Associations between intelligence in adolescence and indicators of health and health behaviors in midlife in a cohort of Swedish women

Associations between intelligence in adolescence and indicators of health and health behaviors in midlife in a cohort of Swedish women

Intelligence 40 (2012) 82–90 Contents lists available at SciVerse ScienceDirect Intelligence Associations between intelligence in adolescence and i...

201KB Sizes 0 Downloads 40 Views

Intelligence 40 (2012) 82–90

Contents lists available at SciVerse ScienceDirect

Intelligence

Associations between intelligence in adolescence and indicators of health and health behaviors in midlife in a cohort of Swedish women Karin Modig a,⁎, Lars R. Bergman b a b

Institute of Environmental Medicine, Division of Epidemiology, Karolinska Institute, Box 210, 71 77 Stockholm, Sweden Stockholm University, Department of Psychology, Sweden

a r t i c l e

i n f o

Article history: Received 22 August 2011 Received in revised form 2 February 2012 Accepted 4 February 2012 Available online 22 February 2012 Keywords: Cognitive epidemiology Health behavior Health Women Intelligence

a b s t r a c t The objective of this study was to investigate associations between intelligence and indicators of health status and health behaviors at age 43 in a cohort of Swedish women (n = 682). Intelligence was measured by standard IQ tests given at ages 10, 13, and 15. At the age of 43, 479 of the women were sampled for a medical examination in which 369 participated (77% participation rate). We performed correlations of IQ and the continuous health variables and we estimated logistic regression models with dichotomous health variables as the dependent variables. No significant correlations were found between IQ and any of the continuous health variables. In unadjusted logistic regression models where the cut-off points were set based on standard health risk levels, four out of sixteen indicators of unfavorable health status and health behaviors showed significant negative associations with intelligence, meaning higher risk with decreasing IQ-score. After adjusting for educational level, two remained statistically significant: being obese, OR 1.51 (95% CI 1.08, 2.12) and having a high systolic blood pressure OR 1.45 (95% CI 1.03, 2.03). For all other health variables, this study finds no support for a sizable association between IQ in adolescence and indicators of health and health behavior in midlife among Swedish women. © 2012 Elsevier Inc. All rights reserved.

1. Introduction The inverse association between early intelligence and later morbidity and mortality, sometimes referred to as cognitive epidemiology (Deary, 2005; Deary & Batty, 2007), is well established (Batty, Deary, & Gottfredson, 2007; Batty et al., 2009), at least among men. This relationship, together with its potential underlying mechanisms, has also been discussed in several papers (Batty, Deary, & Gottfredson, 2007; Batty, Kivimaki, & Deary, 2010; Deary, 2009; Lager, Bremberg, & Vagero, 2009, 2010). In these studies, intelligence was usually measured by global tests of mental ability where the scores on different subtests were added to give a combined score. Of course, such measures cannot automatically be assumed to be synonymous with intelligence in a

⁎ Corresponding author. E-mail address: [email protected] (K. Modig). 0160-2896/$ – see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.intell.2012.02.002

more general sense since there exist different theories of the nature and structure of intelligence (Carrol, 1993; Gardner, 1993; Sternberg, 1985). However, within the dominant gfactor paradigm it is claimed that almost any high-quality test of mental ability that has a reasonably broad set of tasks will load heavily in the g-factor (Gottfredson, 2004; Jensen, 1998). Our definition of intelligence is based on this paradigm and, to avoid confusion with the broader intelligence concept, we henceforth mostly use the term “IQ” instead of intelligence. Despite the established association between early IQ and later health outcomes the underlying mechanisms linking IQ to them are not clear. IQ may act through other variables, or perhaps not at all, if the associations are due to confounding from other factors. Further, as the association between IQ and mortality has been shown in some studies to be present for men only and not for women (Kuh, Richards, Hardy, Butterworth, & Wadsworth, 2004; Lager et al., 2009; Pearce, Deary, Young, & Parker, 2006) it raises questions about

K. Modig, L.R. Bergman / Intelligence 40 (2012) 82–90

whether the mechanisms linking IQ to health outcomes are different for men and women. An often discussed mechanism is mediation by health behaviors, i.e. that IQ affects health behaviors which in turn affects health outcomes. The idea that highly intelligent individuals adopt healthier behaviors is perhaps the most advocated theory among people in general when discussing intelligence and health. It is not unlikely to think that a person with high intelligence would have a greater capacity to collect information about behaviors believed to affect health outcomes, compared to a person who is not so intelligent. Also, the disease and mortality outcomes that are associated with IQ appear to be outcomes related to lifestyle such as cardiovascular diseases (CVD), injuries, accidents, lung cancer, and skin cancer (Batty, Deary, & Gottfredson, 2007; Batty, Deary, Schoon, & Gale, 2007; Batty et al., 2007). Potential mediation from health behaviors has been tested in some studies by adjusting the IQ–mortality association for health behaviors (Batty et al., 2008; Kuh et al., 2004) and by looking at the association of IQ with health behaviors alone (Anstey & Sachdev, 2009; Batty, Deary, & Macintyre, 2007; Chandola, Deary, Blane, & Batty, 2006). Either way, the results are not conclusive and in most cases they are based on men only or men and women were analyzed together. Several studies have looked at early IQ and obesity later in life (Batty, Deary, & Macintyre, 2007; Batty, Deary, Schoon, & Gale, 2007; Chandola et al., 2006; Halkjaer, Holst, & Sorensen, 2003; Hart et al., 2004; Lawlor, Clark, Davey Smith, & Leon, 2006; Yu, Han, Cao, & Guo, 2010) and in most of these studies an association has been found in unadjusted models but after adjustment for adult socioeconomic position and/or education the associations are no longer statistically significant (Batty, Deary, Schoon, & Gale, 2007; Chandola et al., 2006; Halkjaer et al., 2003a; Lawlor et al., 2006). However, one study, based on men only, found a significant association also after adjustment for indicators of adult socioeconomic position (Batty, Deary, & Macintyre, 2007). In another study on men only, Batty et al. found an association of IQ with four of the five individual components comprising the metabolic syndrome: Hypertension, high body mass index, BMI, high triglycerides, and high blood glucose (Batty et al., 2008). Der, Batty, and Deary (2009) found several health outcomes, for example hypertension, to be associated with early IQlevel, however, they did not control the estimates for adult education or socioeconomic position. Several studies have found an inverse association of IQ and smoking status (Anstey & Sachdev, 2009; Batty, Deary, Schoon, & Gale, 2007; Batty, Shipley, et al., 2008; Hemmingsson, Kriebel, Melin, Allebeck, & Lundberg, 2008; Kubicka, Matejcek, Dytrych, & Roth, 2001) but in many of these studies the association vanished or was heavily attenuated after adjustment for indicators of social class. We have also shown in a twin study that the association between IQ and smoking status seems to be confounded by common environmental factors (Wennerstad et al., 2010). Only a few of these studies included women, and those who did analyzed men and women in the same model adjusting for gender, which makes it difficult to see whether the associations were the same for men and women. Another proposed mechanism behind the intelligence– health outcome relationship is what has been referred to as

