Abstract

Background

Smoking should be particularly discouraged among survivors of childhood cancer, who are at increased risk of adverse effects of the cancer and its treatment. We examined the extent of cigarette smoking, factors associated with being a current smoker, and age at initiation of regular smoking among adult survivors of childhood cancer and compared the survivors’ smoking habits with those of the general population.

Methods

We used data from the British Childhood Cancer Survivor Study (BCCSS), a population-based cohort of 17 981 individuals who were diagnosed with childhood cancer between 1940 and 1991 in Britain and had survived for at least 5 years after diagnosis. The 14 836 cohort members who were alive and aged 16 years or older up to September 2006 were eligible to receive a mailed questionnaire that ascertained smoking status and other factors. The general population data were from the 2002 General Household Survey in Britain. Logistic regression was used to investigate factors associated with being a current regular smoker and to compare the prevalence of current regular smoking in the study cohort with that in the general population. Cox regression was used to examine associations between explanatory factors and age at smoking initiation. All statistical tests were two-sided.

Results

Of the 10 326 childhood cancer survivors who returned completed questionnaires, 20.0% were current regular smokers and 29.8% were ever regular smokers, whereas in the comparable general population 28.1% were current regular smokers and 48.8% were ever regular smokers. Current regular smoking was more prevalent among survivors of Wilms tumor or Hodgkin lymphoma than survivors of a central nervous system (CNS) neoplasm; in those aged 10–14 years at diagnosis than 0–4 years; in those not treated with radiotherapy; in those in manual occupations; in those who were separated, widowed, or divorced; in those with lower educational attainment; and in those not currently on long-term regular hospital follow-up. Rates of smoking initiation were lower in women; in those treated with chemotherapy or radiotherapy; and in those with a third party–completed questionnaire. The rate of smoking initiation was highest among those diagnosed at 10–14 years of age. The odds ratio for being a current regular smoker among the survivors compared with the general population was 0.51 (99% confidence interval [CI] = 0.46 to 0.57). Survivors who smoked, smoked fewer cigarettes per day than smokers in the general population; the difference in the multivariable model was 1.5 fewer cigarettes per day (95% CI = 1.03 to 1.99).

Conclusion

The prevalence of smoking varies by subgroup among adult survivors of childhood cancer in the BCCSS but is substantially less overall than that in the general population.

CONTEXT AND CAVEATS
Prior knowledge

Survivors of childhood cancer are at increased risk of second malignant neoplasms due to the cancer and the treatments they received and, for some, because they have a genetic condition that predisposes them to multiple primary neoplasms. For such individuals, cigarette smoking would be an additional source of risk.

Study design

A population-based cohort study of 17 981 survivors of childhood cancer in Britain in which smoking status and age at initiation of regular smoking were ascertained via a questionnaire.

Contribution

The extent of smoking among survivors of childhood cancer diagnosed in Britain between 1940 and 1991 was less than that seen in the comparable general population. Survivors of Hodgkin lymphoma, soft tissue sarcoma, or Wilms tumor were more likely to be current regular smokers, had greater rates of initiating regular smoking, and had some of the highest odds of current smoking prevalence compared with the general population than survivors of other types of childhood cancer.

Implications

Advice on the health risks of smoking should be included in any program of clinical follow-up for survivors of childhood cancer. Smoking prevention and cessation interventions should be developed and implemented in this vulnerable group.

Limitations

The self-reported smoking data were based on responses to a questionnaire rather than on biochemical data. Some questionnaires were completed by a third party rather than by the childhood cancer survivor. The analysis did not include information relating to the survivors who had died before their smoking status could be ascertained.

From the Editors

Cigarette smoking (hereafter referred to as smoking) has been implicated as a risk factor for a variety of illnesses and medical conditions, including cancers of the lung, larynx, pharynx, esophagus, bladder, kidney, pancreas, stomach, liver, and cervix and myeloid leukemia ( 1 ); cardiovascular disease ( 2 ); chronic obstructive pulmonary disease ( 3 ); and infertility ( 4 ). Smoking should therefore be discouraged for all, but especially among survivors of childhood cancer, who are at increased risk of the adverse effects of the cancer and its treatments, including second malignant neoplasms ( 5–8 ), cardiac failure ( 6 , 9 ), stroke ( 10 ), pulmonary complications ( 11 ), and infertility ( 6 , 12 , 13 ). In addition, some individuals who are diagnosed with cancer during childhood have a genetic condition that predisposes them to multiple primary neoplasms, such as hereditary retinoblastoma ( 14 , 15 ), Li–Fraumeni syndrome ( 16 ), or neurofibromatosis ( 17 , 18 ). Such genetically predisposed individuals who have been treated with radiotherapy or chemotherapy and who smoke are therefore exposing themselves to a third, additional, source of risk.

Adult survivors of childhood cancer have been found to have lower rates of smoking than the general population ( 19 , 20 ) and healthy control subjects ( 21–25 ) except in two studies ( 26 , 27 ) that found similar levels of smoking among survivors and control subjects. Factors that have been found to be associated with an increased risk of smoking among childhood cancer survivors include older attained age ( 20 , 24 ), being female ( 20 ), being male ( 24 ), having a low income ( 19 ), being white ( 19 , 24 , 28 ), being less educated ( 19 , 20 , 24 , 25 ), and being 10 years or older at diagnosis ( 19 , 25 ). Having received particular treatments for childhood cancer has also been shown to be associated with smoking prevalence in survivors; those who received potentially pulmonary-toxic cancer treatments, such as bleomycin, carmustine, or chest irradiation ( 19 ), or brain irradiation ( 19 , 25 ) were less likely to begin smoking than those who did not have such treatments.

Although survivors of childhood cancer generally smoke less than healthy control subjects and the general population data, some studies ( 22–24 ) suggest that survivors are less likely to quit smoking than siblings or other control subjects who had not been diagnosed with a childhood cancer. It is therefore important to develop strategies to prevent the initiation of smoking by childhood cancer survivors. In this context, we examined factors that may be associated with initiation of smoking in adult survivors of childhood cancer using data from the British Childhood Cancer Survivor Study (BCCSS), a population-based cohort of most of the adult survivors of childhood cancer in Britain. Because two-thirds of the current smokers in the general population of Britain begin smoking before the age of 18 years ( 29 ), interventions to prevent smoking are unlikely to be appropriate for the majority of adult survivors of childhood cancer. Therefore, we also examined factors that may be associated with being a current regular smoker among adult survivors of childhood cancer with the goal of identifying factors that could be exploited in the development of a smoking cessation intervention tailored to the specialized needs of this group of individuals. Finally, we compared the extent of smoking among adult survivors of childhood cancer with that in the general population of Britain.

Participants and Methods

Participants

The BCCSS was established to investigate the risks of adverse health and social outcomes among survivors of childhood cancer in Britain. The BCCSS objectives, methods, population structure, response rates, and initial descriptive information are described elsewhere ( 30 ). The BCCSS is based on a cohort of 17 981 individuals who were diagnosed with cancer during childhood (ie, from birth through 14 years of age) between 1940 and 1991 in Britain and survived for at least 5 years after diagnosis. The cohort was identified using the population-based National Registry of Childhood Tumors, which includes all neoplasms diagnosed in British residents younger than 15 years of age. Among these cohort members, 14 836 were alive and at least 16 years old up to September 17, 2006 (when the Study Co-ordinating Centre stopped dispatching questionnaires); these survivors were eligible to receive a postal questionnaire via their general practitioner. When a survivor’s current general practitioner could be identified, he/she was requested to send the questionnaire to those eligible survivors for whom there were no known reasons to avoid such contact and for whom the current address was available. The questionnaire was designed to examine the risks of particular adverse health and social outcomes occurring among survivors and thus enable investigation of the variation of such risks in relation to type of childhood cancer and the treatment received (available at http://www.pcpoh.bham.ac.uk/publichealth/cccss/pdfs/completeq_m.pdf [male version of questionnaire] and http://www.pcpoh.bham.ac.uk/publichealth/cccss/pdfs/completeq_f.pdf [female version of questionnaire]). Survivors who were impaired to such an extent that they could not complete the questionnaire themselves were asked to have a relative, friend, or a BCCSS staff member (ie, a third party) complete the form with as much input from the survivor as was possible. Most questionnaires were sent out in batches of approximately 100–200 per week during 2001 and 2002. Questionnaires continued to be sent out from 2003 through 2006 at a much reduced rate, mostly to survivors who had only recently reached 16 years of age or to those for whom there had not been a current general practitioner or accurate survivor address available previously; at the same time, the questionnaires that had been received were coded and computerized. The questionnaires used for this analysis were completed by the survivors from March 1, 2001, to December 20, 2006; 50.4% were completed in 2002, 35.8% in 2001, 9.3% in 2003, 2.6% in 2004, 1.3% in 2005, and 0.6% in 2006. Ethical approval for the study was obtained from a multicenter research ethics committee and from all 212 local research ethics committees in Britain.

Measures of Smoking Prevalence and Age at Initiation

Cohort members were classified as current regular smokers (yes, no) and as ever regular smokers (yes, no) based on their responses to two questions: “Do you smoke cigarettes at all nowadays?” and “Have you ever smoked cigarettes regularly?” Those who answered “yes” to the first question and provided their age (in whole years) when they started smoking regularly were classified as current regular smokers. In addition to the current regular smokers, those who answered “no” to the first question and “yes” to the second question were classified as ever regular smokers.

The survivors were also grouped according to the number of cigarettes they smoked daily (0–9, 10–19, 20–29, or ≥30 cigarettes) and their age at initiation of smoking (0–12, 13–16, 17–20, 21–24, or ≥25 years) using groups established in general population smoking statistics ( 29 ).

Explanatory Factors Investigated in Relation to Smoking Prevalence and Age at Initiation of Smoking

Survivors were categorized by age group at questionnaire completion, region of residence within Great Britain, sex, childhood cancer type, age group at diagnosis, whether they were on long-term regular hospital follow-up, and whether they had received initial treatment with chemotherapy, radiotherapy, or surgery for their childhood neoplasm. The age at questionnaire completion categories we used were the same as those used in compiling data from questionnaires about smoking prevalence in the general population (16–19, 20–24, 25–34, 35–49 years), with the following exception: we combined the two oldest age groups used for the general population data (50–59 and ≥60 years) into one group (ie, ≥50 years) because of the small number of BCCSS members who were 50 years or older ( 29 ). We used the same age-at-diagnosis categories (ie, 0–4, 5–9, and 10–14 years) that were used previously ( 30 ); these groups were chosen because outcomes such as a second malignant neoplasm have been shown to vary among these age-at-diagnosis groups ( 7 ). Region was based on government office regions in Great Britain ( 29 ) and was established using the survivor's address at questionnaire completion or, if not available, the location of their registered general practitioner. The childhood cancer types were classified as follows: central nervous system (CNS) neoplasm, leukemia (including acute lymphoblastic leukemia, acute nonlymphocytic leukemia, and chronic myeloid leukemia), Hodgkin lymphoma, non-Hodgkin lymphoma, neuroblastoma, heritable retinoblastoma (ie, those with a family history of retinoblastoma and/or a diagnosis of bilateral retinoblastoma), nonheritable retinoblastoma, Wilms tumor, bone sarcoma, soft tissue sarcoma, and other neoplasms (which included Burkitt lymphoma, intracranial and intraspinal germ cell tumors, gonadal germ cell tumors, other and unspecified nongonadal germ cell tumors, thyroid carcinomas, malignant melanoma, skin carcinoma, and other less frequent carcinomas). Information on whether the survivor was on long-term regular hospital follow-up (yes, no) came from the survivor's general practitioner. Each questionnaire was initially sent to the survivor's general practitioner, who acted as the “gatekeeper” to the survivor, and the general practitioner was requested to forward the questionnaire to the survivor. The general practitioner was also asked to complete a consent form to enable the BCCSS to contact the survivor directly; this form also asked the general practitioner to indicate whether the survivor was on long-term regular hospital follow-up in relation to his or her childhood neoplastic disease. Initial treatment information was available from the National Registry of Childhood Tumors; however, for some survivors there was no record of whether they were initially treated with chemotherapy, radiotherapy, or surgery and those survivors were excluded from analyses involving these factors (no record: of chemotherapy [n = 3173], of radiotherapy [n = 2930], of surgery [n = 2729]).

