Association Between Smoking and Blood Pressure : Evidence From the Health Survey for England

The HSE is an annual nationwide household survey.⁹ Members of a stratified random sample that is sociodemographically representative of the English population are invited to participate. The annual household response rate is ≈78% overall but slightly lower in men and in inner city residents. Data collection involves an interviewer’s visit (which includes height and weight measurements), followed by a visit from a nurse, who measures BP, records current use of prescribed medicines, and takes a blood sample. Overall, 48 307 adults aged ≥16 years were interviewed in the 1994, 1995, and 1996 surveys.

BP was measured with the automated Dinamap 8100 monitor.¹⁰ With an appropriately sized cuff, 3 BP readings were taken on the right arm with the informant in a seated position after a 5-minute. Informants who had eaten, drunk alcohol, or smoked in the 30 minutes before the measurement were excluded from analysis. Data used in this study are based on the mean of the second and third measurements.

Detailed information on smoking habits was collected. Past smokers were defined as subjects who had stopped smoking >1 year ago; those who stopped within 1 year (n=868) and those with missing data on smoking status (n=81) were excluded. Current smokers were further categorized according to number of cigarettes smoked per day (light 1 to 9, moderate 10 to 19, heavy ≥20). In 1994 and 1996 (but not 1995), blood samples were analyzed for serum cotinine levels.

Body mass index (BMI) was defined as weight (kg)/height (m)². Respondents were classified into 1 of 4 categories of BMI: ≤20 kg/m², underweight; >20 to 25 kg/m², normal weight; >25 to 30 kg/m², overweight; and >30 kg/m², obese.

Respondents were asked about the frequency, quantity, and type of alcoholic drink consumed during the past 12 months. Consumption in units per week was categorized as (1) none, past drinking, or <1 unit; (2) 1 to 10 units for men or 1 to 7 units for women; (3) 11 to 21 units for men and 8 to 14 units for women; and (4) >21 units for men or >14 units for women.

Social class was assigned on the basis of occupation of the head of the household, with the Registrar General’s standard classification.¹¹ Social classes were further grouped into manual (skilled manual, partly skilled, and unskilled occupations) and nonmanual (professional, managerial and technical, and skilled occupations).

The psychosocial measures included in the HSE have varied each year. The self-administered General Health Questionnaire (GHQ12), designed to detect possible psychiatric morbidity, was included in 1994 and 1995. A score of ≥4 was used to identify possible psychiatric disorders.

In 1994, physical activity levels and serum cholesterol levels were assessed. Physical activity was categorized as inactive (those informants who reported <12 occasions of moderate or vigorous activity during the 4 weeks before the interview) and active (the remainder). Total serum cholesterol was categorized as normal (<6.5 mmol/L) and raised (≥6.5 mmol/L). Separate analyses that included these covariates were run only for 1994.

Analyses are presented for the 33 860 persons (70% of those interviewed) who provided data on smoking status and had valid BMI and BP measurements, of whom 11 222, 14 063, and 8575 were never, past, and current smokers, respectively.

The mean ages of male and female respondents were 46.6 years (range 16 to 97 years) and 47.1 years (range 16 to 97 years) respectively. Among men, past smokers were older (mean [SEM] age 52.4 [0.2] years) than never (41.5 [0.2] years) and current (41.9 [0.2] years) smokers. Past smokers also made up the oldest group of women (51.0 [0.2] years), whereas current smokers made up the youngest group (42.6 [0.2] years) and never smokers had a mean age of 46.1 (0.2) years. Among current smoking men and women, heavy smokers tended to be older.

Table 1 shows age-adjusted characteristics of the men and women studied according to smoking status. Similar rates of current smoking were found in men and women, although among current smokers, more men than women were heavy smokers. On average, among men and women, current smokers had significantly lower BMIs than did never and past smokers. However, heavy smokers had the same mean BMI as never smokers. In both genders, there was a higher proportion of heavy drinkers among current smokers than among never smokers, whereas alcohol consumption among past smokers was between that of never and current smokers. The age-adjusted proportion of heavy alcohol consumption was higher among heavy smokers than among light smokers. The proportion of men and women from manual social classes was higher among current smokers than among never or past smokers, and the proportion increased with the number of cigarettes smoked. The proportion of both men and women who scored ≥4 on the GHQ12 questionnaire was higher in smokers than in nonsmokers.

