Annals of Epidemiology 24 (2014) 151e159
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Review article
Personal hair dye use and bladder cancer: a meta-analysis Federica Turati ScD, PhD a, Claudio Pelucchi ScD b, Carlotta Galeone ScD, PhD c, Adriano Decarli ScD, PhD a, c, Carlo La Vecchia MD b, c, * a
Struttura Complessa di Statistica Medica, Biometria e Bioinformatica. Fondazione IRCCS Istituto Nazionale Tumori, Milan, Italy Department of Epidemiology, IRCCS - Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy c Department of Clinical Sciences and Community Health, University of Milan, Milan, Italy b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 17 May 2013 Accepted 3 November 2013 Available online 14 November 2013
Despite considerable research, the issue of hair dyes and bladder cancer is still open to discussion. In January 2013, we searched in PubMed/EMBASE to identify observational studies investigating the association between personal use of hair dyes and bladder cancer incidence/mortality. Pooled relative risks (RRs) and corresponding 95% confidence intervals (CIs) were calculated using random-effects models. Fifteen case-control and two cohort studies were available for meta-analysis (8504 cases/deaths, 14,102 controls, and 617,937 persons at risk). Compared with no use, the pooled RR of bladder cancer for personal use of any type of hair dyes was 0.93 (95% CI, 0.82e1.05), with moderate heterogeneity among studies (I2 ¼ 34.1%, P ¼ .07). Similar RRs were found for females (RR ¼ 0.95) and males (RR ¼ 0.81). Based on seven studies, the pooled RR for personal use of permanent hair dyes was 0.92 (95% CI, 0.77e1.09). Compared with no use, no association was observed for the highest categories of duration of use and lifetime frequency of use of both any type of dyes and permanent dyes. The pooled RR from four studies reporting results for use of dark-colored dyes was 1.29 (95% CI, 0.98e1.71). This meta-analysis allows to definitively exclude any appreciable excess risk of bladder cancer among personal hair dye users. Ó 2014 Elsevier Inc. All rights reserved.
Keywords: Bladder cancer Hair dyes Meta-analysis Risk factors
Introduction Hair dyeing is a common practice in several areas of the world, with an estimated 50%e80% of women having used hair dyes in their life in the United States, Japan, and the European Union [1]. Hair dyes are classified as permanent (or oxidative), semipermanent, and temporary. Semipermanent and temporary hair dyes are nonoxidative and include colored compounds that stain hair directly. Permanent hair dyes, with a market share in the European Union and the United States of approximately 80%, contain one or several “primary intermediates,” that is, arylamines, such as para-phenylenediamine, para-toluenediamine), and other substituted para-diamines, orthoor para-aminophenols, and dye “couplers.” In the presence of hydrogen peroxide, the primary intermediates and couplers react with one another to form colored oligomers [2,3]. A number of aromatic amines contained in hair dyes until the 1980s are mutagenic in vitro [4] and carcinogenic in animals and humans [1,5]. A study identified small amounts of 4-aminobiphenyl, an aromatic amine, which is a recognized human urinary * Corresponding author. Dipartimento di Epidemiologia, IRCCS - Istituto di Ricerche Farmacologiche “Mario Negri”, via Giuseppe La Masa 19, 20156 Milan, Italy. Tel.: þ39 0239014527; fax: þ39 0233200231. E-mail address: [email protected] (C. La Vecchia). 1047-2797/$ e see front matter Ó 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.annepidem.2013.11.003
bladder carcinogen, in eight of the 11 hair dyes tested [6]. Darker colors are formed by using higher concentrations of primary intermediates and have been therefore suggested to pose a higher risk of urinary bladder cancer than nondark-shade hair dyes [2], although the issue remains open to discussion. The International Agency for Research on Cancer (IARC) recently concluded that occupational exposures to hair dyes as a hairdresser or barber were probably carcinogenic to humans, whereas personal use of hair colorants was not classifiable as to its carcinogenicity to humans [1]. When referring to bladder cancer, the IARC Working Group considered limited the available evidence of an increased risk among hairdressers (i.e., occupational exposure) and inadequate the evidence on the role of personal use of hair dyes. The difference between personal and occupational exposure reflects different doses of exposure, but also the occupational studies are subject to a different series of socioeconomic and lifestyle-related confounding factors, including tobacco. With regards to personal use of dyes, although an excess bladder cancer risk has been seen in some studies [7,8], the overall results are not converging [9,10]. An earlier quantitative review found a relative risk (RR) of bladder cancer for ever versus never users between 1.0 and 1.1, based on various assumptions [11]. The most recent meta-analysis, published in 2008 and including 11 casecontrol studies and one cohort study, found no association
152
F. Turati et al. / Annals of Epidemiology 24 (2014) 151e159
between personal hair dye use and bladder cancer risk [12]. Thereafter, three other case-control [13e15] and one cohort [16] studies have been published, including the New England bladder cancer study, with almost 1200 cases [14]. We, therefore, updated the summary evidence on personal use of hair dyes and bladder cancer till January 2013, addressing some critical issues in exposure assessment, including the type of hair-coloring product, color, duration of use, and lifetime frequency of use. Materials and methods Systematic search The meta-analysis was conducted following the Meta-analysis Of Observational Studies in Epidemiology guidelines [17]. We performed a computerized MEDLINE search using PubMed from 1966 to January 2013 to identify potentially eligible studies, using the string “(hair) AND (dye* OR colour OR color) AND (neoplasm* OR cancer OR cancers OR tumour*) AND (urinary OR bladder).” A similar strategy was used to search EMBASE (1980 to January 2013). We also checked the reference lists of every article retrieved and those of some relevant systematic reviews [11,18,19] and metaanalyses [9,10,12], including the IARC Monograph focused on aromatic amines and organic dyes, published in 2010 [1]. Articles not in English, unpublished studies, conference abstracts, and posters were not considered. Two authors (F.T. and C.G.) screened the search results independently against the inclusion criteria to identify eligible articles, and discrepancies were discussed and resolved. Inclusion criteria and data collection The articles were included if (1) they presented original data from case-control or cohort studies; (2) the outcome of interest was incidence or mortality from bladder cancer; (3) they analyzed exposure to personal hair dyes; and (4) they provided RR estimates of the association of interest and the corresponding 95% confidence intervals (CIs) or sufficient information for their calculation (e.g., the distribution of cases and noncases across exposure categories or the P value for the test on the RR). Studies on exposure to hair dyes in occupational settings were not considered. If data from the same study were published in multiple articles [8,20e24], we considered the article presenting results based on a larger number of cases. We abstracted the following information from the included studies in a standard format: country, study design, sex of the population, duration of follow-up (for cohort studies) or calendar years of participants inclusion (for case-control studies), number of participants (cases and controls or cohort size), variables adjusted for in the analysis and/or matching variables, and measures of associations that compared personal use of hair dyes (any type of dyes and permanent dyes, when available) with no use of any type of dyes. We also extracted measures for duration of use and lifetime frequency of use, where available. When multiple results for any measure were presented in a single study, we extracted the one adjusted for the largest number of confounding factors. Statistical analysis The first exposure of interest was personal use of any type of hair dyes, as defined in each study. A study reporting quantitative information on permanent dyes only (and not on any type of dyes) was considered in the analysis of any type of dyes as well [20]. The second exposure of interest was personal use of permanent hair dyes [7,8,14,15,20,25]. One study providing data for exclusive use of permanent hair dyes was included [26].
