Browse the corpus

ith none of the long-term conditions) at age 12 for average confounder values. The age coefficient estimates a linear change in physical activity (or sedentary time) over time, which is the same for all long-term conditions, and additional terms for each health condition reflect any difference in the baseline activity. A model was also fitted to assess if there were interactions between the long-term conditions to assess whether there was any additional effect associated with having multiple conditions. Finally, a model which allowed the change in physical activity over time to depend on health condition; so, for example, MVPA for an adolescent with asthma might decrease at a different rate between ages 12 and 16 than for an adolescent without these conidtions was run. As self-reported pubertal status is susceptible to error27 and pubertal status could be linked to obesity,28 pubertal status was considered in a secondary sensitivity analysis. All analyses were performed in Stata V.15.0.29

a different rate between ages 12 and 16 than for an adolescent without these conidtions was run. As self-reported pubertal status is susceptible to error27 and pubertal status could be linked to obesity,28 pubertal status was considered in a secondary sensitivity analysis. All analyses were performed in Stata V.15.0.29 Results Online supplementary table S1 summarises the characteristics of the imputed data, with summaries for observed data reported in online supplementary table S2. Missing data ranged from <1% (obesity) to 15% (eczema). Between 24% and 27% of adolescents had ever been diagnosed with asthma, between 21% and 27% had ever been diagnosed with eczema and between 13% and 15% were obese. Between 13% and 15% had more than one condition, with the most common combination being asthma and eczema (between 7% and 9% of adolescents). The combinations at each age are summarised graphically in online supplementary figure S1. The average minutes of MVPA per day decreased from 57 min at age 12 to 47 min at age 16, with average sedentary minutes increasing from 354 min to 475 min. Online supplementary figure S2 illustrates how MVPA and sedentary time changes with age among adolescents with the different conditions.

ne supplementary figure S1. The average minutes of MVPA per day decreased from 57 min at age 12 to 47 min at age 16, with average sedentary minutes increasing from 354 min to 475 min. Online supplementary figure S2 illustrates how MVPA and sedentary time changes with age among adolescents with the different conditions. Cross-sectional analyses Online supplementary table S3 summarises the difference in MVPA and sedentary time for adolescents with asthma, eczema, overweight and obese conditions, compared with those without. Tables 1 and 2 present the regression models for MVPA and sedentary time, respectively, adjusting for age, sex, mother’s education and household social class, and report the difference between an adolescent with and without each health condition. These models include all conditions, including combinations; differences are additive if an adolescent has multiple conditions. A minimally adjusted model (see online supplementary table S4), adjusting only for child age and sex, shows very little difference to the coefficients reported in the full model. Table 1 Cross-sectional regression models for MVPA at average ages 12, 14 and 16

Strengths and limitations of this study Objective accelerometer-based assessment of physical activity at ages 12, 14 and 16. Multiple assessments of the physical conditions asthma, eczema and obesity. Large representative population-based cohort allows longitudinal analysis of change over time. Asthma and eczema assessments do not exactly coincide with the physical activity data and so we use lifetime prevalence rather than current asthma/eczema indicators. We are unable to differentiate between mild and severe asthma or eczema—more severe cases tend to be more persistent and it may be that persistence is linked to activity. Introduction Obesity, eczema (synonyms: atopic eczema/dermatitis) and asthma are three of the most prevalent childhood long-term conditions in the UK. Data from the 2015/2016 National Child Measurement Programme showed that 9% of 4–5-year-olds and 20% of 10–11-year-olds in England were obese.1 It has been estimated that 20% of children are diagnosed with eczema2 and 9% with asthma,3 and children with eczema often have asthma. These long-term conditions place substantial burden on quality of life and healthcare expenditure. Understanding factors associated with the prevalence of these conditions is important to aid the development of prevention and management strategies.

czema2 and 9% with asthma,3 and children with eczema often have asthma. These long-term conditions place substantial burden on quality of life and healthcare expenditure. Understanding factors associated with the prevalence of these conditions is important to aid the development of prevention and management strategies. Physical activity is a modifiable lifestyle factor that has been associated with adiposity4 and asthma5 in children, and adults with asthma engage in lower levels of physical activity than those without.6 Low levels of physical activity reduce the quality of life in childhood and increase the risk of other comorbidities.7 Physical inactivity may be a problem among children with long-term conditions. There is some evidence that physical activity levels are lower among children with asthma than children without asthma.8 Children with eczema may be less active than children without eczema due to reluctance to sweat which can irritate eczematous skin, but the current evidence base is very weak.9 It may be that tailored strategies are needed to help those with asthma and eczema be more physically active, but the development of such approaches would only be warranted if there were evidence of lower physical activity.

ctance to sweat which can irritate eczematous skin, but the current evidence base is very weak.9 It may be that tailored strategies are needed to help those with asthma and eczema be more physically active, but the development of such approaches would only be warranted if there were evidence of lower physical activity. There is some evidence to show that physical activity levels are lower among children who are overweight or obese.10 While there have been many studies that have looked at the association between physical activity and obesity, these have tended to focus on cross-sectional surveys, or national surveys that have looked at associations over time but not within participants over time, for which cohort studies are essential. An analysis of the Avon Longitudinal Study of Parents and Children (ALSPAC) cohort reported a graded inverse association between physical activity and obesity at age 1211; an increase of 15 min of moderate to vigorous intensity physical activity (MVPA) per day was associated with 10% lower fat mass in girls and 12% lower in boys.12 However, the authors used an ALSPAC-specific definition of MVPA13 and reported an average MVPA of 20 min per day while the majority of current studies use the Evenson cut-points14 which a comparison study reported was the reliable threshold15 for young people. This difference is important as other studies that use Evenson cut-points typically report MVPA in the range of 50 to 70 min per day16–18 and, as such, it is important to have data from ALSPAC that have been processed using the Evenson cut-points to facilitate international comparisons.

le threshold15 for young people. This difference is important as other studies that use Evenson cut-points typically report MVPA in the range of 50 to 70 min per day16–18 and, as such, it is important to have data from ALSPAC that have been processed using the Evenson cut-points to facilitate international comparisons. The overall objective of this paper is to examine the physical activity of adolescents with three of the most common long-term conditions and if these associations change as young people move through adolescence, using objective assessments of physical activity. This evidence is needed to identify whether there are important differences between children with different long-term conditions. The specific aims were: (1) to compare mean time spent sedentary and engaged in MVPA for adolescents with asthma, eczema and obesity at ages 12, 14 and 16; and (2) to examine prospective associations between sedentary and MVPA time from age 12 to 16 for those with and without these conditions.

ent long-term conditions. The specific aims were: (1) to compare mean time spent sedentary and engaged in MVPA for adolescents with asthma, eczema and obesity at ages 12, 14 and 16; and (2) to examine prospective associations between sedentary and MVPA time from age 12 to 16 for those with and without these conditions. Methods ALSPAC is a birth cohort study based in the former county of Avon (UK).19–21 A total of 15 247 pregnant women with expected delivery dates between April 1991 and December 1992 were recruited, with 14 701 children alive at 1 year of age. Ethical approval for the study was obtained, and written informed consent was obtained for all mothers, with parents providing written informed consent for young people’s participation. Please note that the study website contains details of all the data that are available through a fully searchable data dictionary (http://www.bris.ac.uk/alspac/researchers/data-access/data-dictionary/).

nsent was obtained for all mothers, with parents providing written informed consent for young people’s participation. Please note that the study website contains details of all the data that are available through a fully searchable data dictionary (http://www.bris.ac.uk/alspac/researchers/data-access/data-dictionary/). Accelerometer data Adolescents wore an Actigraph AM7164 (Actigraph LLC, Fort Walton Beach, Florida, USA) accelerometer on their hip for 7 consecutive days. The first accelerometer measurement was taken at an average age of 11 years and 9 months in 2003–2004, with subsequent measurements at ages 13 years and 10 months, and at 15 years and 6 months. Accelerometer data were processed using Kinesoft (V.3.3.75; Kinesoft, Saskatchewan, Canada) and analysis was restricted to those who provided at least 3 days of valid data. A valid day was defined as at least 500 min of data, after excluding intervals of ≥60 min of zero counts allowing up to 2 min of interruptions. The average number of MVPA (≥2296 counts per minute (CPM)) and sedentary minutes per day (<100 CPM) were derived for each adolescent based on the Evenson thresholds.14 Accelerometer data with over 11715 CPM, fewer than 10 sedentary minutes or zero MVPA were considered erroneous and excluded from the analysis (16, 41 and 21 respondents at ages 12, 14 and 16, respectively).22