We also examined the following social and economic factors: legal marital status, highest level of educational attainment, socioeconomic classification, and whether the questionnaire was completed by a third party. The levels for educational attainment relate to formal educational qualifications in Great Britain, which include failure to pass or take any standard public educational examination; level 1 educational attainment, which corresponds to a low level of qualification obtained in Great Britain (eg, General Certificate of Secondary Education); level 2 educational attainment, which corresponds to an intermediate level of educational attainment (eg, Advanced Level qualification); and level 3 educational attainment, which corresponds to the highest level of qualification (ie, a first degree or higher). For socioeconomic classification, the survivor's present or most recent occupation was used to classify survivors into “managerial and professional occupations” (highest socioeconomic classification group), “intermediate occupations,” or “routine and manual occupations” (lowest socioeconomic classification group) as categorized by national statistics ( 29 ). Two further classifications, “student” and “never worked or unemployed,” were added because the BCCSS population had a high proportion in each of these latter classifications.

Statistical Methods

We conducted two types of analyses: an internal analysis and an external analysis. All analyses were carried out using Stata statistical software (version 9.0; Stata Corp., College Station, TX).

Internal Analysis.

The internal analysis examined factors that were associated with being a current regular smoker and age at initiation of regular smoking within the survivor population. We decided a priori that the following factors should be investigated in logistic regression as possible explanatory factors for being a current regular smoker because of their established association with smoking in general population studies ( 29 ) or their relevance to this cohort of childhood cancer survivors: sex; age at questionnaire completion; type of childhood cancer; age at diagnosis; whether the survivor had received chemotherapy, radiotherapy, or surgery; marital status; socioeconomic classification; level of educational attainment; third-party completion of the questionnaire; and whether the survivor was known to be on long-term regular hospital follow-up. Treatment decade and the interval from 5 years after diagnosis to questionnaire completion were not included in the logistic regression because there was sufficient evidence of colinearity of age at questionnaire completion, treatment decade, and interval from 5 years after diagnosis to questionnaire completion, which made it impossible to assess the separate effects of any one factor adjusted for the others. We decided before the analysis was completed to use the CNS neoplasm survivors as the referent group due to the wide ranges in current age and age at diagnosis and the large sample size of this particular group of survivors.

Tests for heterogeneity (based on the likelihood ratio statistic) were used to investigate the association between each factor and the prevalence of being a current regular smoker, first in univariate logistic regression model and then in multivariable logistic regression models that controlled for all factors identified a priori. Tests for linear trend were undertaken for factors with better than categorical measurement properties by assigning increasing consecutive integer values to levels of the factor and a value of zero to the baseline level (ie, score); tests for evidence of departure from a linear trend were also undertaken using likelihood ratio tests comparing the model with the factor as the categorical variable to the model with the factor as scores. Statistical significance was defined as P less than .01, with a two-sided test, because of the large sample size; consequently, 99% confidence intervals (CIs) are reported.

We used survival analysis to investigate the age at initiation of regular smoking. The age at initiation of smoking was analyzed with entry to risk for initiating smoking from the age at which the subject first came under observation for initiation of smoking, which was the survivor's age at 5 years after his or her diagnosis. Survivors who started smoking before this age but subsequent to diagnosis were excluded from the initial survival analysis, although they were included in the further survival analysis (see below). Exit from risk of initiating smoking (ie, the age at which the subject was last under observation for initiating smoking) was age at smoking initiation for the regular smokers and was defined as the age at the midpoint of the year provided on the questionnaire by the survivor; for nonsmokers and never regular smokers, exit from risk (censorship) was the date of questionnaire completion, if available, and otherwise it was the date that we received the returned completed questionnaire. Tests for heterogeneity (from the likelihood ratio statistic) were used to investigate the association between each potential explanatory factor and age at initiation of regular smoking, first in univariate Cox regression models and then in multivariable Cox regression models that controlled for all factors identified a priori, with the exception of socioeconomic classification, marital status, educational attainment, region, and whether on long-term regular hospital follow-up. These latter factors were excluded from the Cox regression because they are time-dependent variables that could have changed over the period at risk; however, we did incorporate these factors (as measured at the time of questionnaire completion) into the Cox multivariable regression model to check the sensitivity of the results for the core demographic, cancer, and treatment factors. Tests for trend and for a departure from a linear trend were undertaken for factors with better than categorical measurement properties as described above. Statistical significance was defined at the 1% level because of the large sample size. The proportional hazards assumption of the Cox regression was investigated both graphically and by formally testing the Schoenfeld residuals, and no violations were identified. The survival analysis, including the Cox regression, was repeated by including the survivors who started smoking after their diagnosis but before they had reached their 5-year survival date. This further analysis also began at the survivor's age at 5 years after diagnosis, but those individuals who started smoking before this age entered risk as regular smokers.

External Analysis.

The external analysis compared smoking prevalence levels and number of cigarettes smoked per day among smokers within the survivor population with smoking prevalence and the number of cigarettes smoked per day in the general population of Britain after adjusting for demographic and social differences. The general population data used for the comparison was from the 2002 General Household Survey (GHS) in Britain ( 29 ), which surveyed 8620 households in Britain in 2002 and 2003, resulting in a sample size of 20 149 people of all ages (range = 0–99 years) with an average response rate of 69%. A total of 14 788 individuals aged 16 years or older reported their smoking status. Ascertainment of smoking status was similar for the survivors and the general population because the questions about smoking in the BCCSS questionnaire were taken verbatim from the GHS.

Multivariable logistic regression was used to obtain odds ratios (ORs) of current regular smoking (and 99% CIs) in the BCCSS vs the general population of Britain ( 31 ), with adjustment for key demographic and social variables identified in the internal analysis and by the GHS ( 29 ). We used a multivariable generalized estimating equation (GEE) logistic regression model to take into account household clustering within the GHS; a weighting factor for the GHS data was used to compensate for nonresponse in the GHS and to match the GHS sample to known population distributions. Statistical significance was defined as P less than .01.

We also used a multivariable GEE linear regression model together with the weighting factor from the GHS to compare the mean number of cigarettes smoked per day among current regular smokers between the BCCSS and the general population. Because of the much smaller sample sizes available (eg, there were only 47 current regular smokers among survivors of heritable retinoblastoma), statistical significance for this analysis only was defined as P less than .05 and 95% CIs were estimated.

Results

Characteristics of Participants

Of the 14 836 survivors who were eligible to receive a questionnaire, 10 483 (70.7%) returned a completed questionnaire; of these, 10 326 (98.5%) could be classified according to their smoking status. A total of 20.0% of the survivors were classified as current regular smokers, and 29.8% were classified as ever regular smokers ( Table 1 ). Approximately 21.1% of the current regular smokers smoked an average of 20 or more cigarettes per day. Among the ever regular smokers, 88.0% had initiated smoking at age 20 years or younger; 52.2% had initiated smoking at age 16 years or younger and 35.8% at 17–20 years of age. Among the ever regular smokers, 51 started smoking before they were diagnosed with their childhood cancer and were excluded from further analyses, and an additional 508 started smoking before 5 years after their diagnosis.

Table 1

Smoking status, amounts smoked, and age of initiation of smoking among survivors of childhood cancer

Classification Current regular cigarette smoker Ever regular cigarette smoker 
Smoking prevalence, No. (%)   
    Yes 2069 (20.0) 3082 (29.8) 
    No 8257 (80.0) 7244 (70.2) 
Number of cigarettes smoked daily, No. (%)   
    0–9 776 (37.8) 1089 (35.6) 
    10–19 845 (41.1) 1226 (40.1) 
    20–29 382 (18.6) 627 (20.5) 
    ≥30 52 (2.5) 114 (3.7) 
    Missing 14 26 
        Mean (SD) 11.8 (7.56) 12.2 (8.08) 
        Median (range)  10.0 (0–60) *  10.0 (0–70) * 
Age started smoking in y, No. (%)   
    0–12 96 (4.6) 124 (4.0) 
    13–16 1041 (50.3) 1478 (48.2) 
    17–20 700 (33.8) 1097 (35.8) 
    21–24 144 (7.0) 219 (7.2) 
    ≥25 88 (4.3) 146 (4.8) 
    Missing 18 
        Mean (SD) 17.4 (3.55) 17.6 (3.60) 
        Median (range) † 16.5 (5.5–41.5) 16.5 (5.5–41.5) 
Classification Current regular cigarette smoker Ever regular cigarette smoker 
Smoking prevalence, No. (%)   
    Yes 2069 (20.0) 3082 (29.8) 
    No 8257 (80.0) 7244 (70.2) 
Number of cigarettes smoked daily, No. (%)   
    0–9 776 (37.8) 1089 (35.6) 
    10–19 845 (41.1) 1226 (40.1) 
    20–29 382 (18.6) 627 (20.5) 
    ≥30 52 (2.5) 114 (3.7) 
    Missing 14 26 
        Mean (SD) 11.8 (7.56) 12.2 (8.08) 
        Median (range)  10.0 (0–60) *  10.0 (0–70) * 
Age started smoking in y, No. (%)   
    0–12 96 (4.6) 124 (4.0) 
    13–16 1041 (50.3) 1478 (48.2) 
    17–20 700 (33.8) 1097 (35.8) 
    21–24 144 (7.0) 219 (7.2) 
    ≥25 88 (4.3) 146 (4.8) 
    Missing 18 
        Mean (SD) 17.4 (3.55) 17.6 (3.60) 
        Median (range) † 16.5 (5.5–41.5) 16.5 (5.5–41.5) 
*

For some survivors, the amounts were zero because they indicated that they smoked less than 1 cigarette per day (current regular smoker, n = 37; ever regular cigarette smoker, n = 79).

For the statistical analysis, the ages provided on the questionnaire (as whole years) by the survivor were taken to the midpoint of the year provided.