83

“system integrity” (Deary, Weiss, & Batty, 2010). It explains the relationship based on the idea that optimal brain development is strongly connected to optimal development of other somatic systems such as the cardiovascular system. This may be due to confounding from common genetic or biological factors. The theory receives some support from the established inverse association with CVD of both height and IQ (Pearce, Deary, Young, & Parker, 2005), or that low birth weight for gestational age and IQ both are related to CVD risk (Bergvall, Iliadou, Johansson, Tuvemo, & Cnattingius, 2006). It is feasible that malnutrition or other non-optimal conditions in fetal life or early postnatal life affects both the development of the brain and also the development of the cardio-vascular system, creating a spurious correlation between intelligence and risk of CVD. Based on the divergent findings of the relationship between IQ and health behaviors and indicators of health, and the lack of studies based on female samples, our aim was to study these relationships for a cohort of Swedish women. In a longitudinal study they have been followed from the age of 10 to age 43 with IQ measured in early adolescence and health variables measured in midlife. 2. Methods 2.1. Sample The present study is based on data on women from the Swedish longitudinal research program Individual Development and Adaptation, IDA (Magnusson, 1988). In IDA a whole school grade cohort of children from the town of Örebro, Sweden, has been followed from age 10 in 1965 to age 49 in 2004 and children who moved into Örebro after the age of 10 have been added to the cohort (n= 682 females). Örebro is a midsized Swedish town with a population of about 100,000 inhabitants. It is in many respects fairly representative of Swedish urban communities, excluding the big cities. However, for the 13 years old children in 1968 the average educational level of the parents and the children's intelligence tests results were slightly higher than the corresponding figures for the average urban community (Bergman, 1973). IQ data from the school years were available for 657 girls, or 96% of the school grade cohort, and data from a medical examination at age 43 (in 1998) were available for 369 women, or 54% of the school grade cohort. However, from a representativeness point of view, the participation rate is higher (77%) since only a subsample of the whole school grade cohort was sampled for the medical examination (n =479; Bergman, 2000). To obtain some information about sample bias in the main investigation variable (IQ), we first compared the average IQ at age 10, 13, and 15 between those having taken part in the medical examination and the others in the school grade cohort and no significant differences were found. We then compared these two groups with regard to their final education obtained in another data collection at age 43 with a participation rate of 89% and again no significant differences were found (although there was a tendency for those who had taken part in the medical examination to have a lower final education than other cohort members). As expected, the IQ scores at age 10, 13 and 15 were highly correlated.

84

K. Modig, L.R. Bergman / Intelligence 40 (2012) 82–90

The correlations of the IQ at age15 with IQ at age 10 and 13 were 0.75 and 0.81, respectively (uncorrected for attenuation; after correction the correlations become 0.80 and 0.86, respectively).

measures (0.97 for height and 0.96 for weight). By this procedure we obtained BMI data for 546 women.

2.2. Variables

For the continuous outcomes we calculated Pearson correlation coefficients together with their 95% confidence intervals and p-values. To characterize shapes of associations we plotted each of the health variables against IQz-score, raw plots and by cumulative frequency percent, to see whether any outliers would affect the correlations. We found no evidence of the associations being other than linear, except for BMI where there was a tendency of lower IQ in the higher range of BMI. For some of the variables it was, however, difficult to detect nonlinearity due to a limited sample size. Therefore, to further investigate this possibility, the continuous variables were also transformed into dichotomous variables based on well-known risk limits previously associated with morbidity (Carlsson, Theobald, Hellenius, & Wandell, 2009; Kaminsky, Savage, Callas, & Ades, 2010; Wandell, Carlsson, & Theobald, 2009; Wood et al., 1998; Wormser et al., 2011). The cut-off points defining risk levels were BMI > 25 for overweight, BMI > 30 for obesity, WHR>85, systolic bp > 140 mm Hg, diastolic blood pressure> 90 mm Hg, S-Cholesterol > 5 mmol/L, HDL cholesterol b 1 mmol/L, and FEV1% b 80%. Following a standard epidemiological procedure, these variables were analyzed with logistic regression analysis (cut off points given in tables) together with the variables that were natural dichotomies (smoking status, giving up smoking, bad self-reported health status, unsatisfactory health status rated by examining physician, regular gynecological examinations, life style change, healthy eating and regular exercise). In the logistic regression analyses the binary health variable was the dependent variable of interest and IQz the independent variable. Analytically, the odds of health outcome were studied (i.e. the number or observations characterized by the health risk divided by the number of observations characterized by the absence of the health risk). It should be remembered that an odds ratio, OR, of 1.0 indicates that the IQ level is not related to the health risk, a OR in excess of 1.0 indicates an increased risk with decreasing IQ and an OR below 1.0 indicates an increased risk with rising IQ. The logistic regression models were carried out un-adjusted and adjusted by educational level as a proxy of socioeconomic position and the models provided odds OR for the more unfavorable health outcome per SD decrease in IQ. Adjusting for education can, however, to some extent be consider as over-adjusting as intelligence and education to some extent is a measure of the same underlying capacity. In this cohort the correlation between IQ and final education was 0.44 (CI 0.37, 0.51). All analyses were conducted in SAS version 9.2.