)) and sedentary minutes per day (<100 CPM) were derived for each adolescent based on the Evenson thresholds.14 Accelerometer data with over 11715 CPM, fewer than 10 sedentary minutes or zero MVPA were considered erroneous and excluded from the analysis (16, 41 and 21 respondents at ages 12, 14 and 16, respectively).22 Exposures Height and weight were measured at the same time as the accelerometer data in a clinic setting by trained fieldworkers. Body mass index (BMI) was used to create age-specific and sex-specific overweight and obesity indicators using 85th and 95th percentiles based on UK BMI reference charts.23 Throughout this paper, the overweight category comprises those participants who were overweight but not the participants who were obese, who are defined as obese throughout. The mother was asked ‘Has a doctor ever told you that your child has asthma?’ at ages 8, 11 and 14, and ‘Has a doctor ever told you that your child has eczema?’ at ages 11 and 14. These were combined to form lifetime doctor-diagnosed asthma and eczema prevalence: ‘has asthma ever been reported as diagnosed by a doctor?’ and ‘has eczema ever been reported as diagnosed by a doctor?’, assessed by ages 11 and 14. As these ages do not correspond to accelerometer collection, current asthma/eczema assessments were unavailable.

rm lifetime doctor-diagnosed asthma and eczema prevalence: ‘has asthma ever been reported as diagnosed by a doctor?’ and ‘has eczema ever been reported as diagnosed by a doctor?’, assessed by ages 11 and 14. As these ages do not correspond to accelerometer collection, current asthma/eczema assessments were unavailable. Confounders The child’s sex as recorded at birth and the age of the child at the accelerometer data collection were included as confounders. The mother’s highest education variable was created by combining from data at 32 w gestation and age 8 to provide the highest education recorded in the data set. Household social class was calculated as the highest occupational social class reported by mother or partner from data at 18 weeks’ gestation (both) and ages 4 (mother) and 8 (partner). Puberty was assessed from yearly questionnaires. For girls, an indicator of onset of menarche was used. For boys, Tanner stages based on pubic hair was used, as self-report data on Tanner stages for genitals has been found to be unreliable.24 Patient and public involvement The concept for this paper was informed by previous research with children with asthma and their parents who expressed challenges in relation to being physically active.25 As this was a secondary analysis of data that were collected 10 years ago, participants were not involved in the design of the specific study or study recruitment which were conducted as part of the ALSPAC cohort. Results will be shared with ALSPAC participants via the study website.

ation to being physically active.25 As this was a secondary analysis of data that were collected 10 years ago, participants were not involved in the design of the specific study or study recruitment which were conducted as part of the ALSPAC cohort. Results will be shared with ALSPAC participants via the study website. Statistical analysis Cross-sectional data were analysed at three timepoints corresponding to accelerometer data at average ages of 12, 14 and 16. We included any child with at least 3 valid days of accelerometer data. Obesity and overweight were calculated at the same timepoints. Asthma and eczema diagnoses did not correspond to the same timepoints, and so were taken from the closest prior measurement, to record lifetime diagnosis up to the point of the accelerometer measurement. Mother’s highest education and household social class were calculated as described above and the same value used for all analyses. Cross-sectional analysis was based on 5735, 4078 and 2198 cases for ages 12, 14 and 16, respectively. For the longitudinal analyses, we combined data from all three timepoints, resulting in 6473 cases with data for at least one timepoint, and 1619 (25%) at all three.

ibed above and the same value used for all analyses. Cross-sectional analysis was based on 5735, 4078 and 2198 cases for ages 12, 14 and 16, respectively. For the longitudinal analyses, we combined data from all three timepoints, resulting in 6473 cases with data for at least one timepoint, and 1619 (25%) at all three. Multiple imputation was used to account for missing data. For the cross-sectional analyses, data were imputed for each data set separately. For the longitudinal analysis, data were imputed if the child had accelerometer data for that timepoint, but not between timepoints, although data from other timepoints was used to impute missing data. Multiple imputation methods were used to create 20 imputed data sets at each timepoint for boys and girls separately, using 20 cycles of regression switching and combined regression coefficients across data sets using Rubin’s rules.26 All exposures, outcomes and confounders were included.

s used to impute missing data. Multiple imputation methods were used to create 20 imputed data sets at each timepoint for boys and girls separately, using 20 cycles of regression switching and combined regression coefficients across data sets using Rubin’s rules.26 All exposures, outcomes and confounders were included. Cross-sectional analyses For each age, characteristics were described separately for all adolescents, those with asthma, eczema, who were overweight or obese, and for adolescents with none of the conditions. Average minutes of MVPA and sedentary time were compared between those with long-term conditions and those without using t-tests. Regression models were fitted for MVPA and sedentary minutes for all adolescents, and for boys and girls separately. Models include the asthma, eczema, overweight and obese indicators, and child age, sex, mother’s education and household social class as confounders. For comparative purposes, minimally adjusted models which adjusted for just child age and sex were also run and are shown in online supplementary table. 10.1136/bmjopen-2018-024858.supp1Supplementary file 1

Cross-sectional analyses For each age, characteristics were described separately for all adolescents, those with asthma, eczema, who were overweight or obese, and for adolescents with none of the conditions. Average minutes of MVPA and sedentary time were compared between those with long-term conditions and those without using t-tests. Regression models were fitted for MVPA and sedentary minutes for all adolescents, and for boys and girls separately. Models include the asthma, eczema, overweight and obese indicators, and child age, sex, mother’s education and household social class as confounders. For comparative purposes, minimally adjusted models which adjusted for just child age and sex were also run and are shown in online supplementary table. 10.1136/bmjopen-2018-024858.supp1Supplementary file 1 Longitudinal analysis Random effect models were fitted to identify changes in physical activity measures over time between the long-term conditions, accounting for individual variability in measurements. As the cross-sectional analyses suggest different patterns among boys and girls, all models were run separately by sex. We included indicators for asthma, eczema, obesity and overweight, and age at time of clinic to capture linear changes over time. All models were adjusted for mother’s highest education and household social class. In this model, the constant term captures the average physical activity of a healthy child (ie, with none of the long-term conditions) at age 12 for average confounder values. The age coefficient estimates a linear change in physical activity (or sedentary time) over time, which is the same for all long-term conditions, and additional terms for each health condition reflect any difference in the baseline activity.

Cross-sectional analyses Online supplementary table S3 summarises the difference in MVPA and sedentary time for adolescents with asthma, eczema, overweight and obese conditions, compared with those without. Tables 1 and 2 present the regression models for MVPA and sedentary time, respectively, adjusting for age, sex, mother’s education and household social class, and report the difference between an adolescent with and without each health condition. These models include all conditions, including combinations; differences are additive if an adolescent has multiple conditions. A minimally adjusted model (see online supplementary table S4), adjusting only for child age and sex, shows very little difference to the coefficients reported in the full model. Table 1 Cross-sectional regression models for MVPA at average ages 12, 14 and 16 Age 12 (n=5735) Age 14 (n=4078) Age 16 (n=2198) Diff* (min) 95 % CI P value Diff* (min) 95 % CI P value Diff* (min) 95 % CI P value All† Asthma −1.47 (−3.19 to 0.25) 0.094 −0.05 (−2.05 to 1.95) 0.960 −2.62 (−5.16 to 0.09) 0.043 Eczema −1.85 (−3.68 to 0.01) 0.049 −0.62 (−2.69 to 1.44) 0.554 0.74 (−1.75 to 3.24) 0.559 Overweight −7.54 (−9.59 to 5.48) <0.0005 −3.12 (−5.53 to 0.72) 0.011 −2.27 (−5.55 to 1.01) 0.175 Obese −8.94 (−10.92 to 6.96) <0.0005 −7.45 (−9.87 to 5.04) <0.0005 −5.29 (−8.50 to 2.07) 0.001 Boys† Asthma −2.03 (−4.76 to 0.70) 0.144 0.47 (−2.71 to 3.65) 0.773 −3.22 (−7.42 to 0.98) 0.133 Eczema −3.52 (−6.70 to 0.34) 0.030 0.05 (−3.51 to 3.61) 0.977 2.28 (−2.16 to 6.71) 0.314 Overweight −11.39 (−14.83 to 7.95) <0.0005 −4.47 (−8.54 to 0.40) 0.031 −5.22 (−10.95 to 0.50) 0.074 Obese −12.52 (−15.77 to 9.26) <0.0005 −10.24 (−14.13 to 6.35) <0.0005 −9.42 (−14.99 to 3.85) 0.001 Girls† Asthma −1.01 (−3.09 to 1.06) 0.339 −0.71 (−3.16 to 1.75) 0.573 −2.48 (−5.52 to 0.56) 0.110 Eczema −0.49 (−2.52 to 1.55) 0.639 −1.17 (−3.75 to 1.41) 0.372 −0.34 (−3.16 to 2.49) 0.816 Overweight −4.07 (−6.44 to 1.70) 0.001 −1.95 (−4.79 to 0.88) 0.177 0.19 (−3.59 to 3.96) 0.923 Obese −5.39 (−7.70 to 3.07) <0.0005 −4.52 (−7.49 to 1.55) 0.003 −1.88 (−5.59 to 1.84) 0.321 *Difference in minutes of MVPA between a child with and without the condition, with all other variables the same.