Risk Factors Associated With Being a Current Regular Smoker

Table 2 presents the frequencies of current regular smoking by potential demographic, cancer, treatment, social, and economic explanatory factors. From the multivariable modeling, age at questionnaire completion, childhood cancer type, treatment with radiotherapy, socioeconomic classification, marital status, level of educational attainment, and whether on long-term regular hospital follow-up were associated with being a current regular smoker ( P < .01) ( Table 2 ). In the multivariable model, an analysis by age at questionnaire completion showed that current regular smoking was most prevalent among those aged 20–34 years at questionnaire completion compared with those aged 16–19 years (OR for 20- to 24-year olds = 1.29, 99% CI = 0.82 to 2.02; OR for 25- to 34-year olds = 1.30, 99% CI = 0.83 to 2.03). An analysis by childhood cancer type showed that compared with those diagnosed with a CNS neoplasm, among whom the prevalence of current smoking was the lowest, those diagnosed with Wilms tumor (OR = 2.75, 99% CI = 1.79 to 4.23), Hodgkin lymphoma (OR = 2.63, 99% CI = 1.70 to 4.05), soft tissue sarcoma (OR = 2.34, 99% CI = 1.60 to 3.43), or leukemia (OR = 2.20, 99% CI = 1.42 to 3.39) were those with the highest prevalence of current regular smokers. Survivors of heritable retinoblastoma (OR = 1.38, 99% CI = 0.66 to 2.89) were the only group of survivors who were not statistically significantly different from survivors of a CNS neoplasm in the likelihood of being a current regular smoker. Survivors who received radiotherapy treatment for their childhood cancer (OR = 0.69, 99% CI = 0.56 to 0.85) were less likely to be a current regular smoker compared with those not so treated. Compared with those in the highest socioeconomic group (ie, those in managerial or professional occupations), those in the lowest socioeconomic group (ie, those in routine or manual occupations) were more likely to be a current regular smoker (OR = 1.59, 99% CI = 1.23 to 2.05). Compared with those who were single, those who were separated, widowed, or divorced were more likely to be a current regular smoker (OR = 1.46, 99% CI = 1.07 to 2.00). Compared with those who had failed to pass or take any standard public educational examination, those with increasingly higher levels of educational attainment were increasingly less likely to be a current regular smoker ( Ptrend < .001). Compared with those not on long-term regular hospital follow-up, those on long-term follow-up were less likely to be a current regular smoker (OR = 0.68, 99% CI = 0.56 to 0.84). In the test for heterogeneity, age at diagnosis was not statistically significantly associated with being a current regular smoker ( Pheterogeneity = .016), but there was a statistically significant trend for increasing likelihood of being a current smoker with increasing age at diagnosis ( Ptrend = .004).

Table 2

Frequencies of current regular smokers among survivors of childhood cancer by specific potential demographic, cancer, treatment, social, and economic explanatory factors, with odds ratios for being a current regular smoker from univariate and multivariable logistic regression models *

  % current regular smoker Univariate model Multivariable model 
Factor Total No. of survivors OR (99% CI) Pheterogeneity† Ptrend ( Pnonlinearity ) ‡ OR (99% CI) Pheterogeneity† Ptrend ( Pnonlinearity ) ‡ 
Overall § 10 275 19.8       
Sex         
    Male 5241 21.1 1.00 (referent)   1.00 (referent)   
    Female 5034 18.5 0.85 (0.75 to 0.96) <.001  0.90 (0.75 to 1.07) .118  
Age at questionnaire completion, y         
    16–19 2001 13.2 1.00 (referent)   1.00 (referent)   
    20–24 1716 20.6 1.70 (1.35 to 2.14)   1.29 (0.82 to 2.02)   
    25–34 3551 22.8 1.95 (1.60 to 2.38)   1.30 (0.83 to 2.03)   
    35–49 2484 20.6 1.71 (1.38 to 2.11)   0.98 (0.60 to 1.60)   
    ≥50 523 18.6 1.50 (1.07 to 2.10) <.001 <.001 (<.001) 0.85 (0.46 to 1.56) .007 .070 (.012) 
Childhood cancer type         
    CNS neoplasm 2164 15.4 1.00 (referent)   1.00 (referent)   
    Leukemia 2824 18.1 1.21 (0.99 to 1.47)   2.20 (1.42 to 3.39)   
    Hodgkin lymphoma 715 25.9 1.91 (1.46 to 2.50)   2.63 (1.70 to 4.05)   
    Non-Hodgkin lymphoma 530 20.2 1.39 (1.01 to 1.90)   1.84 (1.11 to 3.03)   
    Neuroblastoma 416 20.7 1.43 (1.01 to 2.02)   2.01 (1.21 to 3.34)   
    Heritable retinoblastoma ‖ 294 16.0 1.04 (0.67 to 1.61)   1.38 (0.66 to 2.89)   
    Nonheritable retinoblastoma ‖ 409 25.4 1.87 (1.34 to 2.60)   2.06 (1.21 to 3.51)   
    Wilms tumor 957 23.2 1.65 (1.29 to 2.12)   2.75 (1.79 to 4.23)   
    Bone sarcoma 390 20.8 1.44 (1.01 to 2.05)   1.82 (1.09 to 3.03)   
    Soft tissue sarcoma 704 24.4 1.77 (1.35 to 2.33)   2.34 (1.60 to 3.43)   
    Other neoplasm ¶ 872 21.7 1.52 (1.17 to 1.97) <.001  1.54 (1.07 to 2.22) <.001  
Age at diagnosis, y         
    0–4 4759 18.4 1.00 (referent)   1.00 (referent)   
    5–9 2740 20.5 1.14 (0.98 to 1.34)   1.19 (0.94 to 1.52)   
    10–14 2776 21.7 1.23 (1.06 to 1.43) .001 <.001 (.579) 1.35 (1.03 to 1.77) .016 .004 (.755) 
Chemotherapy         
    No  3266 # 20.9 1.00 (referent)   1.00 (referent)   
    Yes  3836 # 20.1 0.95 (0.81 to 1.10) .365  0.76 (0.58 to 1.00) .011  
Radiotherapy         
    No  2166 # 24.2 1.00 (referent)   1.00 (referent)   
    Yes  5179 # 18.8 0.72 (0.62 to 0.85) <.001  0.69 (0.56 to 0.85) <.001  
Surgery         
    No  3346 # 19.4 1.00 (referent)   1.00 (referent)   
    Yes  4200 # 21.4 1.13 (0.98 to 1.31) .032  1.20 (0.88 to 1.63) .123  
Socioeconomic classification **         
    Managerial/professional occupations  2273 # 15.6 1.00 (referent)   1.00 (referent)   
    Intermediate occupations  1791 # 19.9 1.35 (1.09 to 1.67)   1.04 (0.80 to 1.37)   
    Routine/manual occupations  3120 # 27.8 2.08 (1.74 to 2.50)   <.001 (.328) †† 1.59 (1.23 to 2.05)   <.001 (.033) †† 
    Never worked/unemployed  594 # 24.6 1.77 (1.33 to 2.35)   1.18 (0.78 to 1.79)   
    Student  1718 # 9.4 0.56 (0.44 to 0.73) <.001  0.58 (0.35 to 0.94) <.001  
Legal marital status         
    Single  6735 # 19.4 1.00 (referent)   1.00 (referent)   
    Married  2682 # 17.3 0.87 (0.74 to 1.01)   0.60 (0.47 to 0.75)   
    Separated/widowed/divorced  615 # 33.5 2.07 (1.64 to 2.62) <.001  1.46 (1.07 to 2.00) <.001  
Region ‡‡         
    England         
        London  945 # 19.5 1.00 (referent)   1.00 (referent)   
        Northeast  476 # 22.9 1.23 (0.86 to 1.75)   0.83 (0.51 to 1.36)   
        Northwest  1224 # 19.6 1.01 (0.76 to 1.34)   0.81 (0.55 to 1.19)   
        Yorkshire and the Humber  877 # 18.0 0.91 (0.67 to 1.24)   0.75 (0.49 to 1.15)   
        East Midlands  786 # 18.8 0.96 (0.70 to 1.32)   0.74 (0.49 to 1.14)   
        West Midlands  946 # 17.4 0.87 (0.64 to 1.19)   0.72 (0.47 to 1.10)   
        East of England  1098 # 21.0 1.10 (0.82 to 1.46)   0.91 (0.63 to 1.32)   
        Southeast  1481 # 18.7 0.95 (0.72 to 1.25)   0.81 (0.56 to 1.15)   
        Southwest  1031 # 20.5 1.06 (0.80 to 1.42)   0.96 (0.65 to 1.40)   
    Wales  509 # 19.3 0.99 (0.69 to 1.41)   0.69 (0.42 to 1.14)   
    Scotland  890 # 24.0 1.31 (0.98 to 1.75) .024  1.26 (0.85 to 1.87) .010  
Educational attainment §§         
    Failed to pass or take any standard public educational examination  1510 # 23.0 1.00 (referent)   1.00 (referent)   
    Level 1  4134 # 22.6 0.98 (0.82 to 1.18)   0.73 (0.55 to 0.96)   
    Level 2  2839 # 18.4 0.75 (0.62 to 0.92)   0.59 (0.43 to 0.80)   
    Level 3  1418 # 11.4 0.43 (0.33 to 0.56) <.001 <.001 (<.001) 0.31 (0.20 to 0.46) <.001 <.001 (.026) 
Third party–completed questionnaire         
    No 8988 20.6 1.00 (referent)   1.00 (referent)   
    Yes 1287 14.3 0.64 (0.52 to 0.80) <.001  0.74 (0.52 to 1.04) .019  
Hospital follow-up ‖ ‖         
    Not on regular follow-up  6220 # 22.9 1.00 (referent)   1.00 (referent)   
    On regular follow-up  3740 # 14.9 0.59 (0.51 to 0.68) <.001  0.68 (0.56 to 0.84) <.001  
  % current regular smoker Univariate model Multivariable model 
Factor Total No. of survivors OR (99% CI) Pheterogeneity† Ptrend ( Pnonlinearity ) ‡ OR (99% CI) Pheterogeneity† Ptrend ( Pnonlinearity ) ‡ 
Overall § 10 275 19.8       
Sex         
    Male 5241 21.1 1.00 (referent)   1.00 (referent)   
    Female 5034 18.5 0.85 (0.75 to 0.96) <.001  0.90 (0.75 to 1.07) .118  
Age at questionnaire completion, y         
    16–19 2001 13.2 1.00 (referent)   1.00 (referent)   
    20–24 1716 20.6 1.70 (1.35 to 2.14)   1.29 (0.82 to 2.02)   
    25–34 3551 22.8 1.95 (1.60 to 2.38)   1.30 (0.83 to 2.03)   
    35–49 2484 20.6 1.71 (1.38 to 2.11)   0.98 (0.60 to 1.60)   
    ≥50 523 18.6 1.50 (1.07 to 2.10) <.001 <.001 (<.001) 0.85 (0.46 to 1.56) .007 .070 (.012) 
Childhood cancer type         
    CNS neoplasm 2164 15.4 1.00 (referent)   1.00 (referent)   
    Leukemia 2824 18.1 1.21 (0.99 to 1.47)   2.20 (1.42 to 3.39)   
    Hodgkin lymphoma 715 25.9 1.91 (1.46 to 2.50)   2.63 (1.70 to 4.05)   
    Non-Hodgkin lymphoma 530 20.2 1.39 (1.01 to 1.90)   1.84 (1.11 to 3.03)   
    Neuroblastoma 416 20.7 1.43 (1.01 to 2.02)   2.01 (1.21 to 3.34)   
    Heritable retinoblastoma ‖ 294 16.0 1.04 (0.67 to 1.61)   1.38 (0.66 to 2.89)   
    Nonheritable retinoblastoma ‖ 409 25.4 1.87 (1.34 to 2.60)   2.06 (1.21 to 3.51)   
    Wilms tumor 957 23.2 1.65 (1.29 to 2.12)   2.75 (1.79 to 4.23)   
    Bone sarcoma 390 20.8 1.44 (1.01 to 2.05)   1.82 (1.09 to 3.03)   
    Soft tissue sarcoma 704 24.4 1.77 (1.35 to 2.33)   2.34 (1.60 to 3.43)   
    Other neoplasm ¶ 872 21.7 1.52 (1.17 to 1.97) <.001  1.54 (1.07 to 2.22) <.001  
Age at diagnosis, y         
    0–4 4759 18.4 1.00 (referent)   1.00 (referent)   
    5–9 2740 20.5 1.14 (0.98 to 1.34)   1.19 (0.94 to 1.52)   
    10–14 2776 21.7 1.23 (1.06 to 1.43) .001 <.001 (.579) 1.35 (1.03 to 1.77) .016 .004 (.755) 
Chemotherapy         
    No  3266 # 20.9 1.00 (referent)   1.00 (referent)   
    Yes  3836 # 20.1 0.95 (0.81 to 1.10) .365  0.76 (0.58 to 1.00) .011  
Radiotherapy         
    No  2166 # 24.2 1.00 (referent)   1.00 (referent)   
    Yes  5179 # 18.8 0.72 (0.62 to 0.85) <.001  0.69 (0.56 to 0.85) <.001  
Surgery         
    No  3346 # 19.4 1.00 (referent)   1.00 (referent)   
    Yes  4200 # 21.4 1.13 (0.98 to 1.31) .032  1.20 (0.88 to 1.63) .123  
Socioeconomic classification **         
    Managerial/professional occupations  2273 # 15.6 1.00 (referent)   1.00 (referent)   
    Intermediate occupations  1791 # 19.9 1.35 (1.09 to 1.67)   1.04 (0.80 to 1.37)   
    Routine/manual occupations  3120 # 27.8 2.08 (1.74 to 2.50)   <.001 (.328) †† 1.59 (1.23 to 2.05)   <.001 (.033) †† 
    Never worked/unemployed  594 # 24.6 1.77 (1.33 to 2.35)   1.18 (0.78 to 1.79)   
    Student  1718 # 9.4 0.56 (0.44 to 0.73) <.001  0.58 (0.35 to 0.94) <.001  
Legal marital status         
    Single  6735 # 19.4 1.00 (referent)   1.00 (referent)   
    Married  2682 # 17.3 0.87 (0.74 to 1.01)   0.60 (0.47 to 0.75)   
    Separated/widowed/divorced  615 # 33.5 2.07 (1.64 to 2.62) <.001  1.46 (1.07 to 2.00) <.001  
Region ‡‡         
    England         
        London  945 # 19.5 1.00 (referent)   1.00 (referent)   
        Northeast  476 # 22.9 1.23 (0.86 to 1.75)   0.83 (0.51 to 1.36)   
        Northwest  1224 # 19.6 1.01 (0.76 to 1.34)   0.81 (0.55 to 1.19)   
        Yorkshire and the Humber  877 # 18.0 0.91 (0.67 to 1.24)   0.75 (0.49 to 1.15)   
        East Midlands  786 # 18.8 0.96 (0.70 to 1.32)   0.74 (0.49 to 1.14)   
        West Midlands  946 # 17.4 0.87 (0.64 to 1.19)   0.72 (0.47 to 1.10)   
        East of England  1098 # 21.0 1.10 (0.82 to 1.46)   0.91 (0.63 to 1.32)   
        Southeast  1481 # 18.7 0.95 (0.72 to 1.25)   0.81 (0.56 to 1.15)   
        Southwest  1031 # 20.5 1.06 (0.80 to 1.42)   0.96 (0.65 to 1.40)   
    Wales  509 # 19.3 0.99 (0.69 to 1.41)   0.69 (0.42 to 1.14)   
    Scotland  890 # 24.0 1.31 (0.98 to 1.75) .024  1.26 (0.85 to 1.87) .010  
Educational attainment §§         
    Failed to pass or take any standard public educational examination  1510 # 23.0 1.00 (referent)   1.00 (referent)   
    Level 1  4134 # 22.6 0.98 (0.82 to 1.18)   0.73 (0.55 to 0.96)   
    Level 2  2839 # 18.4 0.75 (0.62 to 0.92)   0.59 (0.43 to 0.80)   
    Level 3  1418 # 11.4 0.43 (0.33 to 0.56) <.001 <.001 (<.001) 0.31 (0.20 to 0.46) <.001 <.001 (.026) 
Third party–completed questionnaire         
    No 8988 20.6 1.00 (referent)   1.00 (referent)   
    Yes 1287 14.3 0.64 (0.52 to 0.80) <.001  0.74 (0.52 to 1.04) .019  
Hospital follow-up ‖ ‖         
    Not on regular follow-up  6220 # 22.9 1.00 (referent)   1.00 (referent)   
    On regular follow-up  3740 # 14.9 0.59 (0.51 to 0.68) <.001  0.68 (0.56 to 0.84) <.001  
*

