Secondhand Smoke Exposure During Childhood and Cancer Mortality in Adulthood Among Never Smokers: The Japan Collaborative Cohort Study for Evaluation of Cancer Risk

Abstract

We examined whether secondhand smoke exposure during childhood was associated with cancer mortality in adulthood among never smokers. In the Japan Collaborative Cohort Study for Evaluation of Cancer Risk, we analyzed data from 45,722 Japanese lifetime nonsmokers aged 40–79 years with no history of cancer at baseline (1988–1990) who had completed a lifestyle questionnaire, including information on the number of family members who had smoked at home during their childhood (0, 1, 2, or ≥3 family members). A Cox proportional hazards model and competing-risks regression were used to calculate multivariable hazard ratios and subdistribution hazard ratios with 95% confidence intervals for overall and site-specific cancer mortality according to the number of family members who smoked during the participant’s childhood, after adjusting for potentially confounding factors. During a median follow-up period of 19.2 years, a total of 2,356 cancer deaths were documented. Secondhand smoke exposure was positively associated with the risk of mortality from pancreatic cancer in adulthood; the multivariable hazard ratio for having 3 or more family members who smoked (as compared with none) was 2.32 (95% confidence interval: 1.14, 4.72). Associations were not evident for total cancer risk or risk of other types of smoking-related cancer. In this study, secondhand smoke exposure during childhood was associated with an increased risk of pancreatic cancer mortality in adulthood.

cancer mortality, parental smoking, passive smoking, secondhand smoke

Abbreviations

 
• CI

  confidence interval

 
• HR

  hazard ratio

 
• ICD-10

  International Classification of Diseases, Tenth Revision

 
• JACC

  Japan Collaborative Cohort Study for Evaluation of Cancer Risk

 
• SHR

  subdistribution hazard ratio

Active and passive smoking in adulthood have been associated with increased risks of cancer at various sites. According to a 2014 US Surgeon General’s report, smoking is causally linked to cancers of the oropharynx, larynx, esophagus, lung, stomach, liver, pancreas, kidney, cervix, bladder, and colorectum, as well as acute myeloid leukemia (1). The risks of these smoking-related cancers in adulthood could be increased by early exposure to active smoking, because younger smokers may be more susceptible to physiological changes caused by smoking, such as DNA adduct formation and its longer persistence, although evidence on secondhand smoke is not available (2).

Several prospective cohort studies conducted in Europe and the United States have reported on the association between secondhand smoke exposure during childhood and the risk of cancer in adulthood. The European Prospective Investigation Into Cancer and Nutrition showed that environmental tobacco smoke exposure during childhood could be an independent risk factor for an increased incidence of lung (3), bladder (4), and pancreatic (5) cancer in adulthood. In the Nurses’ Health Study, parental smoking when study participants were living with their parents was associated with an increased risk of pancreatic cancer in adulthood among women, while no association was observed with adult secondhand smoke at home or at work (6).

For children, living in households with smokers is the biggest risk factor for secondhand smoke exposure (7). Parental smoking, low socioeconomic status, low educational level, and less negative attitudes toward secondhand smoke have been associated with a child’s exposure to secondhand smoke at home (8). Therefore, building scientific evidence on the adverse effects of familial smoking at home during childhood may lead to protection of children from future risks of cancer.

However, previous studies have included only Western cohorts. Among the Asian population, little is known about the impact of secondhand smoke exposure during childhood on cancer mortality in adulthood, although the proportion of smoking in adulthood is generally high among men in Asian countries (9). Considering limited and conflicting evidence regarding the effect of secondhand smoke during childhood on the future risk of cancer, we aimed to examine the association between secondhand smoke exposure at home during childhood and adulthood risk of mortality from total cancer and major types of smoking-related cancer using data from a large long-term cohort study on Japanese men and women.

METHODS

Study population and baseline questionnaire

The Japan Collaborative Cohort Study for Evaluation of Cancer Risk (JACC) is a nationwide community-based prospective study that started between 1988 and 1990, enrolling 110,585 subjects (46,395 men and 64,190 women) aged 40–79 years who were living in 45 communities across Japan. The methodology of the study has been described elsewhere (10).