−3.16 to 2.49) 0.816 Overweight −4.07 (−6.44 to 1.70) 0.001 −1.95 (−4.79 to 0.88) 0.177 0.19 (−3.59 to 3.96) 0.923 Obese −5.39 (−7.70 to 3.07) <0.0005 −4.52 (−7.49 to 1.55) 0.003 −1.88 (−5.59 to 1.84) 0.321 *Difference in minutes of MVPA between a child with and without the condition, with all other variables the same. †Models include all conditions, and are adjusted for child age, sex, mother’s education and household social class as confounders. MVPA, moderate to vigorous intensity physical activity. Table 2 Cross-sectional regression models for sedentary time at average ages 12, 14 and 16 Age 12 (n=5735) Age 14 (n=4078) Age 16 (n=2198) Diff* (min) 95 % CI P value Diff* (min) 95 % CI P value Diff* (min) 95 % CI P value All†

†Models include all conditions, and are adjusted for child age, sex, mother’s education and household social class as confounders. MVPA, moderate to vigorous intensity physical activity. Table 2 Cross-sectional regression models for sedentary time at average ages 12, 14 and 16 Age 12 (n=5735) Age 14 (n=4078) Age 16 (n=2198) Diff* (min) 95 % CI P value Diff* (min) 95 % CI P value Diff* (min) 95 % CI P value All† Asthma −5.14 (−9.80 to 0.49) 0.030 −10.77 (−16.90 to 4.65) 0.001 0.89 (−7.30 to9.09) 0.831 Eczema 0.76 (−4.25 to 5.78) 0.765 4.02 (−2.55 to 10.59) 0.230 0.95 (−7.09 to 9.00) 0.816 Overweight 13.82 (8.334 to 19.30) <0.0005 13.53 (6.17 to 20.88) <0.0005 8.63 (−1.96 to 19.22) 0.110 Obese 14.86 (9.59 to 20.13) <0.0005 22.06 (14.68 to 29.43) <0.0005 14.11 (3.75 to 24.48) 0.008 Boys† Asthma −5.86 (−12.33 to 0.60) 0.075 −8.17 (−17.15 to 0.81) 0.074 3.47 (−9.06 to 16.01) 0.587 Eczema 2.64 (−4.63 to 9.90) 0.476 2.80 (−7.19 to 12.80) 0.582 −2.30 (−15.58 to 10.98) 0.734 Overweight 15.07 (7.03 to 23.12) <0.0005 9.49 (−1.78 to 20.75) 0.099 16.70 (−0.21 to 33.62) 0.053 Obese 13.44 (5.82 to 21.05) 0.001 29.28 (18.49 to 40.06) <0.0005 21.78 (5.32 to 38.25) 0.010 Girls† Asthma −4.41 (−11.00 to 2.19) 0.190 −13.26 (−21.81 to 4.72) 0.002 −0.69 (−11.54 to 10.17) 0.901 Eczema −0.85 (−7.67 to 5.96) 0.806 4.55 (−4.29 to 13.40) 0.312 3.33 (−6.97 to 13.63) 0.526 Overweight 12.69 (5.19 to 20.19) 0.001 17.43 (7.74 to 27.12) <0.0005 1.87 (−11.66 to 15.41) 0.782 Obese 15.89 (8.55 to 23.23) <0.0005 14.44 (4.30 to 24.58) 0.005 7.63 (−5.70 to 20.96) 0.262 *Difference in minutes of sedentary time between a child with and without the condition.

13.40) 0.312 3.33 (−6.97 to 13.63) 0.526 Overweight 12.69 (5.19 to 20.19) 0.001 17.43 (7.74 to 27.12) <0.0005 1.87 (−11.66 to 15.41) 0.782 Obese 15.89 (8.55 to 23.23) <0.0005 14.44 (4.30 to 24.58) 0.005 7.63 (−5.70 to 20.96) 0.262 *Difference in minutes of sedentary time between a child with and without the condition. †Models include all conditions, and are adjusted for child age, sex, mother’s education and household social class as confounders.

13.40) 0.312 3.33 (−6.97 to 13.63) 0.526 Overweight 12.69 (5.19 to 20.19) 0.001 17.43 (7.74 to 27.12) <0.0005 1.87 (−11.66 to 15.41) 0.782 Obese 15.89 (8.55 to 23.23) <0.0005 14.44 (4.30 to 24.58) 0.005 7.63 (−5.70 to 20.96) 0.262 *Difference in minutes of sedentary time between a child with and without the condition. †Models include all conditions, and are adjusted for child age, sex, mother’s education and household social class as confounders. The largest and most consistent differences are for adolescents who are overweight or obese. MVPA was 2.3–7.5 min lower among overweight adolescents, and 5.3–8.9 min lower among obese adolescents. Similarly, sedentary time was 8.6–13.8 min higher among overweight adolescents, and 14.1–22.1 min higher among obese adolescents. These associations tend to be slightly stronger in boys than girls. Patterns for adolescents with asthma and eczema are weaker and less consistent. Associations between MVPA and asthma were inconclusive, with a lower MVPA (2.6 min) among asthma sufferers at age 16, but no other ages, and no differences when examining boys and girl separately. There was some association between asthma and sedentary time, with lower sedentary time at ages 12 and 14 (by 5.1 min and 10.8 min, respectively), although evidence is inconclusive when considering boys and girls separately. There was very little association between eczema and either MVPA or sedentary time. We saw a slightly lower MVPA (3.5 min) among boys at age 12, but no other ages, and no associations with sedentary time. It was not possible to explore associations with combinations of long-term conditions due to small numbers (<100) for some combinations. However, there was no evidence of any additional association with having multiple conditions. So, for example, the difference between obese children with asthma and obese children without asthma is the same as the difference between non-obese children with asthma and non-obese children without.

) for some combinations. However, there was no evidence of any additional association with having multiple conditions. So, for example, the difference between obese children with asthma and obese children without asthma is the same as the difference between non-obese children with asthma and non-obese children without. Longitudinal analysis Table 3 summarises the longitudinal models for MVPA and sedentary time, for boys and girls separately. The baseline represents a typical child at age 12 with the most common mother’s highest education (O level—examinations at age 16), and household class (Managerial & Technical). These coefficients are additive, so, for example, the average number of minutes of MVPA for a 12-year-old obese boy is estimated as 69.7–11.1=58.6 min. Figure 1 plots the estimated average MVPA time and sedentary time for the different long-term conditions by age, for boys and girls. Figure 1 Fitted models for average minutes of moderate to vigorous intensity physical activity (MVPA) (left) and sedentary time (right) for different long-term conditions and by sex. Table 3 Comparison of longitudinal models for average minutes of MVPA and sedentary time per day