OR = odds ratio; CI = confidence interval; CNS = central nervous system.

The P value (two-sided) is from the likelihood ratio test for heterogeneity in the probability of being a current regular smoker across different levels of the specified explanatory factor for the univariate analysis and with adjustment for all other factors in the multivariable model. The threshold for statistical significance was .01.

The Ptrend is from the test for trend and the P value in parentheses is from the test for departure from a linear trend. Both P values are two-sided. The threshold for statistical significance was .01.

§

The 51 survivors who started smoking before their diagnosis were excluded from analysis.

Heritable retinoblastoma includes those with a family history of retinoblastoma and/or who were diagnosed with bilateral retinoblastoma; all other cases of retinoblastoma were classified as nonheritable retinoblastoma.

Other neoplasms include Burkitt lymphomas; intracranial and intraspinal germ cell tumors; gonadal germ cell tumors; other and unspecified nongonadal germ cell tumors; thyroid carcinomas; and malignant melanoma, skin carcinoma, and other less frequently occurring carcinomas.

#

These values do not add up to the total because of exclusion of those with no record of treatment (chemotherapy, n = 3173; radiotherapy, n = 2930; surgery, n = 2729) and unusable or missing data (socioeconomic classification [SEC]: missing data or occupations given were inadequately described or could not be used to classify SEC, n = 779; legal marital status: missing data, n = 243; region: survivors who lived in Northern Ireland or the Channel Islands [n = 12] were excluded from the analysis; educational attainment: missing data, n = 374; whether on long-term regular hospital follow-up: missing data, n = 315).

**

For socioeconomic classification, the survivor's present or most recent occupation was used to classify survivors with “managerial and professional occupations,” “intermediate occupations,” or “routine and manual occupations” defined by the 2002 General Household Survey (29); two further categories were added: “student” and “never worked or unemployed” due to the British Childhood Cancer Survivor Study population characteristics.

††

For socioeconomic classification, the tests for trend and for departure from a linear trend were completed using the first three levels of the factor: managerial and professional, intermediate, and routine and manual occupations.

‡‡

Region was based on government office regions in Great Britain (29) and was established using the survivor's address at questionnaire completion, or if not available, the location of their registered general practitioner (GP).

§§

The different levels for the educational attainment relate to formal educational qualifications in Great Britain, which include level 1, which corresponds to a low level of qualification obtained in Great Britain (eg, General Certificate of Secondary Education); level 2, which corresponds to an intermediate level of educational attainment (eg, Advanced Level qualification); and level 3, the highest level of qualification, a first degree or higher.

‖‖

Information obtained from the consent form returned from the GP, which asked whether the survivor was on long-term regular hospital follow-up in relation to his/her childhood neoplastic disease.

Risk Factors Associated With Age at Initiation of Regular Smoking

The actuarial estimated risks of initiating regular smoking by ages 20 and 40 years were 26.2% (99% CI = 24.9% to 27.5%) and 34.0% (99% CI = 32.6% to 35.5%), respectively. When the survivors who started smoking after their diagnosis but before they had survived 5 years were included, risks for smoking initiation by ages 20 and 40 years had increased to 32.0% (99% CI = 30.6% to 33.3%) and 39.2% (99% CI = 37.7% to 40.6%), respectively ( Figure 1 ).

Figure 1

Actuarial percentages of ever regular smokers by age at initiation of smoking.

Figure 1

Actuarial percentages of ever regular smokers by age at initiation of smoking.

In the multivariable model, age at initiation of regular smoking was associated with sex, childhood cancer type, age at diagnosis, treatment with chemotherapy, treatment with radiotherapy, and third-party completion of the questionnaire ( P < .01) ( Table 3 ). Women had a lower rate of initiating regular smoking than men (hazard ratio [HR] = 0.84, 99% CI = 0.74 to 0.95). Survivors of heritable retinoblastoma (HR = 1.15, 99% CI = 0.74 to 1.79) were the only group of survivors who were not statistically significantly different from survivors of a CNS neoplasm in their rate of initiation of regular smoking; survivors of all other childhood cancer types had higher rates of smoking initiation, particularly survivors of Hodgkin lymphoma (HR = 2.30, 99% CI = 1.69 to 3.11) and Wilms tumor (HR = 2.38, 99% CI = 1.82 to 3.13), who had smoking initiation rates that were more than double that of survivors of CNS neoplasms. Survivors who were diagnosed with cancer at ages 10–14 years had the highest rate of initiating smoking at 1.27 times (99% CI = 1.05 to 1.54) that of those who were diagnosed with cancer at ages 0–4 years. The rates of smoking initiation among survivors whose childhood cancer was treated with chemotherapy (HR = 0.73, 99% CI = 0.61 to 0.87) or radiotherapy (HR = 0.81, 99% CI = 0.70 to 0.94) were lower than those among survivors who did not have such treatments. Survivors who completed the questionnaire with the help of a third party had a lower rate of smoking initiation than those who completed the questionnaire themselves (HR = 0.72, 99% CI = 0.57 to 0.91).