Information about intelligence was used from three sources: at age 15 from scores on the Swedish WIT III test battery (stanine scores) and at age 10 and 13 from scores on the Swedish DBA test battery. Both these tests are multifactor tests and only the global IQ score, measuring general intelligence, was used. Reported reliabilities for the two tests are 0.95 and 0.93, respectively. The tests are described more in detail in Backteman and Magnusson (1981). Of the 682 women, 570 had information on intelligence at age 15. Descriptive information of the raw IQ scores at age 10, 13, and 15 are as follows: (1) Means 104.7, 149.5, and 4.8 respectively, (2) SDs 25.3, 26.9, and 1.9, respectively, (3) Ranges 46–168, 52–201, and 1–9, respectively, and (4) Skewness 0.13, −0.67, and − 0.12, respectively. At age 13 a weak tendency for a ceiling effect was present. All IQ-values were standardized to z-scores for each age separately. The IQ measure we used in our analyses was primarily taken from age 15. For those who had missing value on IQ at age 15 we used the IQ z-score at age 13 (53 women) and, if missing also at 13, we used the IQ z-score at age 10 (34 women). By this strategy we ended up with 657 women with an IQ z-score. This score was used in the analyses and denoted with IQz (mean =−0.03 and SD=1.01). Based on self-reported data from a personal interview, given when the women were 43 years old, the variable final educational level was formed. It had three categories: 1. basic education (maximum 10 years), 2. secondary education (Swedish gymnasium), and 3. university degree. From the medical examination we had information on body mass index (BMI, calculated from weight and height measurements and expressed in kg/m 2), systolic and diastolic blood pressures, waist-to-hip ratio (WHR) and Forced Expiratory Volume (FEV, measured with micro spirometer) which we have analyzed as FEV1% predicted. The FEV1 is the volume of air that can be forced out in one second after taking a deep breath, and is an important measure of pulmonary function. It is generally converted to a percentage of normal (predicted), based on gender, age, height and weight (Gruffydd-Jones, 2011). Further, we had information on total cholesterol level (S-Cholesterol) and high density lipoprotein (HDL). Finally the physician who conducted the medical examination estimated the women's health status as “satisfying” or “unsatisfying”. Data on self-reported health status, height, weight, smoking status, exercise levels, nutritional habits, and whether or not going regularly to gynecological examinations were selfreported in questionnaires in 1998. Data on smoking status were self-reported in 2002 together with information about whether or not the women had changed their life style as a result of the medical examination in 1998. Height and weight were measured at the medical examination but also self-reported in questionnaires in the same year, given to a larger sample. To increase the sample size when computing BMI we therefore used the self-reported data since they correlated very highly with the objective

2.3. Data analysis

3. Results Table 1 shows number of observations and mean IQz for dichotomous variables of health and health behaviors. For all of the variables, except eating healthy and excising regularly, there is a tendency of a lower mean IQ for the more unfavorable outcomes. Table 2 presents descriptive data together with correlation coefficients between IQz and all continuous health variables. Although all relationships were in the expected direction,

K. Modig, L.R. Bergman / Intelligence 40 (2012) 82–90

85

Table 1 Number of observations, mean IQz scores together with their standard deviations (SD) and confidence intervals (CI) within parenthesis, for dichotomous health indicators. Indicators of health status

N (%)

Mean IQz (SD) (CI)

BMI > 30 (obesity) >25 (overweight) ≤25 Waist-to-hip ratio > 85 ≤85 Systolic bp > 140 ≤140 Diastolic bp > 90 ≤90 FEV1% b 80 ≥80 S-Cholesterol > 5 ≤5 S-HDL b 1 ≥1 Reported own health as: Bad Good GP health status is: Unsatisfactory Satisfactory

45 (8.6) 195 (37.0) 331 (62.9) 141 (39.7) 214 (60.3) 51 (14.3) 306 (85.7) 21 (5.9) 336 (94.1) 26 (7.3) 329 (92.7) 215 (60.2) 142 (39.8) 6 (1.7) 347 (98.3) 29 (6.2) 440 (93.8) 41 (11.6) 314 (88.4)

− 0.31 (0.96) (− 0.60, − 0.02) − 0.06 (1.03) (− 0.20, 0.09) 0.06 (0.96) (− 0.04, 0.17) − 0.09 (0.95) (− 0.24, 0.07) 0.09 (0.97) (− 0.04, 0.22) − 0.25 (1.04) (− 0.54, 0.05) 0.06 (0.94) (− 0.05, 0.16) − 0.10 (0.98) (− 0.54, 0.35) 0.02 (0.96) (− 0.08, 0.12) − 0.31 (1.08) (− 0.76, 0.12) 0.04 (0.95) (− 0.06, 0.14) − 0.03 (1.00) (− 0.16, 0.11) 0.08 (0.91) (− 0.07, 0.23) − 0.34 (1.14) (− 1.54, 0.86) 0.02 (0.96) (− 0.08, 0.12) − 0.30 (0.93) (− 0.65, 0.06) 0.04 (0.98) (− 0.05, 0.14) − 0.41 (1.10) (−0.76, − 0.06) 0.07 (0.93) (− 0.03, 0.18)

Indicators of health behavior

N (%)

Mean IQz (SD) (CI)

Ever daily smoker: Yes No Go to gynecological examination: Yes No Changed life style after medical examination: Yes No Thinks about eating healthy: Yes No Exercise regularlya: Yes No

338 200 417 52 58 280 279 248 290 236

− 0.06 (0.99) (− 0.17, 0.04) 0.15 (0.97) (0.02, 0.29) 0.04 (0.99) (− 0.05, 0.14) − 0.09 (0.87) (− 0.33, 0.15) − 0.10 (1.01) (− 0.36, 0.17) 0.08 (0.93) (− 0.03, 0.19) 0.01 (0.92) (− 0.10, 0.12) 0.04 (1.07) (− 0.10, 0.17) 0.01 (0.97) (− 0.10, 0.12) 0.05 (1.01) (− 0.08, 0.18)

