5. Association of night shift work and severity of prostate cancer (D’Amico classification) in cases with clinical information (OR: oddsratio; RRR: relative risk ratio; 95% CI: 95% confidence interval)

Controls(N)

Cases withclinical

information(N 5 1068) Low risk (N 5 381) Medium risk (N 5 413) High risk (N 5 274)

Cases(N)

OR(95% CI)1

Cases(N)

RRR(95% CI)1

Cases(N)

RRR(95% CI)1

Cases(N)

RRR(95% CI)1

Never night work 984 713 1 (Ref) 269 1 (Ref) 276 1 (Ref) 168 1 (Ref)

Ever night work 404 355 1.15 (0.95, 1.38) 112 1.04 (0.79, 1.35) 137 1.09 (0.85, 1.40) 106 1.40 (1.05, 1.86)

Types of night work

Permanentnight work

165 155 1.12 (0.87, 1.45) 48 0.98 (0.68, 1.42) 63 1.11 (0.79, 1.56) 44 1.35 (0.91, 1.99)

Rotating night work 239 200 1.17 (0.93, 1.47) 64 1.08 (0.78, 1.50) 74 1.07 (0.78, 1.46) 62 1.44 (1.02, 2.03)

Lifetime cumulative duration of night work (years)

�10 Years 145 124 1.10 (0.83, 1.45) 43 1.05 (0.71, 1.55) 46 1.00 (0.68, 1.45) 35 1.32 (0.86, 2.02)

11–27 Years 130 91 0.95 (0.70, 1.29) 35 1.08 (0.71, 1.63) 26 0.66 (0.42, 1.05) 30 1.26 (0.80, 1.98)

�28 Years 128 136 1.40 (1.05, 1.85) 32 0.98 (0.64, 1.51) 64 1.62 (1.14, 2.31) 40 1.63 (1.08, 2.45)

p-Trend 0.039 0.980 0.023 0.027

Lifetime cumulative frequency of night work (night shifts)

�1152 Nights 99 75 1.03 (0.75, 1.41) 27 0.99 (0.63, 1.53) 24 0.97 (0.64, 1.47) 24 1.17 (0.73, 1.87)

1153–2856Nights

90 69 1.10 (0.79, 1.54) 24 1.16 (0.72, 1.87) 22 0.98 (0.63, 1.52) 23 1.26 (0.76, 2.07)

�2857 Nights 96 94 1.32 (0.99, 1.77) 26 0.97 (0.63, 1.51) 35 1.33 (0.91, 1.94) 33 1.78 (1.17, 2.69)

p-Trend 0.064 0.944 0.168 0.007

1Adjusted for age, center, educational level, family history of prostate cancer, physical activity, smoking status, past sun exposure and daily meatconsumption.

Epidemiology

8 MCC-Spain case-control study

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

and reduces some health risks.41 However, it is unknown ifshift work adaptation (partial or complete) reduces orincreases a night shift worke�rs future cancer risk. Contrary tothe prediction, we found higher prostate cancer risks amongevening types that were expected to have better adapted totheir shifts. Similarly, a dose-response association was foundbetween night work related phase shift and breast cancer inone study26 and increasing number of consecutive nights inanother,43 suggesting that shift adaptation might increasebreast cancer risk. Subjects that adapt are more synchronizedwith the light-dark and sleep-wake cycles, however mightexperience internal desynchronization and temporal misalign-ment (jet-lag) of genetic and metabolic processes that mightin turn increase cancer risk.15,44 However, we found thatprostate cancer risk diminished in the long run among eve-ning types, while it increased for morning types. This sug-gests a possible effect of chronic mismatch among early typeson future cancer risk, also supported by lower levels of mela-tonin production found among night shift workers withmorning preference.45 Thus both morning and evening chro-notypes may be at an increased risk of prostate cancer afternight work and the role of chronotype as an effect modifierneeds to be further evaluated.

Mechanisms involved in the possible association betweennight shift work and prostate cancer risk are complex. Nightshift workers are exposed to light during night and acute sleepdeprivation that may result in suppression or phase shift oftheir normal nocturnal melatonin production and other circa-dian rhythms.45 Long-term exposure to lower melatonin levelsmight relate to a higher prostate cancer risk, since melatoninhas multiple antineoplastic actions. Melatonin directly inhibitsproliferation of cultured prostate cancer cells and reducesgrowth of prostate cancer in rats.46,47 Furthermore melatoninreduces oxidative stress that may accelerate prostate cancerprogression.48,49 Androgens and other sex hormones have acentral role on prostate development and possibly sex hor-mone disruption is implicated in prostate carcinogenesis,although evidence is still inconclusive.50 We found higher lev-els of androgens and progestagens among male night shiftworkers in a cross-sectional study in Barcelona (Papantoniouet al. unpublished data). Sleep deprivation and sleep disruptionrelated with daytime sleep are often observed among nightshift workers and might explain part of the increased cancerrisk31,34; however in the present analysis adjustment for sleepduration or sleeping problems did not change results. Finally,lower exposure to sun and subsequent vitamin D deficiency isan alternative hypothesis for the possible carcinogenicity ofnight shift work that we accounted for, although the patternsof past sun exposure where not significantly different betweenday and night workers.