For both exposures of interest, we estimated the pooled RR of bladder cancer for (1) any use compared with no use, (2) highest duration of use (in years) compared with no use, and (3) highest lifetime frequency of use (total number of times) compared with no use. In these analyses, when a study reported risk estimates for males and females only separately, we included both sex-specific RR estimates. For the analysis on duration of use and lifetime frequency of use, we extracted data for the highest categories of exposure in each study compared with no use. In particular, for duration, the highest categories ranged from 6 or more years [27] to 33 or more years of use [26]. Two subgroups of studies were considered as follows: studies with threshold for the highest category of duration of 20 years or longer and those with threshold for the highest category of duration starting from less than 20 years. For lifetime frequency of use, the highest categories ranged from more than 40 times [15] to more than 504 times [26]. Two subgroups of studies were considered as follows: studies with threshold for the highest lifetime frequency of use of 200 times or more and those with threshold for the highest lifetime frequency of use starting from less than 200 times. For both exposures of interest and for each comparison, we conducted subgroup analyses according to selected factors, such as study design, sex, geographic area, and dark-colored dyes (black dyes when available), to explore potential sources of heterogeneity. We also conducted a sensitivity analysis excluding each study in turn to evaluate the robustness of our results. We used random-effects models to combine the study-specific RRs [28]. Statistical heterogeneity among studies was assessed using the c2 test (results were defined as heterogenous for a value of P < .10) [29], and the potential inconsistency was quantified through the I2 statistic, which describes the percentage of total variation across studies that is due to heterogeneity rather than chance [30]. Publication bias was evaluated through funnel plots [31] and with the Egger [32] and Begg tests [33]. All the statistical analyses were performed using the STATA software (version 11; StataCorp, College Station, TX). Results The flow chart of the procedure to select articles is available in the Appendix (Appendix Figure 1). The PubMed and EMBASE searches identified 87 and 192 publications, respectively. After the exclusion of non relevant studies and duplicate publications, we selected 19 potentially eligible unique publications. Other six pertinent publications were obtained from the review of the reference lists of the articles retrieved. Among the 25 unique full-text articles assessed for eligibility, eight publications not satisfying the inclusion criteria were excluded. In particular, one study was excluded because it had a cross-sectional study design [34] and another one because it investigated the association between bladder cancer risk and use of brilliantine [35]. On the other hand, we included one study on cancer of the lower urinary tract in which 90% of the case series had urinary bladder cancer [27], as well as one case-control study considering together bladder (23 cases) and kidney (12 cases) cancers [36]. Finally, we identified 15 case-control [7,8,13e15,25e27,36e42] and two cohort [16,20] studies on personal hair dye use and bladder cancer, which satisfied the inclusion criteria, for a total of 8504 cases/deaths from bladder cancer (8168 cases and 336 deaths) (Table 1). To improve transparency and replicability of our results, we reported in the column “Notes” in Table 1 all the coding decisions we made on each original article (e.g., RRs pooling within studies) [17].
Table 1 Characteristics of the studies included in the meta-analysis Study design
Sex
Period of enrollment/duration of follow-up
Cases
Controls/cohort size
Adjusting and/or matching factors
Notes
Dunham, 1968, USA (Louisiana)
Case-control, HB
M*
1958e1964
132 M (493*)
136 M (527*)
None
Jain, 1977, Canada
Case-control
Both
NA
107
107
Neutel, 1978, Canada
Case-control
Fy
NA
50
50
Crude OR and corresponding 95% CI calculated from the distribution of cases and controls according to use of personal hair dyes All male controls had benign prostatic hypertrophy and all female controls had stress incontinence Crude OR and corresponding 95% CI calculated from the distribution of cases and controls according to use of personal hair dyes
Howe, 1980, Canada Stavraki, 1981, Canada
Case-control, PB Case-control, PB
Fz F
1974e1976 1976
Hartage, 1982, USA US National Bladder Cancer Study Ohno, 1985, Japan
Case-control, PB
Both
Case-control, PB
Nomura, 1989, USA (Hawaiik)
Sex, age
Age
152 (632z) 70
Sex, age Nonspecified
1977e1978