Briefly, participants were asked to complete self-administered questionnaires, including questions on demographic characteristics, medical history, and lifestyle. Before completing the questionnaire, the participants or community representatives provided informed consent to be involved in this epidemiologic study, according to the guidelines of the Council for International Organizations of Medical Sciences (Geneva, Switzerland) (11). Informed consent was obtained from each participant in 36 of the 45 study areas (written consent in 35 areas and oral consent in 1 area). In the remaining 9 areas, group consent was obtained from the leader of each area. The JACC protocol was approved by the ethics committees of Hokkaido University (Sapporo, Japan), Nagoya University (Nagoya, Japan), and Osaka University (Osaka, Japan).

Secondhand smoke exposure during childhood

We asked the participants, “Did your family member smoke at home when you were an elementary or junior high school student?”, with 3 possible responses: “yes,” “no,” or “cannot remember.” The age of elementary or junior high school students in Japan is between 6 and 15 years. For those who responded yes, we further asked, “Which family members smoked at home?” (possible responses: “mother,” “father,” “brother or sister,” “grandfather or grandmother,” or “other”). According to their responses to these questions, we calculated the number of family members who smoked as 0, 1, 2, or ≥3.

Of the 110,585 participants, we excluded 50,101 participants (37,368 men and 12,733 women) who had ever smoked or did not provide information on their smoking status. A total of 8,725 participants (1,514 men and 7,211 women) living in 8 communities were excluded because the questions on secondhand smoke exposure during childhood were not included in the questionnaire. We excluded 5,403 participants (763 men and 4,640 women) because their answer to the first question was “cannot remember” (271 men and 1,881 women) or they did not provide their history of secondhand smoke exposure during childhood (492 men and 2,759 women). Furthermore, 634 participants (29 men and 605 women) were excluded because of a history of cancer at the time of baseline inquiry. A total of 45,722 participants (6,721 men and 39,001 women) were included in the analyses.

Assessment of confounding variables

Other information on demographic and lifestyle factors—namely age, sex, height, weight, past medical history (e.g., diabetes and hypertension), alcohol drinking status, exercise and walking habits, mental status, educational level, occupation, and eating habits—was derived from a self-administered questionnaire at baseline. Body mass index was calculated as body weight (kg) divided by the square of height (m²).

Mortality surveillance

In each community, a systematic review of death certificates, all of which were forwarded to the local public health center, was conducted. It is believed that all cohort deaths were recorded, except those for participants who died after moving from their original community, in which case the participants’ data were censored. The date of moving from the community was verified using population-registration documents. Mortality data were sent centrally to the Japan Ministry of Health and Welfare, and the underlying cause of death was coded for the National Vital Statistics System according to the International Classification of Diseases, Tenth Revision (ICD-10). In 35 communities, follow-up was conducted until the end of 2009. However, the study was terminated in 4 areas by the end of 1999, in 4 other areas by the end of 2003, and in 2 areas by the end of 2008. The median duration of follow-up was 19.2 years. With reference to the 2014 Surgeon General’s report (1), cause-specific mortality from smoking-related cancers, by cancer site, was defined using ICD-10 codes: lip, oral cavity, pharynx, and larynx (n = 34; ICD-10 codes C000–C149 and C320–C329), nose and nasal cavity (n = 6; ICD-10 codes C300–C301 and C310–C319), esophagus (n = 28; ICD-10 codes C150–C159), stomach (n = 422; ICD-10 codes C160–C169), colorectum (n = 233; ICD-10 codes C180–C189, C199, and C209), liver (n = 232; ICD-10 codes C220–C229), pancreas (n = 232; ICD-10 codes C250–C259), lung (n = 107; ICD-10 codes C340–C349), cervix (n = 28; ICD-10 codes C530–C539), bladder (n = 37; ICD-10 codes C670–C679), or kidney (n = 4; ICD-10 codes C649–C659), or acute myeloid leukemia (n = 27; ICD-10 code C920). If participants died after they had moved from their original community, they were treated as withdrawals from observation when they moved out.