(62.8) (37.2) (88.9) (11.1) (17.2) (82.8) (52.9) (47.1) (55.1) (44.9)

Note. a At least 2 times per week.

there were no significant correlations between IQ and any of the health indicators. For waist-to-hip ratio and systolic bp the relationships were almost significant (95% CI −0.20, 0.01) and (95% CI −0.19, 0.01) respectively. The variables extracted from the medical examination were based on a smaller sample (N= 346–369) than the variables extracted from surveys (N= 485–558). It is possible that there is a general weak positive effect of intelligence on most health indicators which is not strong enough to produce significant relationships, considering the limited power caused by the moderate sample sizes. To study this we constructed two indices based on dichotomized variables, each scored “1” if indicating unfavorable health, otherwise scored “0”. The first index was formed by summing all health status variables and the second by summing all health behavior variables. We then computed the Pearson correlations between the indices and IQz and they were −0.10 (95% CI −0.18, −0.01) and −0.05 (95% CI −0.13, 0.04), respectively. IQ was thus weakly but significantly correlated to the health status index, but not to the health behavior index. Table 3 presents results from the logistic regression analysis of the binary health indicators. Among the indicators of health status, being obese, having a high systolic blood pressure, and having unsatisfactory health status as rated by examining physician were significantly related to a lower IQz score in the unadjusted analysis although for all variables there was a consistent tendency for the estimated odds coefficients to

be somewhat over 1.00, suggesting a slightly increased risk for lower values of IQ. After adjusting for educational level the association was no longer statistically significant for unsatisfactory health status but remained significant for being obese, OR 1.51 (95% CI 1.08, 2.12) and having a high systolic blood pressure (above 140 mm Hg), OR 1.45 (95% CI 1.03, 2.03). This should be interpreted as; according to the model the odds of high blood pressure is multiplied by 1.45 per SD decrease in IQz. Among the indicators of health behaviors, having smoked daily for at least 6 month was the only indicator that was significantly associated with low intelligence in the unadjusted analyses but the significance vanished after adjustment for educational level. Having changed the life style as a result of the medical examination four years earlier was not associated with IQz; neither were thinking about eating healthy, regular exercise or regular gynecological examinations. 4. Discussion In this study we investigated the association between IQ measured in adolescence and several health indicators representing both health status and health behaviors in midlife among Swedish women. We used IQ at age 15 based on the Swedish WIT III test battery and, if missing, IQ measured at age 13 or age 10 based on the Swedish DBA test battery. As reported in the Results section, the correlations between the IQ-scores were 0.76–0.81, about 0.05–0.10 lower than

86

Table 2 Descriptive data, number of observations (N), mean (M), and standard deviation (SD), of continuous variables. Pearson correlation coefficients with their 95% confidence intervals within parenthesis.

IQa BMIb

Waist-to-hip ratioc

Diastolic bpe

Total cholesterolf

HDLg

FEV1%h

Number of cigarettesi

a

(SD) 1.01 4.2

0.84

124.03

75.27

5.35

1.70

98.38

10.22

0.05

16.41

10.07

0.92

0.44

12.73

5.79

IQ − 0.06 (− 0.14, N = 526 − 0.09 (− 0.20, N = 355 − 0.09 (− 0.19, N = 357 − 0.05 (− 0.15, N = 357 − 0.07 (− 0.18, N = 357 0.03 (− 0.08, N = 353 0.06 (− 0.04, N = 355 − 0.06 (− 0.21, N = 164

BMI

c

S bp

D bp

0.70⁎⁎⁎ (0.64, 0.75) N = 368 0.12⁎ (0.02, 0.22) N = 368 − 0.04 (− 0.14, 0.06) N = 364 0.02 (− 0.08, 0.12) N = 366 − 0.05 (− 0.23, 0.12) N = 122

0.15⁎⁎ (0.05, 0.25) N = 368 − 0.08 (− 0.18, 0.03) N = 364 0.02 (− 0.08, 0.12) N = 366 0.02 (− 0.17, 0.19) N = 122

TC

HDL

FEV1

0.07 (− 0.03, 0.17) N = 363 − 0.14 (− 0.31, 0.04) N = 120

− 0.26⁎⁎ (− 0.42, − 0.08) N = 121

0.03)

0.01)

0.42⁎⁎⁎ (0.33, 0.50) N = 367 0.27⁎⁎⁎

0.01)

(0.17, 0.36) N = 367 0.23⁎⁎⁎

0.13⁎⁎ (0.03, 0.23) N = 366 0.15⁎⁎

0.06)

(0.13, 0.32) N = 367 0.15⁎⁎⁎ (0.05, 0.25) N = 368 − 0.37⁎⁎⁎

(0.04, 0.24) N = 366 0.26⁎⁎⁎ (0.16, 0.35) N = 367 − 0.35⁎⁎⁎

(− 0.45, − 0.27) N = 364 − 0.05 (− 0.15, 0.05) N = 367 0.18⁎

(− 0.44, − 0.26) N = 363 − 0.09 (− 0.19, 0.01) N = 365 0.27⁎⁎

(0.03, 0.32) N = 161

(0.09, 0.42) N = 122

0.03)

0.13)

0.16)

0.09)

IQ. Body mass index. Waist-to-hip ratio. d Systolic blood pressure. e Diastolic blood pressure. f Total cholesterol (S-Cholesterol). g High density lipoprotein. h Forced expiratory volume, predicted. i Average number of cigarettes smoked per day among smokers. ⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001. b

WHR

0.09 (− 0.01, 0.19) N = 365 − 0.03 (− 0.13, 0.07) N = 367 0.23⁎ (0.05, 0.39) N = 122

K. Modig, L.R. Bergman / Intelligence 40 (2012) 82–90

Systolic bpd

M − 0.03 24.43

K. Modig, L.R. Bergman / Intelligence 40 (2012) 82–90

87

Table 3 Odds ratios for health risks with their 95% confidence intervals within parenthesis for the relation of a 1 SD decrease in IQz score with health risks in middle age. Results from logistic regression analyses with IQz as the independent variable. Dependent variables