152 (632z) 23 bladder and 12 kidney cancers 2982
5782
Sex, age, area, race, smoking
Fx
1976e1978
66 (293x)
146 (589x)
Age, area, smoking
Case-control, PB
Both
1977e1986
261{
522
Andrew, 1994, USA
Case-control, PB
Both
459
665
Gago-Dominguez, 2001, USA Los Angeles Study
Case-control, PB
Both
Cases: 1994e1998 Controls: 1993e1997 1987e1996
897 (1514#)
897 (1514#)
Sex, age, ethnicity, residence, smoking
Henley and Thun, 2001, USA
Cohort
F
16 y of follow-up from 1982
336 deaths from bladder cancer
547,571 PR
Age, race, smoking, education, blue collar occupation
Kogevinas, 2006, Spain
Case-control, HB
F
1998e2001
152
166
Lin, 2006, USA
Case-control, HB
Both
Case-control: 1999 (ongoing) Case-cohort: 2001 (ongoing)
712
712
Mendelsohn, 2009, China (Shanghai) Shanghai Women Health Study
Cohort
F
1996e2005 7 y of follow-up (average)
32
70,366 PR
The OR for ever use of personal hair dyes was calculated pooling the OR for frequency of use of
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Review article
Personal hair dye use and bladder cancer: a meta-analysis Federica Turati ScD, PhD a, Claudio Pelucchi ScD b, Carlotta Galeone ScD, PhD c, Adriano Decarli ScD, PhD a, c, Carlo La Vecchia MD b, c, * a
Struttura Complessa di Statistica Medica, Biometria e Bioinformatica. Fondazione IRCCS Istituto Nazionale Tumori, Milan, Italy Department of Epidemiology, IRCCS - Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy c Department of Clinical Sciences and Community Health, University of Milan, Milan, Italy b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 17 May 2013 Accepted 3 November 2013 Available online 14 November 2013
Despite considerable research, the issue of hair dyes and bladder cancer is still open to discussion. In January 2013, we searched in PubMed/EMBASE to identify observational studies investigating the association between personal use of hair dyes and bladder cancer incidence/mortality. Pooled relative risks (RRs) and corresponding 95% confidence intervals (CIs) were calculated using random-effects models. Fifteen case-control and two cohort studies were available for meta-analysis (8504 cases/deaths, 14,102 controls, and 617,937 persons at risk). Compared with no use, the pooled RR of bladder cancer for personal use of any type of hair dyes was 0.93 (95% CI, 0.82e1.05), with moderate heterogeneity among studies (I2 ¼ 34.1%, P ¼ .07). Similar RRs were found for females (RR ¼ 0.95) and males (RR ¼ 0.81). Based on seven studies, the pooled RR for personal use of permanent hair dyes was 0.92 (95% CI, 0.77e1.09). Compared with no use, no association was observed for the highest categories of duration of use and lifetime frequency of use of both any type of dyes and permanent dyes. The pooled RR from four studies reporting results for use of dark-colored dyes was 1.29 (95% CI, 0.98e1.71). This meta-analysis allows to definitively exclude any appreciable excess risk of bladder cancer among personal hair dye users. Ó 2014 Elsevier Inc. All rights reserved.
Keywords: Bladder cancer Hair dyes Meta-analysis Risk factors
Introduction Hair dyeing is a common practice in several areas of the world, with an estimated 50%e80% of women having used hair dyes in their life in the United States, Japan, and the European Union [1]. Hair dyes are classified as permanent (or oxidative), semipermanent, and temporary. Semipermanent and temporary hair dyes are nonoxidative and include colored compounds that stain hair directly. Permanent hair dyes, with a market share in the European Union and the United States of approximately 80%, contain one or several “primary intermediates,” that is, arylamines, such as para-phenylenediamine, para-toluenediamine), and other substituted para-diamines, orthoor para-aminophenols, and dye “couplers.” In the presence of hydrogen peroxide, the primary intermediates and couplers react with one another to form colored oligomers [2,3]. A number of aromatic amines contained in hair dyes until the 1980s are mutagenic in vitro [4] and carcinogenic in animals and humans [1,5]. A study identified small amounts of 4-aminobiphenyl, an aromatic amine, which is a recognized human urinary * Corresponding author. Dipartimento di Epidemiologia, IRCCS - Istituto di Ricerche Farmacologiche “Mario Negri”, via Giuseppe La Masa 19, 20156 Milan, Italy. Tel.: þ39 0239014527; fax: þ39 0233200231. E-mail address: [email protected] (C. La Vecchia). 1047-2797/$ e see front matter Ó 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.annepidem.2013.11.003