Table 2 shows the effect of various characteristics on age-adjusted BPs. In men, increasing alcohol consumption was positively associated with BP. In women, mean SBP and DBP showed a U-shaped relationship with alcohol intake. Increasing BMI, manual social class, raised cholesterol level, physical inactivity, and normal GHQ12 score were associated with increasing BP.

Overall, age-adjusted BPs did not differ importantly among never, past, and current smokers (Table 3), although in men, a small significant difference was observed in mean SBP between never smokers (139.9 mm Hg) and current smokers (140.7 mm Hg) (P<0.05), and heavy smokers (141.4 mm Hg) (P<0.05). In women, the lowest BPs were observed among light smokers, who had significantly lower levels than never smokers (135.5/72.4 and 136.8/73.6 mm Hg, respectively; P<0.05).

Mean BP levels after adjustment for age, BMI, alcohol intake, and social class are shown stratified by age and smoking status in Table 4. Among women of both age groups, the lowest BPs were found among light smokers, whereas in men, the lowest mean BPs were found among past smokers. In most age/gender strata, the highest BPs among current smokers were found in the heaviest smokers. However, only among older men and women (>44 years) were SBPs of current smokers greater than SBPs of never smokers and only significantly so for men. No significant differences were seen between DBPs of never and heavy smokers for either gender. The data shown in Table 4 were unaffected by the exclusion of social class from the analytic models.

Among men with normal weight, no association was seen between smoking and SBP, whereas in women of normal weight, past smokers had a mean SBP that was 0.9 mm Hg lower than that of nonsmokers (P=0.04). Among overweight or obese people, male moderate and heavy smokers had an SBP that was 2.1 and 1.8 mm Hg, respectively, higher than that of nonsmokers (P<0.01), whereas no significant differences were observed among women (Figure). A formal test for interaction between BMI and the smoking-SBP relationship was significant in men (P=0.02).

In both never and heavy smoking men, alcohol intake showed a graded effect on mean SBP (P<0.05), and smoking was associated with higher SBP levels only among alcohol drinkers, although the association did not reach statistical significance (Table 5). Further adjustment of these data for BMI did not alter BP patterns by alcohol intake and smoking strata. In nonsmoking women, SBP was lower in moderate than in nondrinkers and heavy drinkers, whereas among light smokers, SBP increased with increasing alcohol intake. Light smokers had lower SBPs than never smokers in all 3 strata of alcohol intake but only significantly so among nondrinkers. The difference in SBP levels between never and light smokers was reduced after adjustment for BMI, regardless of alcohol intake but remained statistically significant among nondrinkers.

Patterns in BPs in the 1994 survey, including additional adjustment for cholesterol levels, physical activity, and GHQ12 scores, were essentially the same as those seen in Tables 4 and 5.

No association between smoking and use of antihypertensive agents was observed in either gender after adjustment for age, alcohol consumption, BMI, and social class, and the results shown in Table 4 did not change substantially when the analyses were repeated and excluded those on antihypertensive treatment (data not shown).

The results of the present study, the largest to investigate this issue to date, show that any independent chronic effect of smoking on BP is small. After adjustment for age, BMI, alcohol, and social class, we found only in older men significantly higher SBPs in heavy and moderate smokers than in never smokers, whereas no such differences were seen for DBP. Therefore, at least in men, our findings do not support those from other studies, which observed lower BPs at increasing levels of cigarette consumption.^{6 7} The finding of elevated SBP, but not DBP, only in older men is an important new observation that is compatible with effects expected in association with chronically enhanced atherogenesis in large capacitance vessels, with which smoking is associated and which produces isolated systolic hypertension.^{14 15} In women, we showed lower mean BP levels in light smokers than in nonsmokers and heavier smokers, although the differences reached statistical significance only for DBP in the younger age group. Because of the strong interrelationships among smoking, alcohol, and BMI and because an interaction was detected, it was important to examine the relationship between smoking and SBP according to BMI categories. Previous studies have been of insufficient size to adjust data appropriately for the potential confounding effects of these other variables. No association between smoking and SBP was seen in men of normal weight, whereas in overweight and obese men, not only heavy, but also moderate, smokers showed a significant increase in SBP. These differences were not observed in women.