Unadjusted OR

OR adjusted for education

1.13 (0.94, 1.35) 1.48⁎ (1.05, 2.08) 1.20 (0.96, 1.50) 1.38⁎ (1.02, 1.88) 1.14 (0.72, 1.79) 1.46 (0.97, 2.19) 1.12 (0.90, 1.40) 1.46 (0.65, 3.31) 1.41 (0.97, 2.05) 1.68⁎ (1.20, 2.37)

1.06 1.51 1.20 1.45 1.26 1.19 1.03 1.20 1.21 1.31

(0.87, (1.08, (0.94, (1.03, (0.76, (0.75, (0.53, (0.49, (0.80, (0.90,

1.29) 2.12) 1.54) 2.03) 2.07) 1.87) 1.45) 2.90) 1.83) 1.92)

1.25⁎ (1.05, 1.50) 0.82 (0.65, 1.04) 0.90 (0.66, 1.22) 0.87 (0.65, 1.17) 1.22 (0.91, 1.65) 1.03 (0.87, 1.22) 1.04 (0.88, 1.24)

1.03 0.91 0.75 0.93 1.26 1.08 1.06

(0.84, (0.70, (0.53, (0.67, (0.91, (0.90, (0.87,

1.26) 1.18) 1.06) 1.28) 1.74) 1.31) 1.28)

Indicators of health status Being overweight (BMI > 25) Being obese (BMI > 30) Waist-to-hip ratio > 85 Systolic bp > 140 mm Hg Diastolic bp > 90 mm Hg FEV1% b 80 S-Cholesterol > 5 S-HDL b 1 Self-reported health as bad GP health status unsatisfactory Indicators of health behavior Ever daily smoker Given up smoking Smoked at least 10 cigarettes per day Regulara gynecological examinations Change of life style as a result the medical examination Thinking about eating healthy Exercise regularly at least two times a week Note. ⁎ p b .05. a Compared to those who do not go at all.

what would typically be expected (Bloom, 1964; Fagan, Holland, & Wheeler, 2007; Larsen, Hartmann, & Nyborg, 2008). This is somewhat puzzling since the Swedish tests we used are broad IQ tests with quite high reliabilities (well over 0.90). Two tentative explanations are (1) a different test was used at age 15 than at age 10 and 13 and (2) at age 13 and 15 the girls are at the beginning/end of their biological growth spurt and the biological age vary considerably between the girls in spite of them being of approximately the same chronological age. Hence, biological age variation might have lowered the correlations. 4.1. IQ and indicators of health — comparison with other studies We found a relatively strong inverted association between intelligence level in adolescence and the odds of being obese in adulthood. The association remained statistically significant, even after adjustment for educational level. An association of IQ with overweight or obesity has been found in previous studies (Batty, Deary, & Macintyre, 2007; Batty, Deary, Schoon, & Gale, 2007; Chandola et al., 2006; Halkjaer et al., 2003b; Lawlor et al., 2006), but adjustment for education or adult socioeconomic position resulted in loss of statistical significance in all but one of these studies (Batty, Deary, & Macintyre, 2007). Only one of the studies analyzed men and women separately with the result of a stronger association between intelligence and obesity for women (Chandola et al., 2006). In the other studies the analyses sometimes included women but they were analyzed together with men with adjustment for gender in the model and thus it is difficult to know whether the associations were different for men and women. One study did not find any significant correlation of early IQ score and later BMI, treated as a continuous variable (Hart et al., 2004), in line with our results, suggesting that the association between intelligence and weight status is not linear. There are a few studies of the relationship between intelligence and the metabolic syndrome

(Batty, Shipley, et al., 2008; Richards, Mishra, Gale, Deary, & Batty, 2009). In the first study an association was found but the cohort was a selected group of former Vietnam-era personnel and consisted only of men. In the second study, childhood IQ was found to be inversely associated with risk of the metabolic syndrome, but the association was almost entirely mediated by educational attainment and achieved occupational social class (Richards et al., 2009). The study included both men and women; they were not separated in the analyses. IQ was inversely related to high systolic blood pressure. This result is supported by some previous studies but not all. Differences between men and women and between types of data (self-reported data vs. medical examinations) might be two explanations. Analyzing blood pressure as a continuous variable might be another as our results from the logistic regression indicate a non-linear association between intelligence and systolic blood pressure since a significant association was found only when setting a cutoff-level for high blood pressure and not when analyzing continuous data. This is in contrast to a previous study by Hart et al. (2004), however they analyzed men and women together. A previous Swedish study analyzing the relationship between IQ and hypertension among 379 men found that the hypertensive men had lower IQ scores than the healthy group (Lindgarde, Furu, & Ljung, 1987). However, due to the cross-sectional study design it is difficult to draw any conclusions about the causal direction of the relationship. Batty and colleagues found no associations of IQ score at age 11 and hypertension in middle age, neither in men nor women (Batty, Deary, & Macintyre, 2007). In their study, hypertension was self-reported and based on whether a physician had informed the patient if they suffered from hypertension, which is a questionable measure. In another study of former Vietnam army personnel, men only, low IQ was associated with high systolic and diastolic blood pressure (Batty, Shipley, et al., 2008), even if the differences were small and not incremental across the

88

K. Modig, L.R. Bergman / Intelligence 40 (2012) 82–90

IQ-distribution. Another study looked at low verbal ability or low processing speed and the use of antihypertensive medication but found no associations. The analyses were adjusted for gender but not separated by gender (Anstey & Sachdev, 2009). When the logistic regression model was not adjusted for education a significant and rather strong inverse association was found between IQ and the physician's ratings of the women's health status as unsatisfactory. This was expected since the rating measured a broad spectrum of health problems and, if there is a weak effect of IQ on many health factors, the broad rating might amplify this tendency. We found no statistically significant association between intelligence and cholesterol levels. This is not a well-studied area but Hart and colleagues reported results in line with our null-findings (Hart et al., 2004). Regarding lung capacity, one study found adolescent cognitive ability to be positively associated with FEV1, although not with rate of decline in FEV1 from 43 to 53 years, which was largely explained by socioeconomic position (Richards, Strachan, Hardy, Kuh, & Wadsworth, 2005). In the only study we are aware of that looked at IQ and FEV1% predicted (like we did), a significant correlation of 0.15 was found (Hart et al., 2004). This is in contrast to our result where we did not find any significant association. The results in their study were, however, based on both men and women and they provided no information of whether the correlations differed by gender.