Table 3

HRs and 99% CIs of age at initiation of regular smoking among survivors of childhood cancer associated with specified demographic, cancer, and treatment factors from a Cox regression *

 Univariate model Multivariable model 
Factor HR (99% CI) Pheterogeneity† Ptrend ( Pnonlinearity ) ‡ HR (99% CI) Pheterogeneity† Ptrend ( Pnonlinearity ) ‡ 
Sex       
    Male 1.00 (referent)   1.00 (referent)   
    Female 0.85 (0.77 to 0.95) <.001  0.84 (0.74 to 0.95) <.001  
Childhood cancer type       
    CNS neoplasm 1.00 (referent)   1.00 (referent)   
    Leukemia 1.19 (1.01 to 1.41)   1.76 (1.30 to 2.38)   
    Hodgkin lymphoma 2.07 (1.66 to 2.59)   2.30 (1.69 to 3.11)   
    Non-Hodgkin lymphoma 1.59 (1.23 to 2.06)   1.79 (1.27 to 2.54)   
    Neuroblastoma 1.58 (1.20 to 2.07)   1.61 (1.16 to 2.25)   
    Heritable retinoblastoma 1.23 (0.89 to 1.69)   1.15 (0.74 to 1.79)   
    Nonheritable retinoblastoma 2.06 (1.61 to 2.63)   1.78 (1.28 to 2.48)   
    Wilms tumor 1.93 (1.59 to 2.34)   2.38 (1.82 to 3.13)   
    Bone sarcoma 1.57 (1.16 to 2.14)   1.52 (1.04 to 2.23)   
    Soft tissue sarcoma 1.76 (1.41 to 2.20)   1.83 (1.40 to 2.38)   
    Other neoplasms 1.66 (1.33 to 2.08) <.001  1.42 (1.09 to 1.85) <.001  
Age at diagnosis, y       
    0–4 1.00 (referent)   1.00 (referent)   
    5–9 0.96 (0.85 to 1.08)   1.05 (0.90 to 1.22)   
    10–14 1.21 (1.04 to 1.40) <.001 .013 (.003) 1.27 (1.05 to 1.54) .005 .003 (.173) 
Chemotherapy       
    No 1.00 (referent)   1.00 (referent)   
    Yes 0.85 (0.75 to 0.96) <.001  0.73 (0.61 to 0.87) <.001  
Radiotherapy       
    No 1.00 (referent)   1.00 (referent)   
    Yes 0.77 (0.68 to 0.87) <.001  0.81 (0.70 to 0.94) <.001  
Surgery       
    No 1.00 (referent)   1.00 (referent)   
    Yes 1.22 (1.08 to 1.37) <.001  1.16 (0.94 to 1.44) .0700  
Third party–completed questionnaire       
    No 1.00 (referent)   1.00 (referent)   
    Yes 0.62 (0.51 to 0.74) <.001  0.72 (0.57 to 0.91) <.001  
 Univariate model Multivariable model 
Factor HR (99% CI) Pheterogeneity† Ptrend ( Pnonlinearity ) ‡ HR (99% CI) Pheterogeneity† Ptrend ( Pnonlinearity ) ‡ 
Sex       
    Male 1.00 (referent)   1.00 (referent)   
    Female 0.85 (0.77 to 0.95) <.001  0.84 (0.74 to 0.95) <.001  
Childhood cancer type       
    CNS neoplasm 1.00 (referent)   1.00 (referent)   
    Leukemia 1.19 (1.01 to 1.41)   1.76 (1.30 to 2.38)   
    Hodgkin lymphoma 2.07 (1.66 to 2.59)   2.30 (1.69 to 3.11)   
    Non-Hodgkin lymphoma 1.59 (1.23 to 2.06)   1.79 (1.27 to 2.54)   
    Neuroblastoma 1.58 (1.20 to 2.07)   1.61 (1.16 to 2.25)   
    Heritable retinoblastoma 1.23 (0.89 to 1.69)   1.15 (0.74 to 1.79)   
    Nonheritable retinoblastoma 2.06 (1.61 to 2.63)   1.78 (1.28 to 2.48)   
    Wilms tumor 1.93 (1.59 to 2.34)   2.38 (1.82 to 3.13)   
    Bone sarcoma 1.57 (1.16 to 2.14)   1.52 (1.04 to 2.23)   
    Soft tissue sarcoma 1.76 (1.41 to 2.20)   1.83 (1.40 to 2.38)   
    Other neoplasms 1.66 (1.33 to 2.08) <.001  1.42 (1.09 to 1.85) <.001  
Age at diagnosis, y       
    0–4 1.00 (referent)   1.00 (referent)   
    5–9 0.96 (0.85 to 1.08)   1.05 (0.90 to 1.22)   
    10–14 1.21 (1.04 to 1.40) <.001 .013 (.003) 1.27 (1.05 to 1.54) .005 .003 (.173) 
Chemotherapy       
    No 1.00 (referent)   1.00 (referent)   
    Yes 0.85 (0.75 to 0.96) <.001  0.73 (0.61 to 0.87) <.001  
Radiotherapy       
    No 1.00 (referent)   1.00 (referent)   
    Yes 0.77 (0.68 to 0.87) <.001  0.81 (0.70 to 0.94) <.001  
Surgery       
    No 1.00 (referent)   1.00 (referent)   
    Yes 1.22 (1.08 to 1.37) <.001  1.16 (0.94 to 1.44) .0700  
Third party–completed questionnaire       
    No 1.00 (referent)   1.00 (referent)   
    Yes 0.62 (0.51 to 0.74) <.001  0.72 (0.57 to 0.91) <.001  
*

HR = hazard ratio; CI = confidence interval; CNS = central nervous system.

The P value (two-sided) is from the likelihood ratio test of heterogeneity for age at smoking initiation across different levels of the specified factor for the univariate analysis and with adjustment for sex, childhood cancer type, age at diagnosis, chemotherapy, radiotherapy, surgery, and third party–completed questionnaire in the multivariable model. The threshold for statistical significance was .01.

The main P value is from the test for trend, and the P value in parentheses is from the test for departure from a linear trend. Both P values are two-sided and the threshold for statistical significance was .01.

Inclusion of the time-dependent variables (as measured at the time of questionnaire completion) in the multivariable Cox regression model had minimal effect on the hazard ratios or the statistical significance of the tests for heterogeneity for the core demographic, cancer, and treatment factors, with the exception of third-party questionnaire completion, which was no longer a statistically significant risk factor for age at initiation of regular smoking (HR = 0.81, 99% CI = 0.62 to 1.06, P = .038).

Inclusion of the survivors who started smoking after diagnosis but within 5 years of their diagnosis in the survival analysis did not appreciably change the hazard ratios or statistical significance of the tests for heterogeneity from the multivariable model for any factor examined except for age at diagnosis; survivors who were diagnosed at ages 10–14 years had an increased rate of initiating smoking compared with those who were diagnosed at ages 0–4 years (HR = 2.51, 99% CI = 2.13 to 2.97, Pheterogeneity < .001). This increase in hazard ratio after including these additional survivors was not unexpected, given that 93% of these survivors were diagnosed at age 10–14 years (whereas only 7% were diagnosed at age 5–9 years and less than 1% at age 0–4 years). For the 5- to 9-year-old diagnosis group, the hazard ratio (HR = 1.10, 99% CI = 0.94 to 1.28) did not change appreciably after including the survivors who started smoking after diagnosis but before they had reached 5 years after their diagnosis.

Smoking Rates Among Childhood Cancer Survivors Compared With the General Population of Britain

Survivors of childhood cancer in the BCCSS had approximately half the odds of being a current regular smoker than the general population of Britain (OR = 0.51, 99% CI = 0.46 to 0.57; Table 4 ). (Given that the rate of current regular smoking was over 20%, however, it should be noted that the odds ratios exaggerate the actual relative risks.) The only survivors who did not have statistically significantly lower odds of being a current regular smoker than the general population were those who had been diagnosed with nonheritable retinoblastoma (OR = 0.79, 99% CI = 0.57 to 1.09). Survivors of Hodgkin lymphoma (OR = 0.74, 99% CI = 0.57 to 0.95) or soft tissue sarcoma (OR = 0.71, 99% CI = 0.54 to 0.92) had the next highest odds of being a current regular smoker compared with the general population, whereas survivors of leukemia (OR = 0.43, 99% CI = 0.36 to 0.51) or a CNS neoplasm (OR = 0.33, 99% CI = 0.27 to 0.40) had the lowest odds. Excluding the survivors who had questionnaires completed by a third party did not alter these findings substantially ( Table 4 ).

Table 4

Percentages and odds ratios (and 99% confidence intervals) for the likelihood of being a current regular smoker in the British Childhood Cancer Survivor Study (BCCSS) compared with the general population of Britain, overall and by childhood cancer type, from a general estimating equation multivariable logistic regression model *

 All completed questionnaires Excluding third party–completed questionnaires 
Diagnosis group † % current regular smoker OR (99% CI) % current regular smoker OR (99% CI) 
BCCSS overall 19.8 0.51 (0.46 to 0.57) 20.6 0.54 (0.48 to 0.60) 
CNS neoplasm 15.4 0.33 (0.27 to 0.40) 17.4 0.39 (0.31 to 0.47) 
Leukemia 18.1 0.43 (0.36 to 0.51) 18.8 0.45 (0.38 to 0.54) 
Non-Hodgkin lymphoma 20.2 0.52 (0.38 to 0.71) 19.1 0.49 (0.35 to 0.67) 
Heritable retinoblastoma 16.0 0.51 (0.33 to 0.80) 15.4 0.46 (0.27 to 0.79) 
Bone sarcomas 20.8 0.61 (0.42 to 0.87) 20.0 0.59 (0.41 to 0.86) 
Neuroblastoma 20.7 0.62 (0.44 to 0.88) 21.2 0.62 (0.43 to 0.89) 
Other neoplasms 21.7 0.63 (0.49 to 0.81) 22.3 0.65 (0.50 to 0.84) 
Wilms tumor 23.2 0.65 (0.51 to 0.81) 23.1 0.66 (0.52 to 0.83) 
Soft tissue sarcomas 24.4 0.71 (0.54 to 0.92) 24.8 0.72 (0.55 to 0.95) 
Hodgkin lymphoma 25.9 0.74 (0.57 to 0.95) 25.3 0.72 (0.56 to 0.93) 
Nonheritable retinoblastoma 25.4 0.79 (0.57 to 1.09) 26.1 0.81 (0.58 to 1.12) 
 All completed questionnaires Excluding third party–completed questionnaires 
Diagnosis group † % current regular smoker OR (99% CI) % current regular smoker OR (99% CI) 
BCCSS overall 19.8 0.51 (0.46 to 0.57) 20.6 0.54 (0.48 to 0.60) 
CNS neoplasm 15.4 0.33 (0.27 to 0.40) 17.4 0.39 (0.31 to 0.47) 
Leukemia 18.1 0.43 (0.36 to 0.51) 18.8 0.45 (0.38 to 0.54) 
Non-Hodgkin lymphoma 20.2 0.52 (0.38 to 0.71) 19.1 0.49 (0.35 to 0.67) 
Heritable retinoblastoma 16.0 0.51 (0.33 to 0.80) 15.4 0.46 (0.27 to 0.79) 
Bone sarcomas 20.8 0.61 (0.42 to 0.87) 20.0 0.59 (0.41 to 0.86) 
Neuroblastoma 20.7 0.62 (0.44 to 0.88) 21.2 0.62 (0.43 to 0.89) 
Other neoplasms 21.7 0.63 (0.49 to 0.81) 22.3 0.65 (0.50 to 0.84) 
Wilms tumor 23.2 0.65 (0.51 to 0.81) 23.1 0.66 (0.52 to 0.83) 
Soft tissue sarcomas 24.4 0.71 (0.54 to 0.92) 24.8 0.72 (0.55 to 0.95) 
Hodgkin lymphoma 25.9 0.74 (0.57 to 0.95) 25.3 0.72 (0.56 to 0.93) 
Nonheritable retinoblastoma 25.4 0.79 (0.57 to 1.09) 26.1 0.81 (0.58 to 1.12) 
*

The percentage of current regular smokers in the general population aged between 16-69 y was 28.1% ( 31 ). The general estimating equation multivariable logistic regression controlled for age group at questionnaire completion [only those younger than 70 years were included in the age group ”≥50 y” because educational attainment was only requested from individuals aged up to and including 69 years in the general household survey (29)], sex, legal marital status; socioeconomic classification, educational attainment, and region and took into account the General Household Survey weighting factor. OR = odds ratio; CI = confidence interval.

Population data from the General Household Survey was used for the reference group ( 31 ).

The current regular smokers within the BCCSS smoked statistically significantly fewer cigarettes per day than current regular smokers in the general population (mean number of cigarettes smoked per day for BCCSS vs the general population: 11.78 vs 14.29; difference between the BCCSS and the general population in number of cigarettes smoked per day from the multivariable linear regression model = 1.5, 95% CI = 1.03 to 1.99) ( Table 5 ). The difference in mean number of cigarettes smoked per day compared with the general population was statistically significant at the 5% level for survivors of Wilms tumor, a CNS neoplasm, nonheritable retinoblastoma, leukemia, heritable retinoblastoma, and Hodgkin lymphoma.