Longitudinal analysis Table 3 summarises the longitudinal models for MVPA and sedentary time, for boys and girls separately. The baseline represents a typical child at age 12 with the most common mother’s highest education (O level—examinations at age 16), and household class (Managerial & Technical). These coefficients are additive, so, for example, the average number of minutes of MVPA for a 12-year-old obese boy is estimated as 69.7–11.1=58.6 min. Figure 1 plots the estimated average MVPA time and sedentary time for the different long-term conditions by age, for boys and girls. Figure 1 Fitted models for average minutes of moderate to vigorous intensity physical activity (MVPA) (left) and sedentary time (right) for different long-term conditions and by sex. Table 3 Comparison of longitudinal models for average minutes of MVPA and sedentary time per day Boys (2510 boys, 4847 measurements) Girls (2744 girls, 5481 measurements) Estimate 95% CI P value Estimate 95% CI P value Average MVPA minutes per day Baseline* (min) 69.68 (67.64 to 71.71) <0.0005 47.52 (46.13 to 48.90) <0.0005 Change with age (min/year) −3.09 (−3.578, to 2.61) <0.0005 −1.78 (−2.10, to 1.45) <0.0005 Asthma† (min) −1.24 (−3.35 to 0.87) 0.249 −1.19 (−2.81 to 0.43) 0.150 Eczema† (min) −1.22 (−3.62 to 1.17) 0.316 −0.87 (−2.52 to 0.79) 0.305 Overweight† (min) −8.05 (−10.46, to 5.63) <0.0005 −2.33 (−3.99, to 0.68) 0.006 Obese† (min) −11.12 (−13.57, to 8.68) <0.0005 −5.01 (−6.78, to 3.25) <0.0005 Average sedentary minutes per day Baseline* (min) 346.64 (341.19 to 352.08) <0.0005 365.73 (31.05 to 370.41) <0.0005 Change with age (min/year) 28.10 (26.94 to 29.26) <0.0005 31.97 (30.92 to 33.03) <0.0005 Asthma† (min) −4.06 (−9.84 to 1.73) 0.169 −6.45 (−11.92 to 0.98) 0.021 Eczema† (min) 2.16 (−4.17 to 8.49) 0.504 1.98 (−3.54 to 7.50) 0.482 Overweight† (min) 14.51 (8.38 to 20.63) <0.0005 12.63 (7.14 to 18.12) <0.0005 Obese† (min) 17.37 (10.97 to 23.78) <0.0005 15.19 (9.28 to 21.09) <0.0005 *Baseline represents the average number of minutes for a healthy individual at age 11 with mother’s highest education of O level (examination at age 16) and household social class of Managerial & Technical (II).

0005 12.63 (7.14 to 18.12) <0.0005 Obese† (min) 17.37 (10.97 to 23.78) <0.0005 15.19 (9.28 to 21.09) <0.0005 *Baseline represents the average number of minutes for a healthy individual at age 11 with mother’s highest education of O level (examination at age 16) and household social class of Managerial & Technical (II). †Change from baseline if child has this health condition. MVPA, moderate to vigorous intensity physical activity. The largest association was between physical activity and obesity, especially among boys. Asthma and eczema showed much weaker associations. A typical boy with no long-term conditions engaged in an average of 72.8 min of MVPA at age 12, declining at a rate of 3.1 min/year with age. They spent an average of 318.5 min in sedentary time at age 12, increasing at a rate of 28.1 min/year. Girls’ average MVPA was lower at 49.3 min at age 12, declining at a rate of 1.8 min/year, while sedentary time was higher, with an average of 333.8 min at age 12, increasing at a rate of 32.0 min/year. We found no association between asthma and either MVPA or sedentary time for boys. Among girls, there was no association between MVPA and asthma, but asthma sufferers spent 6.5 fewer minutes in sedentary time than non-sufferers. We found no association between eczema and either MVPA or sedentary time for boys or girls.

We found no association between asthma and either MVPA or sedentary time for boys. Among girls, there was no association between MVPA and asthma, but asthma sufferers spent 6.5 fewer minutes in sedentary time than non-sufferers. We found no association between eczema and either MVPA or sedentary time for boys or girls. Overweight and obese adolescents had lower MVPA and higher sedentary time. Associations are stronger for boys than for girls, and stronger for obesity than overweight. Being overweight is associated with a decrease in MVPA of 8.1 min and an increase in sedentary time of 14.5 min for boys, compared with a decrease of 2.3 min in MVPA and an increase of 12.6 min in sedentary time for girls. Obesity is associated with a decrease in MVPA of 11.1 min and an increase in sedentary time of 17.4 min among boys, and a decrease in MVPA of 5.0 min and an increase in sedentary time of 15.2 min among girls. There was no evidence of any additional effect associated with having combinations of multiple long-term conditions, via interaction terms. This suggests that any associations between physical activity and long-term conditions may be additive, although low prevalence of multiple conditions in this sample mean that we lacked statistical power to explore this fully. There was no evidence that the rate of change in physical activity over time differed between health condition. Including puberty in the longitudinal model (results not shown) reduced the strength of association with obesity and being overweight slightly but had no effect on associations with asthma or eczema.

s fully. There was no evidence that the rate of change in physical activity over time differed between health condition. Including puberty in the longitudinal model (results not shown) reduced the strength of association with obesity and being overweight slightly but had no effect on associations with asthma or eczema. Discussion In this paper, we found no strong evidence of differences in the mean minutes of MPVA or sedentary time per day between adolescents with asthma or eczema and those without these conditions in the cross-sectional models. We found a small association between asthma and fewer minutes of sedentary time among girls, but no associations for MVPA or eczema. Results suggest that adolescents who are overweight or obese engage in lower levels of physical activity at ages 12, 14 and 16, with stronger associations for boys than girls. These findings were confirmed in the longitudinal models which showed lower levels of MVPA and higher sedentary time with increasing levels of obesity. Specifically, obese boys engage in 11.1 fewer minutes of MVPA and 17.4 more sedentary minutes compared with a boy of similar age and demographics, while obese girls engage in 5.0 fewer minutes of MVPA and 15.2 more sedentary minutes than their non-obese counterparts. This finding is consistent with a previous Mendelian randomisation analysis which suggested a likely bidirectional association between physical activity and adiposity among young people.30

ographics, while obese girls engage in 5.0 fewer minutes of MVPA and 15.2 more sedentary minutes than their non-obese counterparts. This finding is consistent with a previous Mendelian randomisation analysis which suggested a likely bidirectional association between physical activity and adiposity among young people.30 The lack of an association between physical activity, sedentary time and asthma challenges a body of research which has reported that children with asthma are less active than their peers. However, previous literature has relied on relatively small cross-sectional studies using self-reported assessments of physical activity and asthma.8 31–33 The current analysis uses a population-based rather than clinical cohort and therefore contains a lower proportion of children with moderate to severe asthma. Also, because the study combines data measured at different times, we report lifetime asthma prevalence rather than current asthma. Therefore, caution should be exercised in any direct comparison with other studies. Low physical activity and fitness have been suggested as risk factors for asthma onset34 and exercise has been shown to improve markers of asthma control,35 which appear inconsistent with our findings. Concerns that exercise-induced bronchoconstriction might limit physical activity in children with controlled asthma appear to be ill-founded.36 Our results therefore provide reassurance that adolescents with asthma in a general population take part in similar levels of physical activity to their peers and there is no need for condition-specific behaviour change programmes.

t limit physical activity in children with controlled asthma appear to be ill-founded.36 Our results therefore provide reassurance that adolescents with asthma in a general population take part in similar levels of physical activity to their peers and there is no need for condition-specific behaviour change programmes. A previous Mendelian randomisation study from the ALSPAC cohort at age 7 identified obesity as a risk factor for asthma.37 This is consistent with our results. We observed a higher prevalence of obesity among asthma sufferers (20%, 16% and 17% at ages 12, 14 and 16) than those without asthma (14%, 12% and 12%, respectively). Furthermore, obesity with asthma has been implicated as a risk for more frequent asthma exacerbations,38 condition severity and a reduced response to therapy.39 Some small studies have suggested that weight loss may aid asthma management in children and adults.40–42 These studies and the consistent reporting of associations between asthma, obesity prevalence and condition morbidity supported by this study suggest that larger and longer weight management trials in obese individuals with asthma are warranted.

ed that weight loss may aid asthma management in children and adults.40–42 These studies and the consistent reporting of associations between asthma, obesity prevalence and condition morbidity supported by this study suggest that larger and longer weight management trials in obese individuals with asthma are warranted. Previous evidence of any difference in the physical activity patterns of children with and without eczema was unclear.9 There is some evidence of an association in adults,43 and possibly an association with severe eczema in children.44 There may also be a link between eczema and obesity in North America and Asia, although there is no evidence of an association in Europe.45 Our analyses found no differences in accelerometer-measured physical activity for those with lifetime-diagnosed eczema of any severity, which is encouraging given the psychosocial impact of even mild eczema. While parents may restrict activities such as swimming because of concerns about the effects on the child’s skin,46 our findings suggest that the presence of eczema does not overall negatively impact physical activity in this group and therefore there is no need for eczema-specific physical activity programmes.