bladder carcinogen, in eight of the 11 hair dyes tested [6]. Darker colors are formed by using higher concentrations of primary intermediates and have been therefore suggested to pose a higher risk of urinary bladder cancer than nondark-shade hair dyes [2], although the issue remains open to discussion. The International Agency for Research on Cancer (IARC) recently concluded that occupational exposures to hair dyes as a hairdresser or barber were probably carcinogenic to humans, whereas personal use of hair colorants was not classifiable as to its carcinogenicity to humans [1]. When referring to bladder cancer, the IARC Working Group considered limited the available evidence of an increased risk among hairdressers (i.e., occupational exposure) and inadequate the evidence on the role of personal use of hair dyes. The difference between personal and occupational exposure reflects different doses of exposure, but also the occupational studies are subject to a different series of socioeconomic and lifestyle-related confounding factors, including tobacco. With regards to personal use of dyes, although an excess bladder cancer risk has been seen in some studies [7,8], the overall results are not converging [9,10]. An earlier quantitative review found a relative risk (RR) of bladder cancer for ever versus never users between 1.0 and 1.1, based on various assumptions [11]. The most recent meta-analysis, published in 2008 and including 11 casecontrol studies and one cohort study, found no association
152
F. Turati et al. / Annals of Epidemiology 24 (2014) 151e159
between personal hair dye use and bladder cancer risk [12]. Thereafter, three other case-control [13e15] and one cohort [16] studies have been published, including the New England bladder cancer study, with almost 1200 cases [14]. We, therefore, updated the summary evidence on personal use of hair dyes and bladder cancer till January 2013, addressing some critical issues in exposure assessment, including the type of hair-coloring product, color, duration of use, and lifetime frequency of use. Materials and methods Systematic search The meta-analysis was conducted following the Meta-analysis Of Observational Studies in Epidemiology guidelines [17]. We performed a computerized MEDLINE search using PubMed from 1966 to January 2013 to identify potentially eligible studies, using the string “(hair) AND (dye* OR colour OR color) AND (neoplasm* OR cancer OR cancers OR tumour*) AND (urinary OR bladder).” A similar strategy was used to search EMBASE (1980 to January 2013). We also checked the reference lists of every article retrieved and those of some relevant systematic reviews [11,18,19] and metaanalyses [9,10,12], including the IARC Monograph focused on aromatic amines and organic dyes, published in 2010 [1]. Articles not in English, unpublished studies, conference abstracts, and posters were not considered. Two authors (F.T. and C.G.) screened the search results independently against the inclusion criteria to identify eligible articles, and discrepancies were discussed and resolved. Inclusion criteria and data collection The articles were included if (1) they presented original data from case-control or cohort studies; (2) the outcome of interest was incidence or mortality from bladder cancer; (3) they analyzed exposure to personal hair dyes; and (4) they provided RR estimates of the association of interest and the corresponding 95% confidence intervals (CIs) or sufficient information for their calculation (e.g., the distribution of cases and noncases across exposure categories or the P value for the test on the RR). Studies on exposure to hair dyes in occupational settings were not considered. If data from the same study were published in multiple articles [8,20e24], we considered the article presenting results based on a larger number of cases. We abstracted the following information from the included studies in a standard format: country, study design, sex of the population, duration of follow-up (for cohort studies) or calendar years of participants inclusion (for case-control studies), number of participants (cases and controls or cohort size), variables adjusted for in the analysis and/or matching variables, and measures of associations that compared personal use of hair dyes (any type of dyes and permanent dyes, when available) with no use of any type of dyes. We also extracted measures for duration of use and lifetime frequency of use, where available. When multiple results for any measure were presented in a single study, we extracted the one adjusted for the largest number of confounding factors. Statistical analysis The first exposure of interest was personal use of any type of hair dyes, as defined in each study. A study reporting quantitative information on permanent dyes only (and not on any type of dyes) was considered in the analysis of any type of dyes as well [20]. The second exposure of interest was personal use of permanent hair dyes [7,8,14,15,20,25]. One study providing data for exclusive use of permanent hair dyes was included [26].