Table 5

Difference in the mean number of cigarettes smoked per day between survivors of specific cancer types or all survivors and the general population for current regular smokers, adjusted for potential confounders *

Childhood cancer type  Number of current regular smokers † Mean number of cigarettes smoked per day  Difference in mean number of cigarettes smoked per day between survivor group and the general population (95% CI) ‡ 
BCCSS overall 2024 11.78 1.51 (1.03 to 1.99) 
Bone sarcomas 80 13.41 0.24 (−1.20 to 1.68) 
Other neoplasms 188 12.26 0.39 (−0.78 to 1.56) 
Soft tissue sarcomas 172 12.52 1.25 (−0.17 to 2.66) 
Neuroblastoma 85 12.06 1.17 (−0.31 to 2.66) 
Wilms tumor 219 10.90 1.24 (0.25 to 2.23) 
Non-Hodgkin lymphoma 106 12.25 1.39 (−0.13 to 2.91) 
CNS neoplasm 332 12.67 1.58 (0.61 to 2.55) 
Nonheritable retinoblastoma 104 12.53 1.77 (0.39 to 3.15) 
Leukemia 507 10.66 2.02 (1.32 to 2.72) 
Heritable retinoblastoma 47 10.72 2.07 (0.45 to 3.68) 
Hodgkin lymphoma 184 11.86 2.21 (1.14 to 3.28) 
Childhood cancer type  Number of current regular smokers † Mean number of cigarettes smoked per day  Difference in mean number of cigarettes smoked per day between survivor group and the general population (95% CI) ‡ 
BCCSS overall 2024 11.78 1.51 (1.03 to 1.99) 
Bone sarcomas 80 13.41 0.24 (−1.20 to 1.68) 
Other neoplasms 188 12.26 0.39 (−0.78 to 1.56) 
Soft tissue sarcomas 172 12.52 1.25 (−0.17 to 2.66) 
Neuroblastoma 85 12.06 1.17 (−0.31 to 2.66) 
Wilms tumor 219 10.90 1.24 (0.25 to 2.23) 
Non-Hodgkin lymphoma 106 12.25 1.39 (−0.13 to 2.91) 
CNS neoplasm 332 12.67 1.58 (0.61 to 2.55) 
Nonheritable retinoblastoma 104 12.53 1.77 (0.39 to 3.15) 
Leukemia 507 10.66 2.02 (1.32 to 2.72) 
Heritable retinoblastoma 47 10.72 2.07 (0.45 to 3.68) 
Hodgkin lymphoma 184 11.86 2.21 (1.14 to 3.28) 
*

The mean number of cigarettes smoked per day by the current regular smokers in the general population was 14.39 ( 31 ). CI = confidence interval; BCCSS = British Childhood Cancer Survivor Study; CNS = central nervous system.

Excludes the current regular smokers who started smoking before their diagnosis and/or who did not provide the number of cigarettes smoked.

From a general estimating equation multivariable linear regression model that adjusted for sex, age at questionnaire completion, marital status, socioeconomic classification, region, and educational attainment, and took into account the General Household Survey weighting factor. The values are the adjusted regression coefficients and their 95% CIs.

Discussion

Our data indicate that the prevalence of smoking among adult survivors of childhood cancer in the BCCSS was substantially less than that in the general population (OR = 0.51, 99% CI = 0.46 to 0.57) and that overall, the childhood cancer survivors who currently smoked, smoked fewer cigarettes per day on average than the general population. Among survivors in the BCCSS, 20% were classified as being a current regular smoker, approximately 30% were classified as being an ever regular smoker, and an estimated 34%–39% of the survivors had started smoking regularly by age 40 years. Within the comparable general population (ie, those younger than 70 years), 28.1% were classified as a current regular smoker and 48.8% as an ever regular smoker ( 31 ), although the general population includes a higher proportion of older individuals than the BCCSS. The proportions within the BCCSS are similar to those reported in the North American Childhood Cancer Survivor Study (CCSS): among 9709 survivors who were 20 years or younger at cancer diagnosis between 1970 and 1986 and currently 18–47 years old, 17% were current smokers, 28% were ever smokers, and the actuarial estimated incidence of initiating smoking by age 40 years was 32% ( 19 ). Other studies ( 20 , 22–26 , 28 , 32 , 33 ) have reported levels of current smoking among survivors that varied from 14% to 33%. However, comparisons of smoking rates of survivors in different studies are often very difficult because of uncontrolled confounding by many factors, including the types of childhood cancer included, the age of the survivors, and the time from diagnosis; in addition, the methods used to classify smoking status often vary among studies.

The internal analysis revealed that survivors of Hodgkin lymphoma, soft tissue sarcoma, or Wilms tumor were more likely to be current smokers, had greater rates of initiating regular smoking, and had some of the highest odds of current smoking prevalence compared with that in the general population than survivors of other types of childhood cancer. These findings are troubling because survivors of Hodgkin lymphoma and soft tissue sarcoma are at a particularly increased risk of a second malignant neoplasm ( 8 , 34 , 35 ). Other studies ( 36 , 37 ) have reported an increased risk of a second malignant neoplasm in Wilms tumor survivors, particularly those who have undergone abdominal or lung irradiation. It is unclear why more survivors of Hodgkin lymphoma, Wilms tumor, and soft tissue sarcoma would have started smoking than survivors of other cancers. Factors other than sociodemographic factors that have been reported to be associated with engagement in unhealthy lifestyle behaviors among adolescents include personality traits, parental support, family structure, and parental and peer modeling of behaviors ( 38 ), none of which were examined here. In addition, treatment-related factors, knowledge of risks, perceived vulnerability to the health risks of smoking, and an ability to make decisions have also been reported to be associated with engagement in unhealthy lifestyle behaviors among survivors of childhood cancer ( 38 ). Further study is needed to determine which of these factors are associated with current smoking in the survivors from the BCCSS using, for instance, a structured interview with particular reference to the previously highlighted psychosocial factors.

Survivors of childhood cancer have been reported to be at an increased risk of a second malignant neoplasm that is approximately six times higher than that expected, and this increased risk varies with treatment type and intensity ( 7 , 8 ). Being a smoker and smoking for a considerable period are likely to further increase the risk of a second malignant neoplasm among survivors who have an established excess risk attributable to treatment. For example, smoking has been shown to increase the risk of lung cancer in survivors of Hodgkin lymphoma in addition to the increased risks from radiotherapy and chemotherapy ( 39 , 40 ). It is therefore imperative that survivors who have been treated with chemotherapy and/or radiotherapy do not smoke. Survivors in the BCCSS who were treated with radiotherapy for their childhood cancer were less likely to be a current smoker and had a reduced rate of initiating regular smoking than those who did not receive radiotherapy. Chemotherapy treatment, in particular with anthracyclines and epipodophyllotoxins, is also associated with an increased risk of a second malignant neoplasm ( 8 ). We found that BCCSS survivors who were treated with chemotherapy for their childhood cancer had a lower rate of initiating smoking and were less likely to be current smokers (borderline statistical significance) than those who did not receive chemotherapy.

It is encouraging that relatively fewer survivors of heritable retinoblastoma were current smokers than were survivors of all other specific childhood cancer types (including nonheritable retinoblastoma) except CNS neoplasms. The odds of heritable retinoblastoma survivors being a current regular smoker were half those of the corresponding general population. Survivors of heritable retinoblastoma also had one of the lowest rates of initiating regular smoking compared with survivors of other childhood cancers. It is important that hereditary retinoblastoma survivors do not smoke because they are known to have a high risk of a second malignant neoplasm ( 15 , 41 ), which is largely attributable to germline mutations they carry in the RB1 gene ( 42 ). Foster et al. ( 20 ) also reported levels of current smoking similar to those seen here: among 1-year survivors of heritable and nonheritable retinoblastoma with a mean age of 35 years, the majority of whom were diagnosed between 1960 and 1984, the prevalence of current smoking was 17% and 24%, respectively, both of which were statistically significantly lower than the population rates. Foster et al. ( 20 ) suggested that higher educational attainment in the retinoblastoma survivors might explain their lower smoking prevalence compared with that seen in the general population. Here, we found that among all childhood cancer survivors, those with higher educational attainment were less likely to be a smoker than those who failed to pass or to take any standard public educational examination. However, we cannot infer that a higher educational attainment in the heritable retinoblastoma survivors contributed to the lower smoking prevalence because in both the internal and external analyses we included educational attainment in the multivariable models and observed that the heritable retinoblastoma survivors still had one of the lowest prevalences of smoking. Therefore, other unidentified factors could also be contributing to the lower prevalence of smoking in the hereditary retinoblastoma survivors. For example, it is possible that because of their genetic predisposition for a second malignant neoplasm, the heritable retinoblastoma survivors may have received more intensive and focused counseling on lifestyle behaviors, which may have contributed to their lower prevalence of smoking.

Only the survivors of a CNS neoplasm had a lower rate of initiating smoking than survivors of a heritable retinoblastoma. Emmons et al. ( 19 ) also reported a reduced rate of smoking initiation in CNS neoplasm survivors in the CCSS, which they suggested could be a consequence of treatment-related neuropsychological deficits in the CNS neoplasm survivors. This suggestion seems plausible because CNS neoplasm survivors are particularly susceptible to neuropsychological late effects, which range from cognitive impairments to severe learning disabilities that lead to a high level of dependence on others ( 43 , 44 ); such deficits could exclude these survivors from situations where they might start smoking or limit their opportunity to smoke. Neuropsychological function is also associated with age at diagnosis, and survivors who are younger at diagnosis fare worse than those who are older at diagnosis ( 43 ). Neuropsychological deficits present in those diagnosed at a young age may have contributed somewhat to the lower prevalence of smoking among survivors in the BCCSS who were diagnosed before 10 years of age. In the CCSS, a cancer diagnosis at a young age (younger than 10 years) was also associated with a lower risk of smoking initiation than in those diagnosed at age 10 years or older ( 19 ).

Several reports ( 25 , 38 , 45 ) indicate that survivors of childhood cancer do not differ from their peers who were not diagnosed with a childhood cancer with respect to associations between sociodemographic factors and smoking initiation. The survivors in the BCCSS were also not different from the British general population with respect to the sociodemographic factors associated with being a current smoker ( 29 ). Higher rates of smoking have been observed in the general population and also in the BCCSS (albeit at a somewhat lower level of smoking in the latter) among those aged 20–34 years, men, individuals in routine and manual occupations, and those not currently married compared with the complementary groups (as detailed in Table 2 ) ( 29 ). Survivors who failed to pass or take any standard public educational examination were more likely to be a current regular smoker than those with some level of educational attainment. Other studies ( 20 , 24 ) have shown that survivors who were less educated were more likely to be a current smoker, and education has been shown to be associated with smoking initiation in the CCSS ( 19 ). Limited educational attainment has been associated with fatalistic beliefs about cancer prevention, and such fatalistic beliefs have been shown to be associated with smoking ( 46 ). For survivors with fatalistic beliefs, smoking interventions should emphasize the fact that individuals can have an appreciable impact on reducing their risk of a second malignant neoplasm by their lifestyle choices.