eczema. While parents may restrict activities such as swimming because of concerns about the effects on the child’s skin,46 our findings suggest that the presence of eczema does not overall negatively impact physical activity in this group and therefore there is no need for eczema-specific physical activity programmes. Strengths and limitations The main strength of this study is the objective assessment of physical activity at ages 12, 14 and 16 and multiple assessments of physical conditions on a large cohort of adolescents. Although only 25% participants had data at all three timepoints, 61% had data for at least two, making the longitudinal analysis reasonably robust, although the study was not powered to test for interactions between conditions. We used multiple imputation to minimise the amount of missing data but recognise that there are limitations of this approach, including the underlying assumption that data are missing at random.47 48

longitudinal analysis reasonably robust, although the study was not powered to test for interactions between conditions. We used multiple imputation to minimise the amount of missing data but recognise that there are limitations of this approach, including the underlying assumption that data are missing at random.47 48 The study combines data measured at different times. Specifically, the asthma and eczema assessments do not exactly coincide with the physical activity data and so we were unable to include current asthma/eczema indicators. We also have no asthma/eczema information between 11 and 14 and so cannot determine, for example, if asthma precedes obesity. We have not adjusted for smoking we may have impacted on other associations. It is also important to highlight that due to the small number of cases, we were unable to model the interactions between health conditions. Finally, we are unable to differentiate between mild and severe asthma or eczema—more severe cases tend to be more persistent and it may be that persistence is linked to activity. The data were collected approximately a decade ago and it is possible that physical activity patterns and the prevalence of physical conditions, especially obesity, may have changed over this period.

vere asthma or eczema—more severe cases tend to be more persistent and it may be that persistence is linked to activity. The data were collected approximately a decade ago and it is possible that physical activity patterns and the prevalence of physical conditions, especially obesity, may have changed over this period. Conclusions Analysis of the ALSPAC population-based cohort has shown that physical activity and sedentary time did not differ for adolescents with asthma or eczema and those without, but obese adolescents engaged in fewer minutes of MVPA and more sedentary time. Findings reinforce the need for strategies to help obese adolescents be more active but suggest no need to develop bespoke physical activity strategies for adolescents with mild asthma or eczema. Supplementary Material Reviewer comments Author's manuscript We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. Contributors: Conception/design, revising and final approval: all authors. Methodology: RJ, RES, MJR, AJH, ARN, JPHS. Data analysis: RES. Drafting: RJ, RES.

Supplementary Material Reviewer comments Author's manuscript We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. Contributors: Conception/design, revising and final approval: all authors. Methodology: RJ, RES, MJR, AJH, ARN, JPHS. Data analysis: RES. Drafting: RJ, RES. Funding: The UK Medical Research Council and Wellcome (Grant ref: 102215/2/13/2) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and AJH, ARN, RES and RJ will serve as guarantors for the contents of this paper. A comprehensive list of grants funding is available on the ALSPAC website. Physical activity data collection was supported by the US National Heart, Lung, and Blood Institute (R01 HL071248-01A). RJ is partly supported by the National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health Research and Care West (CLAHRC West) at University Hospitals Bristol NHS Foundation Trust. The NIHR Biomedical Research Centre at University Hospitals Bristol NHS Foundation Trust and the University of Bristol supports ARN and JPH-S. MJR is funded by a National Institute for Health Research Post-Doctoral Research Fellowship (PDF-2014-07-013). Competing interests: None declared. Patient consent: Parental/guardian consent obtained.

Funding: The UK Medical Research Council and Wellcome (Grant ref: 102215/2/13/2) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and AJH, ARN, RES and RJ will serve as guarantors for the contents of this paper. A comprehensive list of grants funding is available on the ALSPAC website. Physical activity data collection was supported by the US National Heart, Lung, and Blood Institute (R01 HL071248-01A). RJ is partly supported by the National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health Research and Care West (CLAHRC West) at University Hospitals Bristol NHS Foundation Trust. The NIHR Biomedical Research Centre at University Hospitals Bristol NHS Foundation Trust and the University of Bristol supports ARN and JPH-S. MJR is funded by a National Institute for Health Research Post-Doctoral Research Fellowship (PDF-2014-07-013). Competing interests: None declared. Patient consent: Parental/guardian consent obtained. Ethics approval: ALSPAC Ethics and Law Committee and the Local Research Ethics Committees. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: This is a secondary data analysis based on data from the ALSPAC cohort. The access policy for the ALSPAC data can be found at http://www.bristol.ac.uk/media-library/sites/alspac/documents/researchers/data-access/ALSPAC_Access_Policy.pdf.

Strengths and limitations of this study This study used primary care data from the Clinical Practice Research Datalink (CPRD) which are representative of the UK primary care population. Expert-reviewed codelists were developed to define both the exposure and the outcome which would reduce misclassification bias. Due to the small sample size of people with general practitioner-recorded dementia, we were not able to test the relative effectiveness of varenicline versus nicotine replacement therapy (NRT) on smoking cessation using regression models. Data on smokers who purchase over-the-counter NRT were not available, and therefore the prevalence of NRT might be underestimated. Introduction Smoking is a leading cause of mortality and morbidity worldwide. About 12% of global deaths were linked to smoking in 2015.1 There is substantial evidence that smoking is associated with an increased risk of developing dementia.2 3 For instance, it is estimated that 14% of Alzheimer’s disease (AD) cases worldwide are attributable to smoking.4 Smoking is thought to accelerate the onset of dementia mainly via vascular risk factors such as narrowing of blood vessels in the heart and the brain, thereby triggering oxidative stress.4 5

ing dementia.2 3 For instance, it is estimated that 14% of Alzheimer’s disease (AD) cases worldwide are attributable to smoking.4 Smoking is thought to accelerate the onset of dementia mainly via vascular risk factors such as narrowing of blood vessels in the heart and the brain, thereby triggering oxidative stress.4 5 Few studies report smoking prevalence among people with dementia. In a cross-sectional analysis of patients treated for AD in a neurology clinic during a 2-year period, previous smoking prevalence was 29% (N=21/72).6 In a case–control study of patients with vascular dementia, the rate of current tobacco use was 9% (N=17/198) as compared with 6% (N=11/199) in the control group.7 Beyond the harmful health effects of smoking in people with dementia, there are concerns that smokers in this group may have a higher likelihood of fire accidents due to their compromised cognitive state.8 Since there are currently no available treatments to cure dementia, there is a growing interest in identifying modifiable risk factors for reducing the occurrence of the disease, to delay dementia onset, and reduce its burden.4 9 Smoking cessation could potentially decrease the risk or slow the development of dementia10 and could improve the quality of life of older adults through improved physical, and mental well-being.9 11 Little is known about whether people with dementia are prescribed smoking cessation agents and whether they are effective in this group.

n could potentially decrease the risk or slow the development of dementia10 and could improve the quality of life of older adults through improved physical, and mental well-being.9 11 Little is known about whether people with dementia are prescribed smoking cessation agents and whether they are effective in this group. Based on a Cochrane review of 136 trials, it was reported that nicotine replacement therapy (NRT) (compared with placebo or no treatment) can help people who make a quit attempt to increase their chances of successfully stopping smoking (Hartmann-Boyce et al).12 Data from observational studies13 14 and meta-analyses of randomised controlled trials15 indicate that varenicline is more effective than single form NRT for smoking cessation in the general population. However, it is unclear whether varenicline or NRT could help smokers with dementia to quit smoking. Therefore, in this study we aimed to (1) describe the rates of smoking prevalence and smoking cessation medication prescribing among people with and without general practitioner (GP)-recorded dementia in UK primary care settings from 2007 to 2015 and (2) assess and compare associations of varenicline and NRT on smoking cessation in people with GP-recorded dementia, compared with those without, at 3, 6, 9 months and 1, 2, 4 years after first prescription.