Statistical analysis

Person-years of follow-up were calculated from the date of the baseline questionnaire to the date of death or emigration from the community or the end of 2009 (or 1999, 2003, or 2008), whichever occurred first. Age-adjusted mean values and prevalences of risk factors for cancer were calculated in categories of the number of family members who smoked at home during childhood. Generalized linear models were used to perform trend tests across groups. Sex- and age-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for cancer mortality according to the number of familial smokers in the home during childhood were calculated using a Cox proportional hazards model after adjustment for potentially confounding factors. Competing-risks regression models were employed to estimate subdistribution hazard ratios (SHRs) and 95% CIs, with other causes of death as competing-risk events to minimize the potential impacts of competing-risk bias on the association between childhood secondhand smoke exposure and smoking-related cancer mortality (12). The potentially confounding factors were age (years; continuous), sex (female or male), body mass index (in sex-specific quintiles), secondhand smoke exposure in adulthood at home (never, sometimes, 1–4 days per week, or almost every day) and outside of the home (never, sometimes, 1–4 days per week, or almost every day), alcohol consumption (never drinker, ex-drinker, or current drinker of 0.1–45.9 or ≥46.0 g of ethanol per day), amount of exercise (almost never, 1–4 hours/week, or ≥5 hours/week), amount of walking (almost never, 0.5 hours/day, or >0.5 hours/day), perceived level of mental stress (low, moderate, or high), educational level (age ≤18 years or age ≥19 years upon completion of education), employment status (unemployed or employed), and frequency of vegetable and fruit intake (in quintiles). Furthermore, additional potential confounders were selected for specific cancers: a history of diabetes (yes or no) for pancreatic cancer and red and processed meat intake (in quintiles) for stomach and colorectal cancer.

To eliminate confounding due to adult secondhand smoke exposure, we performed sensitivity analyses by excluding all participants who reported adult secondhand smoke exposure at home or outside of the home.

SAS software, version 9.4 (SAS Institute, Inc., Cary, North Carolina), was used for all statistical analyses. All statistical tests were 2-sided, and P values less than 0.05 were considered statistically significant.

RESULTS

Age-adjusted baseline characteristics of the participants according to the number of family members who smoked during childhood are presented in Table 1. The proportions for the number of family members who smoked were 27.4% for no family members, 56.3% for 1 family member, 14.3% for 2 family members, and 2.0% for 3 or more family members. Most participants (85.3%) were female. Compared with those who had no family members who smoked during childhood, those who reported having 3 or more familial smokers during childhood were more likely to have higher proportions of adult secondhand smoke exposure at home and outside of the home, a higher body mass index, a history of diabetes, and a higher level of mental stress; were less educated; and less often ate vegetables. They were also more likely to be current alcohol drinkers and unemployed.

Table 1

Age-Adjusted Baseline Characteristics of Participants and Adulthood Cancer Incidence According to the Number of Family Members Who Smoked in the Home During Childhood, Japan, 1988–1990^a

	No. of Family Members Who Smoked
	0  (n = 12,539)		1  (n = 25,743)		2  (n = 6,546)		≥3  (n = 894)
Characteristic	No.	%	No.	%	No.	%	No.	%	P for   Trend
Mean age, years	58.0		56.6		56.5		58.4		0.23
Male sex	2,033	16.2	3,792	14.7	797	12.2	99	11.1	<0.001
Mean body mass indexb	23.0		23.0		23.1		23.3		0.005
Secondhand smoke exposure ≥1 day/week in adulthood
At home	3,649	42.3	9,907	52.9	2,669	57.0	395	62.3	<0.001
Outside of the home	1,675	18.0	3,841	19.4	1,059	21.1	174	27.9	<0.001
Current alcohol drinker	2,422	21.8	5,199	22.2	1,403	23.6	178	22.4	0.49
High level of mental stress	2,073	20.2	4,582	20.9	1,332	22.0	182	22.6	0.05
College education or higher	1,266	11.8	2,653	11.6	672	11.6	68	9.2	0.03
Unemployed	2,084	17.8	3,876	18.5	1,026	18.8	158	17.3	0.71
Walking >30 minutes/day	7,652	71.8	15,503	71.3	4,064	71.7	526	68.8	0.09
Exercise ≥1 hour/week	2,883	25.0	5,735	24.8	1,521	24.8	199	23.3	0.28
History of diabetes mellitus	456	3.7	940	4.0	260	4.4	41	4.6	0.17
Mean vegetable intake, no. of times/week	15.5		15.2		14.9		14.6		<0.001
Mean fruit intake, no. of times/week	8.1		8.3		8.3		8.2		0.39
Cancer incidence, no. and % of cases
All cancers	674	5.4	1,323	5.1	304	4.6	55	6.2
Smoking-related cancersc	417	3.3	857	3.3	193	2.9	34	3.8
Lip, oral cavity, pharyngeal, and laryngeal cancer	10	0.1	20	0.1	4	0.1	0	0.0
Nose and nasal cavity cancer	0	0.0	3	0.0	2	0.0	1	0.1
Esophageal cancer	3	0.0	20	0.1	4	0.1	1	0.1
Stomach cancer	125	1.0	244	0.9	43	0.7	10	1.1
Colorectal cancer	63	0.5	138	0.5	28	0.4	4	0.4
Liver cancer	67	0.5	124	0.5	37	0.6	4	0.4
Pancreatic cancer	56	0.4	139	0.5	28	0.4	9	1.0
Lung cancer	76	0.6	132	0.5	35	0.5	3	0.3
Kidney cancer	2	0.0	2	0.0	0	0.0	0	0.0
Bladder cancer	8	0.1	20	0.1	8	0.1	1	0.1
Acute myeloid leukemia	7	0.1	15	0.1	4	0.1	1	0.1
Cervical cancer	9	0.1	16	0.1	2	0.0	1	0.1