This study has several strengths. It is the largest ever such population-based cohort study, and it included all types of childhood cancer and had extended follow-up times for the survivors; 48% of the survivors have been followed up for at least 20 years after reaching 5 years of survival subsequent to their diagnosis ( 30 ).

Our study has three limitations that should be considered when interpreting the results. First, we included third party–completed questionnaires. Exclusion of the third-party questionnaires did not appreciably change the conclusions from the external analysis to when all questionnaires were analyzed. For the internal analysis, we included all survivors, regardless of whether their questionnaires were completed by the survivor or by a third party, and entered third-party completion as a factor in the multivariable model. We found that survivors whose questionnaires were completed by a third party had a lower rate of initiating smoking than those who self-completed a questionnaire. This association could have several possible explanations. On the one hand, the reduced rate of smoking initiation among survivors with third party–completed questionnaires could be linked to a physical or mental disability that prevented self-completion of the questionnaire. Survivors in the BCCSS who were diagnosed during childhood with heritable retinoblastoma or a CNS neoplasm had the highest proportion of third party–completed questionnaires (33.7% and 24.6%, respectively) ( 30 ). Alternatively, the lower rate of smoking initiation in survivors with a third party–completed questionnaire may reflect the fact that the person who completed the questionnaire did not have complete information about the survivor's use of cigarettes. However, we requested that all third party–completed questionnaires be completed in the presence of the survivor whenever possible, which should have led to improved accuracy in the smoking data reported, as has been suggested by others who have used proxy reports on smoking status ( 47 , 48 ).

A second possible limitation of this study is the accuracy of the self-reported smoking data. The smoking data for all survivors were based on responses to a questionnaire rather than on biochemical data, which could call into question their accuracy. However, a review of studies that used biochemical validation of self-reported smoking status found that although sensitivity and specificity of self-reporting varied by study, they were high overall (mean sensitivity and specificity = 87% and 89%, respectively) ( 49 ). The authors of the GHS report acknowledged that cigarette consumption and prevalence may have been underestimated ( 29 ), but the extent of any underestimation among the general population or the survivors in the BCCSS is unknown.

A third limitation is that this analysis did not include information relating to the 2780 of the 17 981 5-year survivors who died before we could ascertain their smoking status. It is possible that some of these deaths may have been related to the survivor's smoking behavior, and, consequently, that the survivors who died may have had a higher prevalence of smoking than reported here for those who were alive. However, given that 43% of the 2780 survivors died at age 16 years or younger, it is unlikely that the smoking behavior of these young individuals who died was the main cause of their death. The most common cause of death in survivors who die in the initial 10-year period following 5-year survival is a cancer recurrence ( 50 ). Cohort members who died from a cancer recurrence are likely to have had less of an opportunity to start smoking than those who did not experience a recurrence, which would have resulted in a lower prevalence of smoking in the entire 17 981-member cohort than what we observed among the survivors who provided their smoking status on the questionnaire.

Although the childhood cancer survivors had a substantially lower smoking prevalence than the general population, the smoking prevalence in this vulnerable group should be reduced even further. Advice on the health risks of smoking should be included in any program of clinical follow-up for survivors of childhood cancer. However, Taylor et al. ( 51 ) reported that 65% of adult survivors of childhood cancer in Britain were not on long-term regular hospital follow-up for their cancer. Clinical follow-up guidelines for children and adolescents in Britain ( www.cclg.org.uk and www.sign.ac.uk ) and in North America ( www.survivorshipguidelines.org ) recommend that survivors should not start smoking and that those who do should stop. The guidelines also provide information on the health risks of smoking in relation to specific cancer treatments, such as alkylating agents, and radiation to the lungs, heart, abdomen, or pelvis. We found that the BCCSS survivors who were on long-term regular hospital follow-up were less likely to be current smokers than those not on follow-up. It is encouraging that those attending clinical follow-up programs appear to have a lower level of smoking than those survivors not attending such a program; however, those on such programs are likely to have different attributes than those not on such programs ( 51 ).

In conclusion, although the extent of smoking among survivors of childhood cancer diagnosed in Britain between 1940 and 1991 was less than that seen in the comparable general population, levels of smoking prevalence among survivors should be reduced further, particularly among survivors of Wilms tumor, Hodgkin lymphoma, and soft tissue sarcoma. Results of this study and data from the CCSS ( 19 ) and the GHS for the general population of Britain ( 29 ) indicate that interventions to prevent initiation of smoking among survivors of childhood cancer need to start early in a survivor's life because the majority of individuals who start to smoke do so by the age of 20 years. A recent review revealed that few interventions have been designed primarily to prevent the initiation of smoking in survivors of childhood cancer ( 52 ). In a randomized controlled trial, Tyc et al. ( 53 ) reported that adolescent survivors who received a risk counseling intervention had statistically significantly better health knowledge scores, higher perceived vulnerability scores, and lower smoking intention scores than those who received standard care. Because factors that predict smoking initiation in childhood cancer survivors have been shown to be similar to those in healthy adolescents ( 25 , 38 ), Klosky et al. ( 45 ) have suggested that smoking prevention interventions that have been developed for healthy adolescents could be used for adolescent survivors of childhood cancer. For example, a well-designed family intervention or a multicomponent community program, which have been shown to assist smoking prevention in healthy adolescents ( 54 , 55 ), could potentially be used in adolescent survivors of cancer. Preventive interventions would not be appropriate for the BCCSS cohort because a large proportion of the participants are older than 20 years of age but should be undertaken among more recently diagnosed survivors of cancer, ideally those aged 12–20 years.

Smoking cessation interventions would be more appropriate for the BCCSS cohort than smoking prevention interventions because a high proportion of the survivors were older than the age at which most individuals initiate smoking (ie, ≤20 years of age). A telephone-based peer-delivered smoking cessation intervention has been used successfully with adult survivors of childhood cancer: those who received the intervention had a statistically significantly higher quit rate than those in a self-help group ( 56 ), and those who used nicotine replacement therapy had more quit attempts than those who did not use this therapy ( 57 ). Numerous smoking cessation methods, such as individual counseling and pharmacological treatments, have been used with success in the general population ( 58 ), although some, such as the antidepressant bupropion, are not advisable for survivors of some types of cancer [eg, those with a history of seizures or a CNS neoplasm ( 45 )]. Novel cessation intervention strategies, such as the nicotine vaccine that is currently being developed, may be an alternative way to help survivors of childhood cancer to stop smoking ( 45 ). Research is required to examine whether interventions that are used or designed to help healthy individuals to stop smoking ( 45 , 58 ) have the same effects on the smoking behavior of childhood cancer survivors. This study has highlighted the groups of childhood cancer survivors that are at higher risk of smoking and among whom successful interventions should be targeted.

Funding

The BCCSS was funded by Cancer Research UK and the Kay Kendall Leukemia Fund. The grant holder for the BCCSS is Professor M.M. Hawkins (grant numbers C386 and A7852). Cancer Research UK and the Kay Kendall Leukemia Fund did not have any role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication.

References

1.
International Agency for Research on Cancer (IARC) and World Health Organisation (WHO)
IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. Tobacco Smoke and Involuntary Smoking. Summary of Data Reported and Evaluation
 , 
2004
, vol. 
Vol 83
 
Lyon, France
IARC
2.
Action on Smoking and Heath (ASH)
Factsheet no. 6. Smoking, The Heart and Circulation
 , 
2007
UK
ASH. Available at
 
3.
Action on Smoking and Heath (ASH)
Factsheet no. 5. Smoking and Respiratory Disease
 , 
2007
UK
ASH
 
4.
British Medical Association (BMA)
Smoking and Reproductive Health
 , 
2007
UK
BMA
 
5.
Bhatia
S
Sklar
C
Second cancers in survivors of childhood cancer
Nature Reviews. Cancer
 , 
2002
, vol. 
2
 
2
(pg. 
124
-
131
)
6.
Hawkins
MM
Long-term survivors of childhood cancers: what knowledge have we gained?
Nat Clin Pract Oncol
 , 
2004
, vol. 
1
 
1
(pg. 
26
-
31
)
7.
Jenkinson
HC
Hawkins
MM
Stiller
CA
Winter
DL
Marsden
HB
Stevens
MCG
Long-term population-based risks of second malignant neoplasms after childhood cancer in Britain
Br J Cancer
 , 
2004
, vol. 
91
 
11
(pg. 
1905
-
1910
)
8.
Neglia
JP
Friedman
DL
Yasui
Y
, et al.  . 
Second malignant neoplasms in five-year survivors of childhood cancer: Childhood Cancer Survivor Study
J Natl Cancer Inst
 , 
2001
, vol. 
93
 
8
(pg. 
618
-
629
)
9.
Mertens
AC
Yasui
Y
Neglia
JP
, et al.  . 
Late mortality experience in five-year survivors of childhood and adolescent cancer: the Childhood Cancer Survivor Study
J Clin Oncol
 , 
2001
, vol. 
19
 
13
(pg. 
3163
-
3172
)
10.
Bowers
DC
Liu
Y
Leisenring
W
, et al.  . 
Late-occuring stroke among long-term survivors of childhood leukemia and brain tumors: a report from the Childhood Cancer Survivor Study
J Clin Oncol
 , 
2006
, vol. 
24
 
33
(pg. 
5277
-
5282
)
11.
Mertens
AC
Yasui
Y
Liu
Y
, et al.  . 
Pulmonary complications in survivors of childhood and adolescent cancer
Cancer
 , 
2002
, vol. 
95
 
11
(pg. 
2431
-
2441
)
12.
Critchley
HO
Thomson
AB
Wallace
WHB
Wallace
WHB
Green
DM
Ovarian and uterine function and reproductive potential
Late Effects of Childhood Cancer
 , 
2004
1st ed
London
Arnold
(pg. 
225
-
237
)
13.
Thomson
AB
Wallace
WHB
Sklar
CA
Wallace
WHB
Green
DM
Testicular function
Late Effects of Childhood Cancer
 , 
2004
1st ed
London
Arnold
(pg. 
239
-
255
)
14.
Draper
GJ
Sanders
BM
Kingston
JE
Second primary neoplasms in patients with retinoblastoma
Br J Cancer
 , 
1986
, vol. 
53
 (pg. 
661
-
671
)
15.
Wong
FL
Boice
JD
Jr
Abramson
DH
, et al.  . 
Cancer incidence after retinoblastoma. Radiation dose and sarcoma risk
JAMA
 , 
1997
, vol. 
278
 (pg. 
1262
-
1267
)
16.
Hisada
M
Garber
JE
Fung
CY
Fraumeni
JF
Jr
Li
FP
Multiple primary cancers in families with Li-Fraumeni syndrome
J Natl Cancer Inst
 , 
1998
, vol. 
90
 
8
(pg. 
606
-
611
)
17.
Meadows
AT
Baum
E
Fossati-Bellani
F
, et al.  . 
Second malignant neoplasms in children: an update from the Late Effects Study Group
J Clin Oncol
 , 
1985
, vol. 
3
 
4
(pg. 
532
-
538
)
18.
Kony
SJ
de Vathaire
F
Chompret
A
, et al.  . 
Radiation and genetic factors in the risk of second malignant neoplasms after a first cancer in childhood
Lancet
 , 
1997
, vol. 
350
 
9071
(pg. 
91
-
95
)
19.
Emmons
K
Li
FP
Whitton
J
, et al.  . 
Predictors of smoking initiation and cessation among childhood cancer survivors: a report from the Childhood Cancer Survivor Study
J Clin Oncol
 , 
2002
, vol. 
20
 