e with and without general practitioner (GP)-recorded dementia in UK primary care settings from 2007 to 2015 and (2) assess and compare associations of varenicline and NRT on smoking cessation in people with GP-recorded dementia, compared with those without, at 3, 6, 9 months and 1, 2, 4 years after first prescription. Methods Data source and population We conducted a retrospective cohort study using electronic medical records from 683 general practices in England from 2007 to 2015 using data from the Clinical Practice Research Datalink (CPRD). Patient data from the CPRD are broadly representative of the UK primary care population in terms of age, sex and ethnicity.16 These data have been validated, audited and quality checked.17 The study’s protocol (15_115R) was approved by the Independent Scientific Advisory Committee for MHRA Database Research (https://www.cprd.com/isac/). Code lists We defined variables using medical and product codes within the CPRD. All code lists were developed using a list from a previously published study.13 These codelists were derived from the the British National Formulary (BNF) and the International Classification of Diseases (ICD-10) and then agreed on by field experts (DR, KHT). A previous systematic review that checked the validity of coding of various diagnoses in what was then the General Practice Research Database (now CPRD) suggests that coding for dementia and Alzheimer’s is relatively accurate.18

national Classification of Diseases (ICD-10) and then agreed on by field experts (DR, KHT). A previous systematic review that checked the validity of coding of various diagnoses in what was then the General Practice Research Database (now CPRD) suggests that coding for dementia and Alzheimer’s is relatively accurate.18 Study subjects During the study period (2007–2015), we included people (aged ≥18 years) with information about their smoking status (either smokers or non-smokers) for smoking prevalence estimates, and we included smokers prescribed either varenicline or NRT for prescribing prevalence and for comparing quit rates. We used an open cohort design, with new patients entering the cohort throughout the observation period. For the primary objective, people were categorised as having dementia if: (i) they had ever been diagnosed with dementia (based on ICD-10 diagnoses F00-F03), or (ii) if they were prescribed dementia medications: (BNF chapter 4.11) (see online supplementary eTables 1 and 2 for a list of all Read and product codes used in this study). Then, the earliest record of GP-recorded dementia in the CPRD was taken forward ensuring that all records used were within the registration period of each patient. Patients with no records of the above-mentioned diagnosis/prescriptions were considered to have no GP-recorded dementia (for clarity, we hence forth refer to this as dementia). 10.1136/bmjopen-2018-027569.supp1Supplementary data

For the primary objective, people were categorised as having dementia if: (i) they had ever been diagnosed with dementia (based on ICD-10 diagnoses F00-F03), or (ii) if they were prescribed dementia medications: (BNF chapter 4.11) (see online supplementary eTables 1 and 2 for a list of all Read and product codes used in this study). Then, the earliest record of GP-recorded dementia in the CPRD was taken forward ensuring that all records used were within the registration period of each patient. Patients with no records of the above-mentioned diagnosis/prescriptions were considered to have no GP-recorded dementia (for clarity, we hence forth refer to this as dementia). 10.1136/bmjopen-2018-027569.supp1Supplementary data For the secondary objective, we constructed a cohort of eligible first varenicline/NRT prescriptions (see online supplementary eFigure 1 for a flowchart of numbers of patients excluded and reasons for exclusion). Within that cohort of eligible prescriptions, we considered individuals with dementia to be those with recorded Read codes for ever dementia/dementia medications prior to first varenicline/NRT prescription; we did this to ensure that a diagnosis of dementia preceded the exposure (prescription of a smoking cessation medicine).

that cohort of eligible prescriptions, we considered individuals with dementia to be those with recorded Read codes for ever dementia/dementia medications prior to first varenicline/NRT prescription; we did this to ensure that a diagnosis of dementia preceded the exposure (prescription of a smoking cessation medicine). Variables Smoking and prescribing prevalence estimates For smoking prevalence estimates, a patient’s smoking status (aged ≥18 years) was defined by a record indicating smoking/non-smoking or prescription of NRT/varenicline in that year. In case of missing information about smoking, the patient’s smoking status was carried forward until there was evidence of a change in smoking status or carried backwards if smoking status was only recorded in the final year of registration. Records that were outside the registration period for each patient were excluded. Exposure Exposure was defined as prescription of varenicline or NRT (eg, patches and so on, on prescription as opposed to over-the-counter, hence forth we refer to this as NRT). Prescriptions used to define exposure groups were issued between 1 September 2006 and 31 August 2016, with no prior record of use of a related product in the preceding 18 months. We used the first treatment episode to ensure that intervention groups were ‘new users’ of the medication.19 We did not model multiple and repeated prescriptions of smoking cessation medications during follow-up because this is likely to be strongly related to patient characteristics.

roduct in the preceding 18 months. We used the first treatment episode to ensure that intervention groups were ‘new users’ of the medication.19 We did not model multiple and repeated prescriptions of smoking cessation medications during follow-up because this is likely to be strongly related to patient characteristics. Outcome: smoking cessation Smoking cessation was defined as having an electronic record indicating a non-smoking status. The closest smoking record to each follow-up period was selected to determine each study participant’s smoking status; that is, the most recent smoking record identified between cohort entry and each follow-up period (eg, 3 months, 6 months). People with missing smoking data (beyond 180 days) were assumed to be continuing smokers20 which has been previously found to be robust in sensitivity analyses.13

udy participant’s smoking status; that is, the most recent smoking record identified between cohort entry and each follow-up period (eg, 3 months, 6 months). People with missing smoking data (beyond 180 days) were assumed to be continuing smokers20 which has been previously found to be robust in sensitivity analyses.13 Covariates Covariates included patients’ age at time of prescription, sex, index of multiple deprivation score (IMD), mean number of GP visits 1 year prior to first prescription, year of first prescription of a smoking cessation medication, body mass index (BMI), days registered in the CPRD, the Charlson Index (a measure of chronic illness),21 alcohol misuse, history of mental disorder or psychoactive medication prescriptions, evidence of other psychoactive medication prescription or other less common psychiatric disorders. We used multiple imputation to handle missing data on BMI and IMD. This was done using the ICE command in Stata where we produced 20 imputed datasets (see online supplementary eTables 3). We included all exposures, covariates and outcomes in the imputation model.22 Follow-up The secondary outcome was smoking cessation at 2 years, and this was also assessed at 3, 6, 9 months and 1 and 4 years after first prescription of varenicline or NRT.

Covariates Covariates included patients’ age at time of prescription, sex, index of multiple deprivation score (IMD), mean number of GP visits 1 year prior to first prescription, year of first prescription of a smoking cessation medication, body mass index (BMI), days registered in the CPRD, the Charlson Index (a measure of chronic illness),21 alcohol misuse, history of mental disorder or psychoactive medication prescriptions, evidence of other psychoactive medication prescription or other less common psychiatric disorders. We used multiple imputation to handle missing data on BMI and IMD. This was done using the ICE command in Stata where we produced 20 imputed datasets (see online supplementary eTables 3). We included all exposures, covariates and outcomes in the imputation model.22 Follow-up The secondary outcome was smoking cessation at 2 years, and this was also assessed at 3, 6, 9 months and 1 and 4 years after first prescription of varenicline or NRT. Statistical analysis Smoking prevalence Smoking rates were calculated by dividing the number of people with dementia who had Read codes indicating current smoking for each year between 2007 and 2015 by the total number of people with dementia and a smoking status code (indicating current or non/ex-smoking) each year between 2007 and 2015. For comparison, smoking prevalence was also estimated among people without dementia.

th dementia who had Read codes indicating current smoking for each year between 2007 and 2015 by the total number of people with dementia and a smoking status code (indicating current or non/ex-smoking) each year between 2007 and 2015. For comparison, smoking prevalence was also estimated among people without dementia. Prescribing prevalence The prevalence of varenicline and NRT prescribing among current smokers was calculated by dividing the number of prescriptions each year from 2007 to 2015 (there were no varenicline prescriptions for patients with dementia in 2006) by the number of current smokers in each year. Prevalence was estimated for people with and without dementia. Individuals with missing smoking information were excluded from the denominators. Association of varenicline and NRT prescriptions with smoking cessation We had originally planned to compare the effectiveness of NRT and varenicline for smoking cessation in individuals with dementia. However, given the small numbers of individuals prescribed varenicline, we had insufficient power to conduct this analysis. Therefore, we compared the effectiveness of being prescribed either varenicline or NRT on smoking cessation in individuals with dementia compared with individuals without dementia. This was determined by estimating quit rates at each follow-up period for individuals prescribed either of these medications. This was calculated by dividing the number of non-smokers in each group by the total number of people in that group at each follow-up period. All analyses were conducted using Stata V.14 MP.