For both exposures of interest, we estimated the pooled RR of bladder cancer for (1) any use compared with no use, (2) highest duration of use (in years) compared with no use, and (3) highest lifetime frequency of use (total number of times) compared with no use. In these analyses, when a study reported risk estimates for males and females only separately, we included both sex-specific RR estimates. For the analysis on duration of use and lifetime frequency of use, we extracted data for the highest categories of exposure in each study compared with no use. In particular, for duration, the highest categories ranged from 6 or more years [27] to 33 or more years of use [26]. Two subgroups of studies were considered as follows: studies with threshold for the highest category of duration of 20 years or longer and those with threshold for the highest category of duration starting from less than 20 years. For lifetime frequency of use, the highest categories ranged from more than 40 times [15] to more than 504 times [26]. Two subgroups of studies were considered as follows: studies with threshold for the highest lifetime frequency of use of 200 times or more and those with threshold for the highest lifetime frequency of use starting from less than 200 times. For both exposures of interest and for each comparison, we conducted subgroup analyses according to selected factors, such as study design, sex, geographic area, and dark-colored dyes (black dyes when available), to explore potential sources of heterogeneity. We also conducted a sensitivity analysis excluding each study in turn to evaluate the robustness of our results. We used random-effects models to combine the study-specific RRs [28]. Statistical heterogeneity among studies was assessed using the c2 test (results were defined as heterogenous for a value of P < .10) [29], and the potential inconsistency was quantified through the I2 statistic, which describes the percentage of total variation across studies that is due to heterogeneity rather than chance [30]. Publication bias was evaluated through funnel plots [31] and with the Egger [32] and Begg tests [33]. All the statistical analyses were performed using the STATA software (version 11; StataCorp, College Station, TX). Results The flow chart of the procedure to select articles is available in the Appendix (Appendix Figure 1). The PubMed and EMBASE searches identified 87 and 192 publications, respectively. After the exclusion of non relevant studies and duplicate publications, we selected 19 potentially eligible unique publications. Other six pertinent publications were obtained from the review of the reference lists of the articles retrieved. Among the 25 unique full-text articles assessed for eligibility, eight publications not satisfying the inclusion criteria were excluded. In particular, one study was excluded because it had a cross-sectional study design [34] and another one because it investigated the association between bladder cancer risk and use of brilliantine [35]. On the other hand, we included one study on cancer of the lower urinary tract in which 90% of the case series had urinary bladder cancer [27], as well as one case-control study considering together bladder (23 cases) and kidney (12 cases) cancers [36]. Finally, we identified 15 case-control [7,8,13e15,25e27,36e42] and two cohort [16,20] studies on personal hair dye use and bladder cancer, which satisfied the inclusion criteria, for a total of 8504 cases/deaths from bladder cancer (8168 cases and 336 deaths) (Table 1). To improve transparency and replicability of our results, we reported in the column “Notes” in Table 1 all the coding decisions we made on each original article (e.g., RRs pooling within studies) [17].
Table 1 Characteristics of the studies included in the meta-analysis Study design
Sex
Period of enrollment/duration of follow-up
Cases
Controls/cohort size
Adjusting and/or matching factors
Notes
Dunham, 1968, USA (Louisiana)
Case-control, HB
M*
1958e1964
132 M (493*)
136 M (527*)
None
Jain, 1977, Canada
Case-control
Both
NA
107
107
Neutel, 1978, Canada
Case-control
Fy
NA
50
50
Crude OR and corresponding 95% CI calculated from the distribution of cases and controls according to use of personal hair dyes All male controls had benign prostatic hypertrophy and all female controls had stress incontinence Crude OR and corresponding 95% CI calculated from the distribution of cases and controls according to use of personal hair dyes
Howe, 1980, Canada Stavraki, 1981, Canada
Case-control, PB Case-control, PB
Fz F
1974e1976 1976
Hartage, 1982, USA US National Bladder Cancer Study Ohno, 1985, Japan
Case-control, PB
Both
Case-control, PB
Nomura, 1989, USA (Hawaiik)
Sex, age
Age
152 (632z) 70
Sex, age Nonspecified
1977e1978
152 (632z) 23 bladder and 12 kidney cancers 2982
5782
Sex, age, area, race, smoking
Fx
1976e1978
66 (293x)
146 (589x)
Age, area, smoking
Case-control, PB
Both
1977e1986
261{
522
Andrew, 1994, USA
Case-control, PB
Both
459
665
Gago-Dominguez, 2001, USA Los Angeles Study
Case-control, PB
Both
Cases: 1994e1998 Controls: 1993e1997 1987e1996
897 (1514#)
897 (1514#)
Sex, age, ethnicity, residence, smoking
Henley and Thun, 2001, USA
Cohort
F
16 y of follow-up from 1982
336 deaths from bladder cancer
547,571 PR
Age, race, smoking, education, blue collar occupation
Kogevinas, 2006, Spain
Case-control, HB
F
1998e2001
152
166
Lin, 2006, USA
Case-control, HB
Both
Case-control: 1999 (ongoing) Case-cohort: 2001 (ongoing)
712
712
Mendelsohn, 2009, China (Shanghai) Shanghai Women Health Study
Cohort
F
1996e2005 7 y of follow-up (average)
32
70,366 PR
The OR for ever use of personal hair dyes was calculated pooling the OR for frequency of use of