6
(pg. 
1608
-
1616
)
20.
Foster
MC
Kleinerman
RA
Abramson
DH
Seddon
JM
Tarone
RE
Tucker
MA
Tobacco use in adult long-term survivors of retinoblastoma
Cancer Epidemiol Biomarkers Prev
 , 
2006
, vol. 
15
 
8
(pg. 
1464
-
1468
)
21.
Bauld
C
Toumbourou
JW
Anderson
V
Coffey
C
Olsson
CA
Health-risk behaviours among adolescent survivors of childhood cancer
Pediatr Blood Cancer
 , 
2005
, vol. 
45
 
5
(pg. 
706
-
715
)
22.
Haupt
R
Byrne
J
Connelly
RR
, et al.  . 
Smoking habits in survivors of childhood and adolescent cancer
Med Pediatr Oncol
 , 
1992
, vol. 
20
 
4
(pg. 
301
-
306
)
23.
Larcombe
I
Mott
M
Hunt
L
Lifestyle behaviours of young adult survivors of childhood cancer
Br J Cancer
 , 
2002
, vol. 
87
 
11
(pg. 
1204
-
1209
)
24.
Tao
M
Guo
MD
Weiss
R
, et al.  . 
Smoking in adult survivors of childhood acute lymphoblastic leukaemia
J Natl Cancer Inst
 , 
1998
, vol. 
90
 
3
(pg. 
219
-
225
)
25.
Carswell
K
Chen
Y
Nair
RC
, et al.  . 
Smoking and binge drinking among Canadian survivors of childhood and adolescent cancers: a comparative, population-based study
Pediatr Blood Cancer
  
doi: 10.1002/pbc.21568
26.
Troyer
H
Holmes
GE
Cigarette smoking among childhood cancer survivors
Am J Dis Child
 , 
1988
, vol. 
142
 
2
pg. 
123
 
27.
Verrill
JR
Schafer
J
Vannatta
K
Noll
RB
Aggression, antisocial behaviour, and substance abuse in survivors of pediatric cancer: possible protective effects of cancer and its treatment
J Pediatr Psychol
 , 
2000
, vol. 
25
 
7
(pg. 
493
-
502
)
28.
Castellino
SM
Casillas
J
Hudson
MM
, et al.  . 
Minority adult survivors of childhood cancer: a comparison of long-term outcomes, health care utilization, and health-related behaviors from the Childhood Cancer Survivor Study
J Clin Oncol
 , 
2005
, vol. 
23
 
27
(pg. 
6499
-
6507
)
29.
Richards
L
Fox
K
Roberts
C
Fletcher
L
Goddard
E
Living in Britain. No31. Results from the 2002 General Household Survey
2004
London
Office for National Statistics, Her Majesty's Stationery Office
 
Report No. No31
30.
Hawkins
MM
Lancashire
ER
Winter
DL
, et al.  . 
The British Childhood Cancer Survivor Study: objectives, methods, population structure, response rates and initial descriptive information
Pediatr Blood Cancer
 , 
2008
, vol. 
50
 
5
(pg. 
1018
-
1025
)
31.
The Data Archive, University of Essex (GHS database from the internet)
 
Available at: http://www.data-archive.ac.uk/ Accessed May 15, 2007
32.
Denmark-Wahnefried
W
Werner
C
Clipp
EC
, et al.  . 
Survivors of childhood cancer and their guardians. Current health behaviours and receptivity to health promotion programs
Cancer
 , 
2005
, vol. 
103
 
10
(pg. 
2171
-
2180
)
33.
Mulhern
RK
Tyc
VL
Phipps
S
, et al.  . 
Health-related behaviors of survivors of childhood cancer
Med Pediatr Oncol
 , 
1995
, vol. 
25
 
3
(pg. 
159
-
165
)
34.
Taylor
AJ
Winter
DL
Stiller
CA
Murphy
M
Hawkins
MM
Risk of breast cancer in female survivors of childhood Hodgkin's disease in Britain: a population-based study
Int J Cancer
 , 
2007
, vol. 
120
 
2
(pg. 
384
-
391
)
35.
Cohen
RJ
Curtis
RE
Inskip
PD
Fraumeni
JFJ
The risk of developing second cancers among survivors of childhood soft tissue sarcoma
Cancer
 , 
2005
, vol. 
103
 
11
(pg. 
2391
-
2396
)
36.
Breslow
NE
Takashima
JR
Whitton
JA
Moksness
J
D’Angio
GJ
Green
DM
Second malignant neoplasms following treatment for Wilms’ tumor: a report from the National Wilms’ Tumor Study Group
J Clin Oncol
 , 
1995
, vol. 
13
 (pg. 
1851
-
1859
)
37.
Taylor
AJ
Winter
DL
Pritchard-Jones
K
, et al.  . 
Second Primary Neoplasms in Survivors of Wilms’ Tumour—a Population-based Cohort Study from the British Childhood Cancer Survivor Study
Int J Cancer
 , 
2008
, vol. 
122
 
9
(pg. 
2085
-
2093
)
38.
Ford
JS
Ostroff
JS
Health behaviours of childhood cancer survivors: what we’ve learned
J Clin Psychol Med Settings
 , 
2006
, vol. 
13
 
2
(pg. 
151
-
167
)
39.
Travis
LB
Gospodarowicz
M
Curtis
RE
, et al.  . 
Lung cancer following chemotherapy and radiotherapy for Hodgkin's disease
J Natl Cancer Inst
 , 
2002
, vol. 
94
 
3
(pg. 
182
-
192
)
40.
van Leeuwen
FE
Klokman
WJ
Stovall
M
, et al.  . 
Roles of radiotherapy and smoking in lung cancer following Hodgkin's disease
J Natl Cancer Inst
 , 
1995
, vol. 
87
 
20
(pg. 
1530
-
1537
)
41.
Hawkins
MM
Draper
GJ
Kingston
JE
Incidence of second primary tumours among childhood cancer survivors
Br J Cancer
 , 
1987
, vol. 
56
 
3
(pg. 
339
-
347
)
42.
Yandell
DW
Poremba
C
Genetics of retinoblastoma: implications for other human cancers
Med Pediatr Oncol Suppl
 , 
1996
, vol. 
1
 
S1
(pg. 
25
-
28
)
43.
Mulhern
RK
Phipps
S
White
H
Wallace
WHB
Green
DM
Neuropsychological outcomes
Late Effects of Childhood Cancer
 , 
2004
1st ed
London
Arnold
(pg. 
18
-
36
)
44.
Rourke
MT
Kazak
AE
Schwartz
CL
Hobbie
WL
Constine
LS
Ruccione
KS
Psychological aspects of long-term survivorship
Survivors of Childhood and Adolescent Cancer. A Multidisciplinary Approach
 , 
2005
2nd ed
Berlin
Springer
(pg. 
295
-
304
)
45.
Klosky
JL
Tyc
VL
Garces-Webb
DM
Buscemi
J
Klesges
RC
Hudson
MM
Emerging issues in smoking among adolescent and adult cancer survivors. A comprehensive review
Cancer
 , 
2007
, vol. 
110
 
11
(pg. 
2408
-
2419
)
46.
Niederdeppe
J
Levy
AG
Fatalistic beliefs about cancer prevention and three prevention behaviours
Cancer Epidemiol Biomarkers Prev
 , 
2007
, vol. 
16
 
5
(pg. 
998
-
1003
)
47.
Chen
Y
Rennie
DC
Dosman
JA
The reliability of cigarette consumption reports by spousal proxies
Am J Public Health
 , 
1995
, vol. 
85
 
12
(pg. 
1711
-
1712
)
48.
Passaro
KT
Noss
J
Savitz
DA
Little RE, and the ALSPAC Study Team
Agreement between self and partner reports of paternal drinking and smoking
Int J Epidemiol
 , 
1997
, vol. 
26
 
2
(pg. 
315
-
320
)
49.
Patrick
DL
Cheadle
A
Thompson
DC
Diehr
P
Koepsell
T
Kinne
S
The validity of self-reported smoking: a review and meta-analysis
Am J Public Health
 , 
1994
, vol. 
84
 
7
(pg. 
1086
-
1093
)
50.
Robertson
CM
Hawkins
MM
Kingston
JE
Late deaths and survival after childhood cancer: implications for cure
BMJ
 , 
1994
, vol. 
309
 
6948
(pg. 
162
-
166
)
51.
Taylor
A
Hawkins
MM
Griffiths
A
, et al.  . 
Long-term follow-up of survivors of childhood cancer in the UK
Pediatr Blood Cancer
 , 
2004
, vol. 
42
 
2
(pg. 
161
-
168
)
52.
Clarke
S
Eiser
C
Health behaviours in childhood cancer survivors: a systematic review
Eur J Cancer
 , 
2007
, vol. 
43
 (pg. 
1373
-
1384
)
53.
Tyc
VL
Rai
SN
Lensing
S
Klosky
JL
Stewart
DB
Gattuso
J
Intervention to reduce intentions to use tobacco among pediatric cancer survivors
J Clin Oncol
 , 
2003
, vol. 
21
 
7
(pg. 
1366
-
1372
)
54.
Sowden
A
Stead
L
Community interventions for preventing smoking in young people
Cochrane Database Syst Rev.
 , 
2000
, vol. 
1
  
CD001291. doi: 10.1002/14651858.CD001291
55.
Thomas
RE
Baker
P
Lorenzetti
D
Family-based programmes for preventing smoking by children and adolescents
Cochrane Database Syst Rev.
 , 
2007
, vol. 
1
  
CD004493. doi: 10.1002/14651858.CD004493.pub2
56.
Emmons
KM
Puleo
E
Park
E
, et al.  . 
Peer-delivered smoking counselling for childhood cancer survivors increases rate of cessation: the Partnership for Health Study
J Clin Oncol
 , 
2005
, vol. 
23
 
27
(pg. 
6516
-
6523
)
57.
Park
ER
Puleo
E
Butterfield
RM
, et al.  . 
A process evaluation of a telephone-based peer-delivered smoking cessation intervention for adult survivors of childhood cancer: the partnership for health study
Prev Med
 , 
2006
, vol. 
42
 
6
(pg. 
435
-
442
)
58.
Lancaster
T
Stead
L
Silagy
C
Sowden A. For the Cochrane Tobacco Addiction Review Group
Effectiveness of interventions to help people stop smoking: findings from the Cochrane Library
BMJ
 , 
2000
, vol. 
321
 
7257
(pg. 
355
-
358
)
The BCCSS is a national collaborative undertaking guided by a Steering Group that comprises Professor Douglas Easton (chair), Professor Michael Hawkins (secretary), Dr Helen Jenkinson, Dr Meriel Jenney, Dr Emma Lancashire, Professor Kathryn Pritchard-Jones, Professor Michael Stevens, Mr Charles Stiller, Dr Elaine Sugden, Dr Andrew Toogood, and Dr Hamish Wallace. The BCCSS benefits from the contributions of the Officers, Centers, and individual members of the Children's Cancer and Leukemia Group, the Childhood Cancer Research Group, and the Regional Pediatric Cancer Registries. The BCCSS acknowledges the collaboration of the Office for National Statistics, the General Register Office for Scotland, the National Health Service Central Registers, the regional cancer registries, health authorities, and area health boards for providing general practitioner names and addresses and the general practitioners nationwide who facilitated direct contact with survivors. We are particularly thankful to all survivors who completed a 40-page questionnaire.