Abbreviation: ICD-10, International Classification of Diseases, Tenth Revision.

^a Data are age-adjusted mean values for continuous variables and numbers and age-adjusted percentages for categorical variables.

^b Weight (kg)/height (m)².

^c Smoking-related cancers included cancer of the lip, oral cavity, pharynx, and larynx (ICD-10 codes C000–C149 and C320–C329), nose and nasal cavity (ICD-10 codes C300–C301 and C310–C319), esophagus (ICD-10 codes C150–C159), stomach (ICD-10 codes C160–C169), colorectum (ICD-10 codes C180–C189, C199, and C209), liver (ICD-10 codes C220–C229), pancreas (ICD-10 codes C250–C259), lung (ICD-10 codes C340–C349), cervix (ICD-10 codes C530–C539), bladder (ICD-10 codes C670–C679), and kidney (ICD-10 codes C649–C659) and acute myeloid leukemia (ICD-10 code C920).

During the median follow-up period of 19.2 years, a total of 2,356 (478 men and 1,878 women) deaths from cancer were documented. Table 2 shows the multivariable HRs and SHRs for cancer mortality according to the number of family members who smoked during childhood. The results shown in Table 2 are limited to cancer sites with at least 100 observed cases. In age- and sex-adjusted analyses, no association was observed between secondhand smoke exposure during childhood and the risks of mortality from all cancers or smoking-related cancers in adulthood, except for pancreatic cancer. The excess risk of mortality from pancreatic cancer associated with having 3 or more familial smokers was statistically significant. After adjustment for additional potentially confounding factors, the association did not change; the multivariable HR for pancreatic cancer mortality was 2.32 (95% CI: 1.14, 4.72) among persons with ≥3 family members who smoked. Such an association was not evident for other cancer sites. In the competing-risk regression models, a similar association with pancreatic cancer mortality was observed; the multivariable SHR for pancreatic cancer mortality was 2.25 (95% CI: 1.10, 4.60) among those with ≥3 familial smokers, compared with no family members who smoked.

Table 2

Hazard Ratios for Adult Cancer Mortality According to the Number of Family Members Who Smoked in the Home During Childhood, Japan, 1988–1990