Strengths and limitations

This study extends knowledge on circadian disruption andprostate cancer risk and provides evidence based on an exten-sive evaluation of exposure. It is the first study to analyze chro-

notype information with relation to prostate cancer with theuse of a validated questionnaire,24,29 and assess disease severityusing a well established clinical classification. A possible limita-tion of the study is that chronotype was assessed only once, inthe present and not when subjects worked at night. Chronotypeand sleep habits depend on age, although there is evidence thatchronotype tracks through life and current chronotype highlycorrelates with self-reported chronotype at adolescence andearly adulthood.23 We examined lifetime occupational historywith detailed shift work characteristics for each job and usedmetrics such as cumulative duration and frequency. The highpercentage of missing values in cumulative frequency waslargely attributed to the fact that this question was only doneto subjects that had previously reported night work. However,our night shift work definition was not based on self-reportsbut time-schedules for each job as they were considered morereliable. We applied a common definition for night shift workand reclassified subjects to fit the predefined groups. Given thevariety and often irregularity of time schedules and shift pat-terns in the real life work environment it is difficult to classifythe exposure and it is unknown which time window of work isthe most detrimental. Exposure misclassification is inevitable,thus, being nondifferential would attenuate risk estimatestoward the null. Although the study is big including [mt]1,000cases, in stratified analyses numbers were smaller and we mighthave lacked statistical power to detect significant associationsand interactions. The fact that a variety of occupations wasincluded in the present study and not a single professionincreases the external validity of the results. However, con-founding by other occupational exposures/agents related tonight shift work and cancer might have occurred. We checkedthe type of occupations in day and non-day jobs as well astypes of night shift jobs in cases and controls and found no dif-ferences. Recall bias is not likely since the probable carcinoge-nicity of night shift work is still not a major concern in Spainand the subjects were told that the study was designed to assessenvironmental and genetic risk factors for cancer and not spe-cifically the effects of night shift work. Selection bias in caseswas minimized by enrolling all new prostate cancer cases in thehospitals participating in the study. Selection bias of controls ispossible and controls that were currently engaged in night shiftwork may have participated less frequently. However, phonecalls were performed repeatedly in different times of the day.Furthermore, results did not change after excluding currentworkers (25% of the study population) from the analysis, show-ing that this bias is not likely to have affected risk estimates.The assessment of a wide range of potential confoundersincluding diet, sleep habits, exposure to sun as a proxy for defi-ciency of vitamin D and other lifestyle characteristics is anadditional asset of the analysis and enabled us to deal with pos-sible confounding.

In conclusion, our study suggests a possible role of nightshift work in prostate cancer risk. In this large population-based study, we found a higher risk for prostate canceramong night shift workers with longer exposure over lifetime.

Epidemiology

Papantoniou et al. 9

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

Overall risk was higher among subjects with an evening chro-notype, but also increased in morning chronotypes after lon-ger durations of night shift work. Night shift work wasassociated with disease severity at diagnosis suggesting a pos-sible association between night shift work and lower prostatecancer survival.

Authors’ ContributionKP performed the statistical analysis and wrote the manuscript. GCV andMK were involved in the design of the study, in the acquisition of data and

contributed in writing the final version of the manuscript. AE participated inthe statistical analysis and contributed in writing the manuscript. NA, BPG,JB, IGA, JL, RP, JJJM, FA, AT and MP participated in study design, acquisi-tion of data and enrolment of patients. All authors read and approved thefinal manuscript.

AcknowledgementsThe authors thank all participants, interviewers and the data manager of thestudy Judith Cirac. They would also like to thank Juan Alguacil, Miguel San-tiba~nez, Victor Moreno and Laura Costas for their valuable comments in thetext. The authors declare no conflicts of interest.

References

1. Ferlay J, Steliarova-Foucher E, Lortet-Tieulent J,et al. Cancer incidence and mortality patterns ineurope: estimates for 40 countries in 2012. Eur JCancer 2013;49:137421403.

2. Siegel R, Naishadham D, Jemal A. Cancer statis-tics, 2013. CA Cancer J Clin 2013;63:11–30.

3. Lorelei A, Mucci LBS, Hans-Olov Adami. Pros-tate Cancer. In: Adami HO, Hunter D, Tricho-poulos D, eds. Textbook of cancer epidemiology,2nd edn. Oxford University Press, 2008. Chapter20. 517–54.