4.2. IQ and indicators of health behaviors — comparison with other studies Several studies have found associations between intelligence and smoking status (Anstey & Sachdev, 2009; Batty, Deary, Schoon, & Gale, 2007; Batty, Shipley, et al., 2008; Hemmingsson et al., 2008). We did, however, find in a previous twin study that the association between intelligence and smoking status seemed to be confounded by early environmental factors (Wennerstad et al., 2010) since no association was found within pairs of twins. In the same study (based on men only), we found a stronger association of low IQ with current than past smoking, suggesting that among smokers, intelligence might be positively related to giving up smoking. This finding was not supported in the present study, based on women, where no association between IQ and giving up smoking was found, and the finding was not supported by a study by Anstey and Sachdev (2009). Further, we found no statistically significant association between IQ and number of cigarettes smoked per day. This is supported by the result in a previous study where no association was found between IQ and nicotine dependence (Modig, Silventoinen, Tynelius, Kaprio, & Rasmussen, 2011). The association between intelligence and physical activity has been sparsely studied. Two studies found some support for intelligence to be positively correlated with higher levels of physical activity (Anstey & Sachdev, 2009; Singh-Manoux et al., 2009) but these studies included both men and women in the analyses, although adjusted for gender. The relationships were also rather small. Further, in the second study IQ was measured at the same time as physical activity, raising questions about reverse causality. Our results gave no evidence of an association between IQ and physical activity among Swedish women.

4.3. The mechanism behind the inverted relationship between IQ and health risks The association between early intelligence and later mortality risk is consistent among men whereas conflicting findings have been reported regarding women with some studies showing associations (Jokela, Batty, Deary, Gale, & Kivimaki, 2009; Jokela, Elovainio, Singh-Manoux, & Kivimaki, 2009; Kuh et al., 2009; Leon, Lawlor, Clark, Batty, & Macintyre, 2009) and others showing no associations (Kuh et al., 2004; Pearce et al., 2006), including a study of Swedish women (Lager et al., 2009). If the association between intelligence and mortality only holds for men, it raises questions about the underlying mechanisms. For example, the previously mentioned system integrity theory, suggesting confounding by genetic and/or environmental factors, such as the fetal environment, is unlikely to be an important explanation of the relationship since, if true, the effects it predict would likely operate in a similar way for both genders. Mediation by health behaviors, however, might still be a plausible explanation if it is related to the IQ-level among men but not women, or if the consequences of these behaviors for health outcomes differ by gender. In order to receive information about the underlying mechanism of intelligence and health outcomes, future long-term longitudinal studies are needed that examine the whole chain if relationships between intelligence, health behaviors, and health outcomes, and where men and women are analyzed separately. Preferably, intelligence should then be measured in childhood before streaming occurs in school and health outcomes should be measured in old age when morbidity and mortality have become prevalent. 4.4. Limitations and strengths Even though the sample size of the present study was moderate, the quality of the data was high. It is a strength that the present study was based on long-term longitudinal data with an almost complete set of variables for both health behaviors and indicators of health and that, in contrast to most studies, women were studied. Most of the data for the health indicators were also extracted from medical examinations and not self-reported. It is a limitation that the health indicators were measured at a comparatively young age (age 43) when most health problems have not yet emerged and it is possible that stronger relationships with IQ would have been found had the measurements been taken at a higher age. On the other hand, concerning health-related behaviors it is a strength that they were measured in midlife because it is probable that the presumed causal effect of them should operate during the decades preceding the emergence of the major diseases in old age. In another study based on the present sample it was also found that the four-year stability of health-related behaviors was high (Benzies, Wångby, & Bergman, 2008). At older ages they might however change as an effect of health which is another advantage of looking at health behaviors in midlife. For two reasons it is a strength that IQ was measured already in adolescence. Firstly, the risk of an inversed causal relationship between IQ and health-related factors is strongly reduced. Secondly, the confounding of IQ and education is reduced because the well-documented effect of education on