ia. This was determined by estimating quit rates at each follow-up period for individuals prescribed either of these medications. This was calculated by dividing the number of non-smokers in each group by the total number of people in that group at each follow-up period. All analyses were conducted using Stata V.14 MP. Patient and public involvement This study was based on the analysis of anonymised primary care data. No patients were involved during the design and analysis of this study. Results Smoking prevalence and smoking cessation medication prescribing estimates Unadjusted smoking prevalence among people with dementia steadily decreased from 11% (N=2965/27 432 in 2007 to 7% (N=2690/36 249) in 2015 (figure 1). These estimates were consistently lower than in people without dementia, 26% (N=1 010 530/3 860 169) in 2007 and 21% (N=628 444/3 068 743) in 2015, respectively (see online supplementary eTable 4). However, after age and sex standardisation, the smoking prevalence among people with dementia was slightly higher than in people without (see online supplementary eFigure 2). Figure 1 Percentage (%) of primary care patients with an electronic medical record indicating smoking, from 2007 to 2015, in people with or without dementia. The rate of NRT prescribing in people without dementia was 7% (68 935/1 010 530) in 2007 which decreased to 2% (N=13 626/628 444) by 2015, whereas NRT prescribing rates among people with dementia increased during the same period. Compared with people without dementia, people with dementia had lower prescribing rates of varenicline (figure 2).

ing in people without dementia was 7% (68 935/1 010 530) in 2007 which decreased to 2% (N=13 626/628 444) by 2015, whereas NRT prescribing rates among people with dementia increased during the same period. Compared with people without dementia, people with dementia had lower prescribing rates of varenicline (figure 2). Figure 2 Prescription prevalence of varenicline or NRT in primary care, from 2007 to 2015, in smokers with dementia, compared with smokers without dementia. NRT, nicotine replacement therapy. Smoking cessation amongst individuals prescribed NRT and varenicline Of the 235 314 people included in this analysis, 447 were people with dementia, whereas 234 867 were people without (see online supplementary eFigure 1). Overall, 159 736 smokers were prescribed NRT and 75 578 prescribed varenicline (table 1). The mean age of people with dementia at the time of smoking medication prescription was about 72 years (SD=12.6), while that of people without dementia was 46 years (SD=14.8) People with dementia were about 25 years older, had more GP visits 1 year prior to the first prescription, suffered from more comorbidities and received more mental health-related prescriptions than those without dementia (table 1). Table 1 Baseline characteristics of people with or without dementia by exposure group, N (%)

Smoking cessation amongst individuals prescribed NRT and varenicline Of the 235 314 people included in this analysis, 447 were people with dementia, whereas 234 867 were people without (see online supplementary eFigure 1). Overall, 159 736 smokers were prescribed NRT and 75 578 prescribed varenicline (table 1). The mean age of people with dementia at the time of smoking medication prescription was about 72 years (SD=12.6), while that of people without dementia was 46 years (SD=14.8) People with dementia were about 25 years older, had more GP visits 1 year prior to the first prescription, suffered from more comorbidities and received more mental health-related prescriptions than those without dementia (table 1). Table 1 Baseline characteristics of people with or without dementia by exposure group, N (%) Characteristic People with dementia (N=447) People without dementia (N=234 867) NRT (N=409) Varenicline (N=38) Total NRT (N=159 327) Varenicline (N=75 540) Total Age at time of first prescription* 71.1 (12.2) 66.2 (15.1) 70.7 (12.6) 46.2 (15.5) 44.4 (13.2) 45.6 (14.8) Sex (male) 186 (45.5%) 19 (50.0%) 205 (45.9%) 73 674 (46.2%) 37 676 (49.9%) 111 350 (47.4%) Index of multiple deprivation score (IMD)†‡ 3 4 3 3 3 3 No of GP visits 1 year prior to first prescription* 11.5 (9.0) 15.3 (9.9) 11.8 (9.1) 8.9 (7.4) 7.3 (6.1) 8.4 (7.0) BMI*† 24.6 (5.1) 25.7 (6.3) 24.7 (5.4) 26.5 (5.7) 26.5 (5.4) 26.5 (5.6) Year of first prescription† 2010 2010 2010 2009 2010 2009 Days of history* 3573.8 (2181.2) 3450.3 (2327.4) 3563.3 (2191.5) 3052.9 (1907.1) 3164.8 (1986.2) 3088.9 (1933.6) Comorbidity ever (Charlson Index) 354 (86.6%) 28 (73.7%) 382 (85.5%) 59 489 (37.3%) 24 017 (31.8%) 83 506 (35.6%) Alcohol misuse ever 104 (25.4%) 11 (29.0%) 115 (25.7%) 13 890 (8.7%) 4759 (6.3%) 18 649 (7.9%) Self-harm ever 67 (16.4%) 9 (23.7%) 76 (17.0%) 17 232 (10.8) 6652 (8.8%) 23 884 (10.2%) Ever anxiety and stress related disorders 151 (36.9%) 16 (42.1%) 167 (37.4%) 44 381 (27.9%) 17 377 (23.0%) 61 758 (26.3%) Other behavioural/neurological disorder ever 30 (7.3%) 6 (15.8%) 36 (8.1%) 8693 (5.5%) 2956 (3.9%) 11 649 (5.0%) Ever depression 217 (53.1%) 26 (68.4%) 243 (54.4%) 65 343 (41.0%) 26 097 (34.6%) 91 440 (38.9%) Ever antidepressants 273 (66.7%) 28 (73.7%) 301 (67.3%) 79.584 (50.0%) 32 230 (42.7%) 111 814 (47.6%) Ever antipsychotics 175 (42.8%) 13 (34.2%) 188 (42.1%) 28 972 (18.2%) 9792 (13.0%) 38 764 (16.5%) Ever hypnotics/anxiolytics 238 (58.2%) 20 (52.6%) 258 (57.7%) 60 092 (37.7%) 25 134 (33.3%) 85 226 (36.3%) *Data presented are mean and SD.

s 273 (66.7%) 28 (73.7%) 301 (67.3%) 79.584 (50.0%) 32 230 (42.7%) 111 814 (47.6%) Ever antipsychotics 175 (42.8%) 13 (34.2%) 188 (42.1%) 28 972 (18.2%) 9792 (13.0%) 38 764 (16.5%) Ever hypnotics/anxiolytics 238 (58.2%) 20 (52.6%) 258 (57.7%) 60 092 (37.7%) 25 134 (33.3%) 85 226 (36.3%) *Data presented are mean and SD. †Data presented are median. ‡Missing data: BMI data were missing for 14.1% (N=33 059); IMD data were missing for 43.6% (N=102 657). BMI, body mass index; GP, general practitioner. Smokers with dementia were 74% (95% CI 64% to 82%) less likely to be prescribed varenicline than NRT, compared with smokers without dementia (see online supplementary eTable 5). The proportion of people with and without dementia who quit smoking after being prescribed either varenicline or NRT increased throughout the study’s follow-up period (3 months to 4 years). At 2-year follow-up, people with dementia were more likely to quit smoking (30.6%, 95% CI 25.8% to 35.1%) than those without (25.7%, 95% CI 25.4% to 25.8%) when prescribed either varenicline or NRT (figure 3) (see online supplementary eTable 6). However, after adjusting for all covariates, we found no evidence for a difference in quit rates between individuals with and without dementia (OR=1.0, 95% CI 0.81 to 1.23) (see online supplementary eTable 7). Figure 3 Percentage (%) of people with an electronic medical record indicating smoking cessation at up to 2 years follow-up, in people with and without dementia.

Smokers with dementia were 74% (95% CI 64% to 82%) less likely to be prescribed varenicline than NRT, compared with smokers without dementia (see online supplementary eTable 5). The proportion of people with and without dementia who quit smoking after being prescribed either varenicline or NRT increased throughout the study’s follow-up period (3 months to 4 years). At 2-year follow-up, people with dementia were more likely to quit smoking (30.6%, 95% CI 25.8% to 35.1%) than those without (25.7%, 95% CI 25.4% to 25.8%) when prescribed either varenicline or NRT (figure 3) (see online supplementary eTable 6). However, after adjusting for all covariates, we found no evidence for a difference in quit rates between individuals with and without dementia (OR=1.0, 95% CI 0.81 to 1.23) (see online supplementary eTable 7). Figure 3 Percentage (%) of people with an electronic medical record indicating smoking cessation at up to 2 years follow-up, in people with and without dementia. Discussion People with dementia were less likely to be prescribed varenicline compared with those without dementia. There was no clear evidence for a difference in quit rates in individuals with and without dementia following prescription of NRT or varenicline.