		Risk of Incident Cancer in Adulthood
Type of Cancer and No. of Family Members Who Smoked	No. of   Cases	HRa	95% CI	Multivariable HRb	95% CI	Multivariable SHRb	95% CI
All cancers
0	674	1.00	Referent	1.00	Referent	1.00	Referent
1	1,323	1.03	0.94, 1.13	1.03	0.94, 1.13	1.04	0.94, 1.14
2	304	0.94	0.82, 1.08	0.94	0.82, 1.08	0.95	0.83, 1.10
≥3	55	1.20	0.91, 1.59	1.21	0.92, 1.59	1.18	0.89, 1.56
P for trend		0.93		0.85		0.85
≥1 vs. 0		1.02	0.93, 1.11	1.03	0.94, 1.12	1.03	0.94, 1.13
Smoking-related cancersc
0	417	1.00	Referent	1.00	Referent	1.00	Referent
1	857	1.09	0.97, 1.23	1.10	0.97, 1.23	1.10	0.97, 1.23
2	193	0.98	0.83, 1.17	0.98	0.82, 1.17	0.99	0.83, 1.18
≥3	34	1.20	0.84, 1.70	1.20	0.84, 1.70	1.15	0.81, 1.64
P for trend		0.56		0.59		0.60
≥1 vs. 0		1.07	0.96, 1.20	1.08	0.96, 1.21	1.08	0.96, 1.21
Non–smoking-related cancersd
0	257	1.00	Referent	1.00	Referent	1.00	Referent
1	466	0.93	0.80, 1.08	0.95	0.81, 1.11	0.94	0.81, 1.10
2	111	0.88	0.70, 1.10	0.89	0.71, 1.12	0.90	0.72, 1.13
≥3	21	1.21	0.78, 1.89	1.24	0.79, 1.94	1.20	0.77, 1.88
P for trend		0.52		0.67		0.66
≥1 vs. 0		0.93	0.80, 1.07	0.95	0.81, 1.10	0.94	0.81, 1.09
Pancreatic cancer
0	56	1.00	Referent	1.00	Referent	1.00	Referent
1	139	1.31	0.96, 1.79	1.30e	0.95, 1.77	1.33e	0.97, 1.83
2	28	1.04	0.66, 1.64	1.04e	0.66, 1.64	1.06e	0.66, 1.68
≥3	9	2.30	1.14, 4.65	2.32e	1.14, 4.72	2.25e	1.10, 4.60
P for trend		0.14		0.13		0.14
≥1 vs. 0		1.29	0.95, 1.74	1.32	0.97, 1.79	1.31	0.96, 1.78
Liver cancer
0	67	1.00	Referent	1.00	Referent	1.00	Referent
1	124	0.98	0.72, 1.31	1.00	0.74, 1.36	1.00	0.73, 1.36
2	41	1.14	0.77, 1.68	1.12f	0.75, 1.66	1.11f	0.74, 1.66
≥3
P for trend		0.60		0.63		0.66
≥1 vs. 0		1.01	0.76, 1.34	1.03	0.77, 1.38	1.02	0.76, 1.38
Lung cancer
0	76	1.00	Referent	1.00	Referent	1.00	Referent
1	132	0.91	0.68, 1.20	0.91	0.68, 1.21	0.90	0.67, 1.20
2	38	0.91f	0.62, 1.34	0.92f	0.62, 1.37	0.92f	0.62, 1.37
≥3
P for trend		0.56		0.60		0.60
≥1 vs. 0		0.91	0.69, 1.19	0.91	0.69, 1.20	0.90	0.68, 1.19
Stomach cancer
0	125	1.00	Referent	1.00	Referent	1.00	Referent
1	244	1.05	0.85, 1.31	1.03g	0.83, 1.29	1.02g	0.82, 1.28
2	43	0.75	0.53, 1.06	0.74g	0.52, 1.05	0.74g	0.52, 1.05
≥3	10	1.20	0.63, 2.29	1.15g	0.60, 2.20	1.09g	0.57, 2.09
P for trend		0.49		0.40		0.36
≥1 vs. 0		1.00	0.81, 1.23	0.98	0.79, 1.22	0.97	0.79, 1.21
Colorectal cancer
0	63	1.00	Referent	1.00	Referent	1.00	Referent
1	138	1.16	0.86, 1.56	1.15g	0.85, 1.56	1.14g	0.84, 1.54
2	32	0.92	0.60, 1.41	0.93f,g	0.60, 1.43	0.93g	0.60, 1.44
≥3
P for trend		0.95		0.94		0.94
≥1 vs. 0		1.11	0.83, 1.48	1.10	0.82, 1.48	1.10	0.82, 1.47

Abbreviations: CI, confidence interval; HR, hazard ratio; ICD-10, International Classification of Diseases, Tenth Revision; SHR, subdistribution hazard ratio.

^a Adjusted for age and sex.

^b Results were adjusted for age (years; continuous), sex, body mass index (weight (kg)/height (m)²; sex-specific quintiles), secondhand smoke exposure in adulthood at home (never, sometimes, 1–4 days per week, or almost every day) and outside of home (never, sometimes, 1–4 days per week, or almost every day), alcohol consumption (never drinker, ex-drinker, or current drinker of 0.1–45.9 or ≥46.0 g of ethanol per day), amount of exercise (almost never, 1–4 hours/week, or ≥5 hours/week), amount of walking (almost never, 0.5 hours/day, or >0.5 hours/day), perceived level of mental stress (low, moderate, or high), educational level (age ≤18 years or age ≥19 years upon completion of education), employment status (unemployed or employed), and frequency of vegetable and fruit intake (quintiles).