4. Eeles RA, Olama AA, Benlloch S, et al. Identifica-tion of 23 new prostate cancer susceptibility lociusing the iCOGS custom genotyping array. NatGenet 2013;45:385–91, 91e1-2.

5. Sharifi N, Gulley JL, Dahut WL. Androgen depri-vation therapy for prostate cancer. JAMA 2005;294:238–44.

6. Sigurdardottir LG, Valdimarsdottir UA, Fall K,et al. Circadian disruption, sleep loss, and pros-tate cancer risk: a systematic review of epidemio-logic studies. Cancer Epidemiol Biomarkers Prev2012;21:1002–11.

7. IARC. Painting, firefighting and shiftwork IARCmonographs, vol. 98. IARC, 2010. http://monographs.iarc.fr.

8. Jia Y, Lu Y, Wu K, et al. Does night workincrease the risk of breast cancer? A systematicreview and meta-analysis of epidemiological stud-ies. Cancer Epidemiol 2013;37:197–206.

9. Conlon M, Lightfoot N, Kreiger N. Rotating shiftwork and risk of prostate cancer. Epidemiology2007;18:182–3.

10. Kubo T, Oyama I, Nakamura T, et al. Industry-based retrospective cohort study of the risk ofprostate cancer among rotating-shift workers. IntJ Urol 2012;18:206–11.

11. Kubo T, Ozasa K, Mikami K, et al. Prospectivecohort study of the risk of prostate cancer amongrotating-shift workers: findings from the japancollaborative cohort study. Am J Epidemiol 2006;164:549–55.

12. Parent ME, El-Zein M, Rousseau MC, et al. Nightwork and the risk of cancer among men. Am JEpidemiol 2012;176:751–9.

13. Stevens RG, Hansen J, Costa G, et al. Considera-tions of circadian impact for defining ’shift work’in cancer studies: IARC working group report.Occup Environ Med 2010;68:154–62.

14. Fritschi L, Glass DC, Heyworth JS, et al. Hypoth-eses for mechanisms linking shiftwork and can-cer. Med Hypotheses 2011;77:430–436.

15. Fu L, Lee CC. The circadian clock: pacemakerand tumour suppressor. Nat Rev Cancer 2003;3:350–61.

16. Takahashi JS, Shimomura K, Kumar V. Searchingfor genes underlying behavior: lessons from circa-dian rhythms. Science 2008;322:909–12.

17. Sancar A, Lindsey-Boltz LA, Kang TH, et al. Cir-cadian clock control of the cellular response toDNA damage. FEBS Lett 2010;584:2618–25.

18. Filipski E, King VM, Li X, et al. Host circadianclock as a control point in tumor progression.J Natl Cancer Inst 2002;94:690–7.

19. Mao L, Dauchy RT, Blask DE, et al. Circadiangating of epithelial-to-mesenchymal transition inbreast cancer cells via melatonin-regulation ofGSK3beta. Mol Endocrinol 2012;26:1808–20.

20. Flynn-Evans EE, Mucci L, Stevens RG, et al.Shiftwork and prostate-specific antigen in thenational health and nutrition examination survey.J Natl Cancer Inst 2013;105:1292–7.

21. Orsted DD, Nordestgaard BG, Jensen GB, et al.Prostate-specific antigen and long-term predictionof prostate cancer incidence and mortality in thegeneral population. Eur Urol 2012;61:865–74.

22. Hansen J. 0200 Night shiftwork and breast cancersurvival in Danish women. Occup Environ Med2014;71Suppl1:A26.

23. Roenneberg T, Kuehnle T, Juda M, et al. Epide-miology of the human circadian clock. Sleep MedRev 2007;11:429–38.

24. Zavada A, Gordijn MC, Beersma DG, et al. Com-parison of the munich chronotype questionnairewith the Horne-ostberg’s Morningness-eveningness score. Chronobiol Int 2005;22:267–78.

25. Erren TC. Shift work and cancer research: canchronotype predict susceptibility in night-shiftand rotating-shift workers? Occup Environ Med2013;70:283–4.

26. Fritschi L, Erren TC, Glass DC, et al. The associ-ation between different night shiftwork factorsand breast cancer: a case-control study. Br J Can-cer 2013;109:2472–80.

27. Hansen J, Lassen CF. Nested case-control studyof night shift work and breast cancer risk amongwomen in the Danish military. Occup EnvironMed 2012;69:551–6.

28. Ramin C, Devore EE, Pierre-Paul J, et al. Chro-notype and breast cancer risk in a cohort of USnurses. Chronobiol Int 2013;30:1181–6.