K. Modig, L.R. Bergman / Intelligence 40 (2012) 82–90

IQ (Cliffordson & Gustafsson, 2008) is less of a problem in the present study: IQ was measured before any streaming in school had occurred. The moderate sample size in the present study lowered the precision of the estimates and reduced the power to find significant true relationships. However, examining the correlation coefficients between IQ and the continuous health variables and their confidence intervals lead us to the conclusion that it is unlikely that any true substantial associations exist. 5. Conclusion This study finds no support for a sizable association between intelligence in adolescence and indicators of health and health behavior in midlife, measured by a range of indicators, among Swedish women, with the exception for an association with low intelligence and the odds of being obese and having a high systolic blood pressure. The proposed mediating effect of health behavior on the IQ–mortality association is not supported by the results of this study, neither the idea of a general sizeable correlation of intelligence and health status. Acknowledgment This study was made possible by access to data from the longitudinal research program Individual Development and Adaptation. The scientific leader is Lars R. Bergman. Responsible for the planning, implementation and financing of the collection of data before 1996 was David Magnusson and after that time Lars R. Bergman. The data collections were supported by grants from the Swedish National Board of Education, the Swedish Committee for the Planning and Coordination of Research, The Bank of Sweden Tercentenary Foundation, the Swedish Social Research Council, and the Örebro County Council. References Anstey, K. J. L. L. F. C. H., & Sachdev, P. (2009). Level of cognitive performance as a correlate and predictor of health behaviours that protect against cognitive decline in late life: The path through life study. Intelligence, 37, 600–606. Backteman, G., & Magnusson, D. (1981). Longitudinal stability of personalitycharacteristics. Journal of Personality, 49(2), 148–160. Batty, G. D., Deary, I. J., & Gottfredson, L. S. (2007). Premorbid (early life) IQ and later mortality risk: Systematic review. Annals of Epidemiology, 17(4), 278–288. Batty, G. D., Deary, I. J., & Macintyre, S. (2007). Childhood IQ in relation to risk factors for premature mortality in middle-aged persons: The Aberdeen Children of the 1950s study. Journal of Epidemiology and Community Health, 61(3), 241–247. Batty, G. D., Deary, I. J., Schoon, I., & Gale, C. R. (2007). Mental ability across childhood in relation to risk factors for premature mortality in adult life: The 1970 British Cohort Study. Journal of Epidemiology and Community Health, 61(11), 997–1003. Batty, G. D., Gale, C. R., Mortensen, L. H., Langenberg, C., Shipley, M. J., & Deary, I. J. (2008). Pre-morbid intelligence, the metabolic syndrome and mortality: The Vietnam experience study. Diabetologia, 51(3), 436–443. Batty, G. D., Kivimaki, M., & Deary, I. J. (2010). Intelligence, education, and mortality. BMJ, 340, c563. Batty, G. D., Shipley, M. J., Mortensen, L. H., Boyle, S. H., Barefoot, J., Gronbaek, M., et al. (2008). IQ in late adolescence/early adulthood, risk factors in middle age and later all-cause mortality in men: The Vietnam experience study. Journal of Epidemiology and Community Health, 62(6), 522–531. Batty, G. D., Wennerstad, K. M., Smith, G. D., Gunnell, D., Deary, I. J., Tynelius, P., et al. (2007). IQ in early adulthood and later cancer risk: Cohort study of one million Swedish men. Annals of Oncology, 18(1), 21–28.

89

Batty, G. D., Wennerstad, K. M., Smith, G. D., Gunnell, D., Deary, I. J., Tynelius, P., et al. (2009). IQ in early adulthood and mortality by middle age: Cohort study of 1 million Swedish men. Epidemiology, 20(1), 100–109. Benzies, K. M., Wångby, M., & Bergman, L. R. (2008). Stability and change in health-related behaviours of midlife Swedish women. Health Care for Women International, 29, 997–1018. Bergman, L. R. (1973). Parents' education and mean change in intelligence. Scandinavian Journal of Psychology, 14, 273–281. Bergman, L. R. (2000). Women's health, work, and education in a life-span perspective. Technical report 1: Theoretical background and overview of the data collection. IDA Report series, No. 71.: Stockholm University, Department of Psychology. Sweden. Bergvall, N., Iliadou, A., Johansson, S., Tuvemo, T., & Cnattingius, S. (2006). Risks for low intellectual performance related to being born small for gestational age are modified by gestational age. Pediatrics, 117(3), e460–e467. Bloom, B. S. (1964). Stability and change in human characteristics. New York: Wiley & Sons. Carlsson, A. C., Theobald, H., Hellenius, M. L., & Wandell, P. E. (2009). Cardiovascular and total mortality in men and women with different blood pressure levels — A 26-year follow-up. Blood Pressure, 18(3), 105–110. Carrol, J. B. (1993). Human cognitive abilities: A survey of factor-analytic studies. New York: Cambridge University Press. Chandola, T., Deary, I. J., Blane, D., & Batty, G. D. (2006). Childhood IQ in relation to obesity and weight gain in adult life: The National Child Development (1958) Study. International Journal of Obesity, 30(9), 1422–1432. Cliffordson, C., & Gustafsson, J. -E. (2008). Effects of age and schooling on intellectual performance: Estimates obtained from analysis of continuous variation in age and length of schooling. Intelligence, 36, 143–152. Deary, I. J. (2005). Intelligence, health and death: The new field of cognitive epidemiology. The Psychologist, 18, 610–613. Deary, I. J. (2009). Introduction to the special issue on cognitive epidemiology. Intelligence, 37, 517–519. Deary, I. J., & Batty, G. D. (2007). Cognitive epidemiology. Journal of Epidemiology and Community Health, 61(5), 378–384. Deary, I. J., Weiss, A., & Batty, D. (2010). Intelligence and personality as predictors of illness and death: How researchers in differential psychology and chronic disease epidemiology are collaborating to understand and address health inequalities. Psychological Science in the Public Interest, 11(2), 53–79. Der, G., Batty, G. D., & Deary, I. J. (2009). The association between IQ in adolescence and a range of health outcomes at 40 in the 1979 US National Longitudinal Study of Youth. Intelligence, 37(6), 573–580. Fagan, J. F., Holland, C. R., & Wheeler, K. (2007). The prediction, from infancy, of adult IQ and achievement. Intelligence, 35(3), 225–231. Gardner, H. (1993). Multiple intelligences: The theory in practice. New York: Basic Books. Gottfredson, L. S. (2004). Intelligence: Is it the epidemiologists' elusive “fundamental cause” of social class inequalities in health? Journal of Personality and Social Psychology, 86(1), 174–199. Gruffydd-Jones, K. L. C. (2011). The 2010 NICE COPD Guidelines: How do they compare with the GOLD guidelines? Primary Care Respiratory Journal, 20 (x) Ahead of print. Halkjaer, J., Holst, C., & Sorensen, T. I. (2003). Intelligence test score and educational level in relation to BMI changes and obesity. Obesity Research, 11(10), 1238–1245. Hart, C. L., Taylor, M. D., Smith, G. D., Whalley, L. J., Starr, J. M., Hole, D. J., et al. (2004). Childhood IQ and cardiovascular disease in adulthood: Prospective observational study linking the Scottish Mental Survey 1932 and the Midspan studies. Social Science & Medicine, 59(10), 2131–2138. Hemmingsson, T., Kriebel, D., Melin, B., Allebeck, P., & Lundberg, I. (2008). How does IQ affect onset of smoking and cessation of smoking — Linking the Swedish 1969 conscription cohort to the Swedish survey of living conditions. Psychosomatic Medicine, 70(7), 805–810. Jensen, A. R. (1998). The g-factor: The science of mental ability. : Westport: CT: Greenwood. Jokela, M., Batty, G. D., Deary, I. J., Gale, C. R., & Kivimaki, M. (2009). Low childhood IQ and early adult mortality: The role of explanatory factors in the 1958 British birth cohort. : Pediatrics. Jokela, M., Elovainio, M., Singh-Manoux, A., & Kivimaki, M. (2009). IQ, socioeconomic status, and early death: The US National Longitudinal Survey of Youth. Psychosomatic Medicine, 71(3), 322–328. Kaminsky, D. A., Savage, P. D., Callas, P. W., & Ades, P. A. (2010). Lung function and cardiovascular risk: Effect of cardiac rehabilitation. Journal of Cardiopulmonary Rehabilitation and Prevention, 30(6), 384–390. doi: 10.1097/HCR.0b013e3181d6f976. Kubicka, L., Matejcek, Z., Dytrych, Z., & Roth, Z. (2001). IQ and personality traits assessed in childhood as predictors of drinking and smoking behaviour in middle-aged adults: A 24-year follow-up study. Addiction, 96(11), 1615–1628.