^c Smoking-related cancers included cancer of the lip, oral cavity, pharynx, and larynx (ICD-10 codes C000–C149 and C320–C329), nose and nasal cavity (ICD-10 codes C300–C301 and C310–C319), esophagus (ICD-10 codes C150–C159), stomach (ICD-10 codes C160–C169), colorectum (ICD-10 codes C180–C189, C199, and C209), liver (ICD-10 codes C220–C229), pancreas (ICD-10 codes C250–C259), lung (ICD-10 codes C340–C349), cervix (ICD-10 codes C530–C539), bladder (ICD-10 codes C670–C679), and kidney (ICD-10 codes C649–C659) and acute myeloid leukemia (ICD-10 code C920).

^d Non–smoking-related cancers included cancer at all sites not listed above.

^e HR and SHR were further adjusted for self-reported diabetes status.

^f Multivariable HR and SHR for ≥2 family members who smoked versus 0. We collapsed the categories of 2 and ≥3 familial smokers when fewer than 5 participants were included in the ≥3 familial smokers category.

^g HR and SHR were further adjusted for red meat and processed meat intake.

Because there were few cases of liver, lung, and colorectal cancer among participants with ≥3 familial smokers, we combined the categories of 2 and ≥3 familial smokers for those analyses. None of the multivariable HRs or SHRs reached statistical significance.

Compared with no familial smokers, having any family member who smoked (≥1 familial smokers) during childhood was associated with an increased risk of mortality from pancreatic cancer, although the association did not reach statistical significance.

In the sensitivity analyses, exclusion of all participants who reported adult secondhand smoke exposure at home or outside of the home did not change the results materially (see Web Table 1, available at https://doi.org/10.1093/aje/kwab284).

DISCUSSION

In this large prospective study of Japanese lifetime nonsmokers aged 40–79 years with a median follow-up of 19.2 years, exposure to smoking by 3 or more family members during childhood was associated with approximately double the risk of mortality from pancreatic cancer in adulthood, compared with having no familial smokers in childhood. The result did not change materially after the exclusion of participants with adult secondhand smoke exposure. In the JACC, Lin et al. (13) previously reported that the presence of a smoker among family members in childhood was not associated with an increased risk of pancreatic cancer mortality in adulthood (relative risk = 1.21, 95% CI: 0.87, 1.68) among female lifetime nonsmokers. However, they adjusted only for age, body mass index, and history of diabetes, not for passive smoking status in adulthood. We conducted more comprehensive analyses, adjusting for multiple potentially confounding factors, and examined associations with the number of family members who smoked during childhood among male and female lifetime nonsmokers. To the best of our knowledge, this is the first study to report a positive association between secondhand smoke exposure, estimated by the number of family members who smoked during childhood, and adulthood risk of pancreatic cancer in an Asian population. We did not observe any significant associations for the risk of total cancer mortality or mortality from other smoking-related cancers. A potential explanation for the lack of association could be the relatively small number of cases among these mostly lifetime-nonsmoking women.

Our suggestive findings on the effects of childhood secondhand smoke exposure on the future risk of pancreatic cancer are consistent with findings from previous studies carried out in Western countries. In the European Prospective Investigation Into Cancer and Nutrition, a cohort study of 112,430 European lifetime nonsmokers, daily exposure to secondhand smoke during childhood was associated with an increased risk of pancreatic cancer in adulthood, compared with no exposure; the multivariable HR was 2.09 (95% CI: 1.14, 3.84) (14). However, such an association was not evident for the risks of total cancer and other types of smoking-related cancer. In the Nurses’ Health Study (n = 86,673 women), a history of parental smoking exposure significantly increased the risk of pancreatic cancer incidence in adulthood; the multivariable relative risk of developing pancreatic cancer among persons exposed to parental smoking was 1.41 (95% CI: 1.03, 1.91), compared with nonexposure (6).

An underlying mechanism for the association between childhood secondhand smoke exposure and risk of adulthood pancreatic cancer could be a chronic pancreatic inflammatory process. Fibrosis induced by secondhand smoke exposure was found in an animal study (15). Another potential mechanism of pancreatic carcinogenesis could be the formation of DNA adducts caused by carcinogens in cigarette smoke, such as nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone and its major metabolite, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (16). A previous animal experiment showed that the number of DNA adducts in the pancreas increased when rats were given water containing 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone and 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (17). Once DNA adducts are formed, they can be removed by DNA repair enzymes in the body, and the DNA returns to normal. However, if the DNA is seriously damaged and repair is insufficient, DNA adducts can persist and result in miscoding during DNA replication and the development of cancer (18).