Among smokers, a U-shaped relationship between the number of cigarette smoked and relative body weight has been found in several studies, with those smoking 10 to 20 cigarettes/d being the leanest.¹⁶ Although this seems paradoxical given that smoking is associated with increased 24-hour energy expenditure,¹⁷ heavy smokers may weigh more because of other habits, such as high alcohol and saturated fat intakes. Other dietary variables, such as electrolyte intake, may further confound the BP-smoking association, because smokers have been reported to have different diets than nonsmokers.¹⁸ We did not have dietary data but showed that heavy smokers tend to drink more than nonsmokers or lighter smokers and, in the 1994 survey, had higher cholesterol levels.

Alcohol intake, which is strongly associated with smoking,¹² also appears to affect the smoking-BP relationship: a difference in the relationship between alcohol consumption and BP among smokers and nonsmokers was observed in 3 German cross-sectional studies.^{19 20} In 1 of these studies,¹⁹ effect modification by smoking was stronger in men than in women. Moreover, BMI, which has well-established effects on BP,²¹ is also affected by alcohol intake, and this, too, must be taken into account when assessing the relationship between smoking and BP because of the apparent opposing effects that alcohol and smoking exert on body weight.

Given these complex interrelationships, we examined the relationship between smoking and SBP separately by alcohol intake status, before and after adjustments for BMI. Among men, we observed higher levels of SBP with increasing alcohol intake at each level of smoking, whereas in nondrinkers, heavy smokers did not have higher SBPs. In contrast, among women, SBP was lower among light smokers than among never smokers regardless of alcohol intake, and alcohol showed only a small dose-response effect on SBP among light smokers and not among never smokers. Hence, differences in the BP-smoking association observed between men and women appear to be due to the stronger interrelation between smoking and drinking in men. For any given smoking category, women tend to consume smaller amounts of alcohol than men. Among men, adjustment for BMI did not affect the increase in SBP with increasing alcohol intake among never or heavy smokers: the effect of smoking on SBP was small, variable in direction, and nonsignificant. In women, the difference in SBP levels between never and light smokers was reduced, but still significant, after adjustment for BMI in nondrinkers, appearing to be in part due to the confounding effect of body weight.

One of the possible limitations of the study is that the HSE collects information by means of a questionnaire, and therefore some of the risk factor data are self-reported, which may be inaccurate.²² However, agreement between self-reported smoking status and serum cotinine levels was good: only 3.5% of men and 0.8% of women who reported never smoking had a cotinine level of ≥20 ng/mL, a level that is strongly suggestive of smoking. Reclassification of these respondents as smokers did not affect the observed results for SBP and DBP (data not shown). The use of the Dinamap 8100 for BP measurement has been challenged.²³ However, this machine is considered to be suitable and sufficiently accurate in the setting of a large multicenter survey.¹⁰

In summary, these data from a large nationally representative database suggest that there are no consistent independent differences of clinical significance in BP values between smokers and nonsmokers. The BP differences associated with smoking that were observed in this study differed with age and between men and women and may well be explained at least in part by differential confounding effects of BMI and alcohol intake. However, because smoking and BP have been shown to exert a synergistic adverse effect on the risk of coronary heart disease,³ it is critical that persons with raised BP are advised to stop smoking. Furthermore, because BP levels in smokers are rarely recorded during or immediately after smoking when acute rises in BP occur, usual BP levels of smokers tend to be systematically underestimated.

The Health Survey for England is commissioned by the Department of Health and was carried out by the Joint Health Survey Unit of National Center for Social Research (formerly SCPR) and Department of Epidemiology and Public Health at University College London.