Figure 3 Percentage (%) of people with an electronic medical record indicating smoking cessation at up to 2 years follow-up, in people with and without dementia. Discussion People with dementia were less likely to be prescribed varenicline compared with those without dementia. There was no clear evidence for a difference in quit rates in individuals with and without dementia following prescription of NRT or varenicline. A strength of this study is that we used primary care data from the CPRD which are representative of the UK primary care population.16 Hence, smoking rates in people with dementia in this study are likely to be generalisable to the dementia population in the UK and in similar countries. Additionally, we used expert-reviewed codelists to define both the exposure and the outcome which would reduce misclassification bias (ie, classifying people with dementia as people without and vice versa).

entia in this study are likely to be generalisable to the dementia population in the UK and in similar countries. Additionally, we used expert-reviewed codelists to define both the exposure and the outcome which would reduce misclassification bias (ie, classifying people with dementia as people without and vice versa). There are several limitations to this research. Due to the small sample size of people with dementia, we were not able to test the relative effectiveness of varenicline versus NRT on smoking cessation using regression models. We had no data on smokers who purchase over-the-counter NRT, therefore we might be underestimating the prevalence of NRT use, particularly among people without dementia. Hence, it is likely that the prevalence of NRT use among people without dementia is larger than the prevalence of NRT prescribing in this study. Moreover, outcome definition (smoking vs non-smoking status) was based on self-reported data rather than biochemical verification of smoking status. Additionally, social desirability bias may occur when unsuccessful quitters don’t disclose their smoking status truthfully. We also relied on point estimates (ie, smoking status reported at a single timepoint) for making conclusions about smoking status. This may not have captured long-term abstinence. In other words, it is possible that smoking status may have fluctuated between GP visits. A further limitation is having no information about patient adherence in taking their prescribed smoking cessation medications.

mepoint) for making conclusions about smoking status. This may not have captured long-term abstinence. In other words, it is possible that smoking status may have fluctuated between GP visits. A further limitation is having no information about patient adherence in taking their prescribed smoking cessation medications. We are not aware of previous population-based studies that estimated the smoking rates among people with dementia as the available evidence has been limited to small cross-sectional studies. For instance, a community study in China found that about 17% (N=69/186) of the elderly sample with dementia were current smokers compared with 25% (N=415/1664) in people without dementia.23 Results from the Toyama dementia survey in Japan show that only 4% of people with dementia smoked compared with 10% in those without.24 This high variability in the results points to the need for larger and more representative studies in people with dementia to be conducted. Meanwhile, we found that smoking prevalence has decreased steadily among people without dementia, from 26% in 2007 to 21% in 2015. This was fairly similar to the general population in England as evidenced by results from the Smoking Toolkit Study (24.2% in 2007 to 18.7% in 2015)25 and therefore speaks to the external validity of our study.

hat smoking prevalence has decreased steadily among people without dementia, from 26% in 2007 to 21% in 2015. This was fairly similar to the general population in England as evidenced by results from the Smoking Toolkit Study (24.2% in 2007 to 18.7% in 2015)25 and therefore speaks to the external validity of our study. We observed a low prevalence of varenicline prescribing during our study period in people with dementia. Our estimate for individuals without dementia was consistent with findings from a previous study that examined the use of varenicline for smoking cessation treatment in UK primary care using data from THIN database in 2011. Compared with our results from that year, our estimates appear slightly lower (1.1% vs 1.8% in the THIN study).26 While NRT prescribing rates increased from 4% in 2007 to 5% in 2015 in people with dementia, these rates declined over time in people without dementia. A recent report from the British Lung Foundation found that NRT prescribing through primary care in England has dropped about 75% during the last 10 years. That was mainly due to cuts to public health funding that would have adversely impacted specialist stop smoking services.27

ed over time in people without dementia. A recent report from the British Lung Foundation found that NRT prescribing through primary care in England has dropped about 75% during the last 10 years. That was mainly due to cuts to public health funding that would have adversely impacted specialist stop smoking services.27 Our study is among the first to investigate longer term smoking cessation after being prescribed varenicline and NRT among individuals with dementia in a real world setting. Our results suggest that both varenicline and NRT could produce long-term smoking cessation in people with dementia as they do in those without. Almost one-third of smokers with dementia quit smoking after 2-year follow-up. Regardless of the smoking cessation medication prescribed to people with dementia, it is important to acknowledge that achieving smoking cessation in this group may carry health benefits which would potentially improve their general health status and and/or extend life expectancy.28 This should ideally be coupled with improving diet quality and increasing physical activites that may shield quitters from weight gain after smoking cessation.29

ing smoking cessation in this group may carry health benefits which would potentially improve their general health status and and/or extend life expectancy.28 This should ideally be coupled with improving diet quality and increasing physical activites that may shield quitters from weight gain after smoking cessation.29 It is not clear why individuals with dementia are less likely to be prescribed varenicline than NRT compared with individuals without. Previous clinical and observational studies have established that varenicline is superior to single form NRT in achieving smoking cessation in different groups.13 30 31 Additionally, varenicline did not seem to be associated with an increased risk of documented cardiovascular events, depression or self-harm when compared with NRT in primary care in England.32 On the other hand, a recent study based on CPRD data concluded that NRT appears to increase cardiovascular events for patients prescribed NRT, compared with those receiving smoking cessation advice after 52 weeks of follow-up.33 This was consistent with the evidence shown by a meta-analysis of 120 studies involving 177 390 individuals.12 It is possible that GPs are less likely to prescribe varenicline to individuals with dementia because of lower likelihood of adherence; in a recent systematic review of the literature, older patients with dementia were found to have a low level of medication adherence.34

of 120 studies involving 177 390 individuals.12 It is possible that GPs are less likely to prescribe varenicline to individuals with dementia because of lower likelihood of adherence; in a recent systematic review of the literature, older patients with dementia were found to have a low level of medication adherence.34 In summary, age-adjusted and sex-adjusted smoking prevalence among individuals with dementia was similar to those without dementia and smoking cessation rates were similar following prescription of smoking cessation medications between these groups. However, individuals with dementia were less likely to be prescribed varenicline than individuals without dementia. Supplementary Material Reviewer comments Author's manuscript Contributors: TI contributed to data cleaning, data analysis, interpretation of results and writing the manuscript. RM, GT, ND, AT and KT contributed to study conceptualisation, study design, interpretation of results, data analysis and writing the manuscript. MM and DR contributed to study conceptualisation, study design, interpretation of results and writing the manuscript. TJ extracted the data and contributed to writing the manuscript. TI, AT and ND had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

ted to study conceptualisation, study design, interpretation of results and writing the manuscript. TJ extracted the data and contributed to writing the manuscript. TI, AT and ND had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Funding: KT is funded by a National Institute for Health Research postdoctoral fellowship (PDF-2017-10-068). TJ receives funding from the National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care (NIHR CLAHRC) West. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. RM is supported by the NIHR Bristol Biomedical Research Centre, a partnership between the University Hospitals Bristol NHS Foundation Trust and the University of Bristol; and by a Cancer Research UK Programme Grant (C18281/A19169). GT is funded by a Cancer Research UK Population Researcher Postdoctoral Fellowship Award (C56067/A21330). This research was supported by Global Research Awards for Nicotine Dependence (GRAND), an independently reviewed competitive grants program supported by Pfizer, to the University of Bristol. Disclaimer: The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Competing interests: None declared. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.

Funding: KT is funded by a National Institute for Health Research postdoctoral fellowship (PDF-2017-10-068). TJ receives funding from the National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care (NIHR CLAHRC) West. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. RM is supported by the NIHR Bristol Biomedical Research Centre, a partnership between the University Hospitals Bristol NHS Foundation Trust and the University of Bristol; and by a Cancer Research UK Programme Grant (C18281/A19169). GT is funded by a Cancer Research UK Population Researcher Postdoctoral Fellowship Award (C56067/A21330). This research was supported by Global Research Awards for Nicotine Dependence (GRAND), an independently reviewed competitive grants program supported by Pfizer, to the University of Bristol. Disclaimer: The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Competing interests: None declared. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed. Data availability statement: Data may be obtained from a third party and are not publicly available.