The strength of the present study was its prospective design, which eliminated recall bias in exposure assessment. Furthermore, since the prevalence of smoking in Japan was relatively high during the participants’ childhood, we had a sufficient number of family members who smoked (n = 894 for ≥3 familial smokers) to find an association.

However, this study had several limitations. First, misclassification due to the recall of secondhand smoke exposure during childhood could be a potential problem. However, the proportion of participants who reported having 1 or more family members who smoked at home during their childhood was approximately 70%. This result was almost perfectly consistent with the prevalence of smoking in Japan during the 1950s and 1960s (19). Additionally, a previous study examined the validity of offspring-reported parental smoking status during the offspring’s childhood (20). The result showed that adults reported their parents’ smoking status during their childhood quite accurately; the sensitivity of offspring’s correct reporting of parental smoking status during childhood (age ≤10 years) was 0.82 (95% CI: 0.81, 0.84), and specificity was 0.95 (95% CI: 0.95, 0.96). Therefore, the effect of misclassification in the present study may have been limited. Second, we used mortality data, not incidence data, which may lead to misclassification in the diagnosis of cancer at different sites. However, the widespread use of computed tomography scanning or magnetic resonance imaging in local Japanese hospitals since the 1980s has probably made death certificate diagnoses of cancer sufficiently accurate (21).

Third, we used self-reported number of family members who smoked during childhood as a proxy for exposure to secondhand smoke in childhood; therefore, the duration and intensity of secondhand smoke exposure in childhood were not measured, resulting in nondifferential misclassification. Fourth, these findings could have been due to chance, since they were based on a small number of cases and are counterintuitive with regard to a previous report on a rapid decrease in pancreatic cancer risk among adult former smokers (5). Finally, despite our efforts to adjust for potentially confounding factors, we cannot rule out the effects of unmeasured factors or residual confounding, such as prenatal smoke exposure, behavioral change over time, and quantity of individual and passive smoking in adulthood.

In conclusion, secondhand smoke exposure during childhood could be an independent risk factor for adulthood pancreatic cancer regardless of conventional risk factors and adult lifestyle, while no association was evident for the risk of total cancer mortality or mortality from other types of smoking-related cancer. As previously reported, interventions designed to achieve smoking cessation among parents for the sake of their children’s health may be recommended (22). The present study reinforces the importance of such interventions, because smoking cessation among parents not only benefits their health but also benefits their children in the long run.

ACKNOWLEDGMENTS

Author affiliations: Public Health Unit, Department of Social Medicine, Graduate School of Medicine, Osaka University, Osaka, Japan (Masayuki Teramoto, Hiroyasu Iso); Department of Preventive Medicine, Graduate School of Medicine, Nagoya University, Nagoya, Japan (Kenji Wakai); and Department of Preventive Medicine, Graduate School of Medicine, Hokkaido University, Sapporo, Japan (Akiko Tamakoshi).

This study was supported by Grants-in-Aid for Scientific Research (KAKENHI) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT; Monbu-kagaku-shō) of Japan; Grants-in-Aid for Scientific Research on Priority Areas of Cancer Epidemiology from MEXT (KAKENHI grants 61010076, 62010074, 63010074, 1010068, 2151065, 3151064, 4151063, 5151069, 6279102, 11181101, 17015022, 18014011, 20014026, 20390156, and 26293138); and Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research (KAKENHI grant 16H06277). This research was also supported by Grants-in-Aid from the Ministry of Health, Labour and Welfare (Health and Labor Sciences Research Grants; “Comprehensive Research on Cardiovascular Disease and Life-Style Related Diseases” grants H20-Junkankitou [Seishuu]-Ippan-013 and H23-Junkankitou [Seishuu]-Ippan-005); the Intramural Research Fund for Cardiovascular Diseases (grant 22-4-5) of the National Cerebral and Cardiovascular Center; and “Comprehensive Research on Cardiovascular Disease and Life-Style Related Diseases” grants H26-Junkankitou [Seisaku]-Ippan-001, H29-Junkankitou [Seishuu]-Ippan-003, and 20FA1002.

We thank all staff members involved in this study for their valuable help in conducting the baseline survey and follow-up.

Conflict of interest: none declared.