29. Roenneberg T, Wirz-Justice A, Merrow M. Lifebetween clocks: daily temporal patterns of humanchronotypes. J Biol Rhythms 2003;18:80–90.

30. White IR, Royston P, Wood AM. Multiple impu-tation using chained equations: issues and guid-ance for practice. Stat Med 2011;30:377–99.

31. Gapstur SM, Diver WR, Stevens VL, et al. Workschedule, sleep duration, insomnia, and risk offatal prostate cancer. Am J Prev Med 2014;46:S26–33.

32. Schwartzbaum J, Ahlbom A, Feychting M. Cohortstudy of cancer risk among male and female shiftworkers. Scand J Work Environ Health 2007;33:336–43.

33. Yong M, Blettner M, Emrich K, Nasterlack M,Oberlinner C, Hammer GP. A retrospectivecohort study of shift work and risk of incidentcancer among German male chemical workers.Scand J Work Environ Health, 2014;40:502–10.

34. Kakizaki M, Inoue K, Kuriyama S, et al. Sleepduration and the risk of prostate cancer: theohsaki cohort study. Br J Cancer 2008;99:176–8.

35. Sigurdardottir LG, Valdimarsdottir UA, MucciLA, et al. Sleep disruption among older men andrisk of prostate cancer. Cancer Epidemiol Bio-markers Prev 2013;22:872–9.

36. Kloog I, Haim A, Stevens RG, et al. Global co-distribution of light at night (LAN) and cancersof prostate, colon, and lung in men. ChronobiolInt 2009;26:108–25.

37. Hernandez DJ, Nielsen ME, Han M, et al. Con-temporary evaluation of the D’amico risk classi-fication of prostate cancer. Urology 2007;70:931–5.

38. Egevad L, Granfors T, Karlberg L, et al. Prognos-tic value of the gleason score in prostate cancer.BJU Int 2002;89:538–42.

39. Archer SN, Robilliard DL, Skene DJ, et al. Alength polymorphism in the circadian clock genePer3 is linked to delayed sleep phase syndromeand extreme diurnal preference. Sleep 2003;26:413–5.

40. Zhu Y, Brown HN, Zhang Y, et al. Period3 struc-tural variation: a circadian biomarker associatedwith breast cancer in young women. Cancer Epi-demiol Biomarkers Prev 2005;14:268–70.

41. Saksvik IB, Bjorvatn B, Hetl H, et al. Individualdifferences in tolerance to shift work–a systematicreview. Sleep Med Rev 2011;15:221–35.

42. Folkard S. Do permanent night workers show cir-cadian adjustment? A review based on the endog-enous melatonin rhythm. Chronobiol Int 2008;25:215–24.

43. Lie JA, Kjuus H, Zienolddiny S, et al. Night workand breast cancer risk among Norwegian nurses:assessment by different exposure metrics. Am JEpidemiol 2011;173:1272–9.

44. Blask DE, Hill SM, Dauchy RT, et al. Circadianregulation of molecular, dietary, and metabolicsignaling mechanisms of human breast cancergrowth by the nocturnal melatonin signal and theconsequences of its disruption by light at night.J Pineal Res 2011;51:259–69.

45. Papantoniou K, Pozo OJ, Espinosa A, et al. Circa-dian variation of melatonin, light exposure, and

Epidemiology

10 MCC-Spain case-control study

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC

diurnal preference in day and night shift workersof both sexes. Cancer Epidemiol Biomarkers Prev2014;23:1176–86.

46. Jung-Hynes B, Schmit TL, Reagan-Shaw SR, et al.Melatonin, a novel Sirt1 inhibitor, imparts anti-proliferative effects against prostate cancer invitro in culture and in vivo in TRAMP model.J Pineal Res 2011;50:140–9.

47. Marelli MM, Limonta P, Maggi R, et al. Growth-inhibitory activity of melatonin on humanandrogen-independent DU 145 prostate cancercells. Prostate 2000;45:238–44.

48. Galano A, Tan DX, Reiter RJ. Melatonin as anatural ally against oxidative stress: a physico-chemical examination. J Pineal Res 2011;51:1–16.

49. Nguyen HL, Zucker S, Zarrabi K, et al. Oxidativestress and prostate cancer progression are elicitedby membrane-type 1 matrix metalloproteinase.Mol Cancer Res 2011;9:1305–18.

50. Roddam AW, Allen NE, Appleby P, et al. Endog-enous sex hormones and prostate cancer: a col-laborative analysis of 18 prospective studies.J Natl Cancer Inst 2008;100:170–83.

Epidemiology

Papantoniou et al. 11

Int. J. Cancer: 00, 00–00 (2015) VC 2014 UICC