90

K. Modig, L.R. Bergman / Intelligence 40 (2012) 82–90

Kuh, D., Richards, M., Hardy, R., Butterworth, S., & Wadsworth, M. E. (2004). Childhood cognitive ability and deaths up until middle age: A post-war birth cohort study. International Journal of Epidemiology, 33(2), 408–413. Kuh, D., Shah, I., Richards, M., Mishra, G., Wadsworth, M., & Hardy, R. (2009). Do childhood cognitive ability or smoking behaviour explain the influence of lifetime socio-economic conditions on premature adult mortality in a British post war birth cohort? Social Science & Medicine, 68(9), 1565–1573. Lager, A., Bremberg, S., & Vagero, D. (2009). The association of early IQ and education with mortality: 65 year longitudinal study in Malmo, Sweden. BMJ, 339, b5282. Lager, A., Bremberg, S., & Vagero, D. (2010). Intelligence and mortality. Only ignorance stops progress. BMJ, 340, c2765. Larsen, L., Hartmann, P., & Nyborg, H. (2008). The stability of general intelligence from early adulthood to middle age. Intelligence, 36(1), 29–34. Lawlor, D. A., Clark, H., Davey Smith, G., & Leon, D. A. (2006). Childhood intelligence, educational attainment and adult body mass index: Findings from a prospective cohort and within sibling-pairs analysis. International Journal of Obesity, 30(12), 1758–1765. Leon, D., Lawlor, D. A., Clark, H., Batty, G. D., & Macintyre, S. (2009). The association of childhood intelligence with mortality risk from adolescence to middle age: Findings from the Aberdeen Children of the 1950s cohort study. Intelligence, 37(6), 520–528. Lindgarde, F., Furu, M., & Ljung, B. O. (1987). A longitudinal study on the significance of environmental and individual factors associated with the development of essential hypertension. Journal of Epidemiology and Community Health, 41, 220–226. Magnusson, D. (1988). Individual development from an interactional perspective: A longitudinal study. In D. Magnusson (Ed.), Paths through life, Vol. 1, Hillsdale, NJ: Erlbaum. Modig, K., Silventoinen, K., Tynelius, P., Kaprio, J., & Rasmussen, F. (2011). Genetics of the association between intelligence and nicotine dependence: A study of male Swedish twins. Addiction, 106(5), 995–1002.

Pearce, M. S., Deary, I. J., Young, A. H., & Parker, L. (2005). Growth in early life and childhood IQ at age 11 years: The Newcastle Thousand Families Study. International Journal of Epidemiology, 34(3), 673–677. Pearce, M. S., Deary, I. J., Young, A. H., & Parker, L. (2006). Childhood IQ and deaths up to middle age: The Newcastle Thousand Families Study. Public Health, 120(11), 1020–1026. Richards, M. B. S., Mishra, G., Gale, C., Deary, I. J., & Batty, G. D. (2009). IQ in childhood and the metabolic syndrome in middle age: Extended followup of the 1946 British Birth Cohort Study. Intelligence, 37, 567–572. Richards, M., Strachan, D., Hardy, R., Kuh, D., & Wadsworth, M. (2005). Lung function and cognitive ability in a longitudinal birth cohort study. Psychosomatic Medicine, 67(4), 602–608. Singh-Manoux, A., Sabia, S., Kivimaki, M., Shipley, M., Ferrie, J. E., & Marmot, M. (2009). Cognition and incident coronary heart disease in late midlife: The Whitehall II study. Intelligence, 37, 529–534. Sternberg, R. J. (1985). Beyond IQ: A triarchic theory of intelligence. Cambridge: Cambridge University Press. Wandell, P. E., Carlsson, A. C., & Theobald, H. (2009). The association between BMI value and long-term mortality. International Journal of Obesity, 33(5), 577–582. Wennerstad, K. M., Silventoinen, K., Tynelius, P., Bergman, L., Kaprio, J., & Rasmussen, F. (2010). Associations between IQ and cigarette smoking among Swedish male twins. Social Science & Medicine, 70(4), 575–581. Wood, D., De Backer, G., Faergeman, O., Graham, I., Mancia, G., & Pyorala, K. (1998). Prevention of coronary heart disease in clinical practice: recommendations of the Second Joint Task Force of European and other Societies on Coronary Prevention. Atherosclerosis, 140(2), 199–270. Wormser, D., Kaptoge, S., Di Angelantonio, E., Wood, A. M., Pennells, L., Thompson, A., et al. (2011). Separate and combined associations of bodymass index and abdominal adiposity with cardiovascular disease: Collaborative analysis of 58 prospective studies. Lancet, 377(9771), 1085–1095. Yu, Z. B., Han, S. P., Cao, X. G., & Guo, X. R. (2010). Intelligence in relation to obesity: A systematic review and meta-analysis. Obesity Reviews, 11(9), 656–670.