Stroke Is Predicted by Low Visuospatial in Relation to Other Intellectual Abilities and Coronary Heart Disease by Low General Intelligence

Background Low intellectual ability is associated with an increased risk of coronary heart disease and stroke. Most studies have used a general intelligence score. We studied whether three different subscores of intellectual ability predict these disorders. Methods We studied 2,786 men, born between 1934 and 1944 in Helsinki, Finland, who as conscripts at age 20 underwent an intellectual ability test comprising verbal, visuospatial (analogous to Raven's progressive matrices) and arithmetic reasoning subtests. We ascertained the later occurrence of coronary heart disease and stroke from validated national hospital discharge and death registers. Results 281 men (10.1%) had experienced a coronary heart disease event and 131 (4.7%) a stroke event. Coronary heart disease was predicted by low scores in all subtests, hazard ratios for each standard deviation (SD) lower score ranging from 1.21 to 1.30 (confidence intervals 1.08 to 1.46). Stroke was predicted by a low visuospatial reasoning score, the corresponding hazard ratio being 1.23 (95% confidence interval 1.04 to 1.46), adjusted for year and age at testing. Adjusted in addition for the two other scores, the hazard ratio was 1.40 (1.10 to 1.79). This hazard ratio was little affected by adjustment for socioeconomic status in childhood and adult life, whereas the same adjustments attenuated the associations between intellectual ability and coronary heart disease. The associations with stroke were also unchanged when adjusted for systolic blood pressure at 20 years and reimbursement for adult antihypertensive medication. Conclusions Stroke is predicted by low visuospatial reasoning scores in relation to scores in the two other subtests. This association may be mediated by common underlying causes such as impaired brain development, rather than by mechanisms associated with risk factors shared by stroke and coronary heart disease, such as socio-economic status, hypertension and atherosclerosis.


Introduction
Children and young adults who gain lower scores in intellectual ability tests have shorter life expectancy [1,2,3,4,5,6,7] and a higher risk of coronary heart disease in later life [8,9,10,11,12,13,14]. Recent studies suggest that they also have a higher risk of stroke [8,10,14,15]. In some studies this association is at least as powerful as that for coronary heart disease, whereas other studies have shown a weaker association [8] or none at all [9]. Whether the predictive ability of low intellectual ability for coronary heart disease is different from that for stroke is potentially important because such differences may help to understand the mechanisms that link intellectual ability with disease.
Most previous studies have used a general measure of intellectual ability. While the predictive validity of the concept of ''general intelligence'' is well established, intelligence consists of different abilities which may in part have different developmental origins [16], different neural correlates [17] and different predictive value on educational and occupational outcomes [18,19]. We studied whether verbal, visuospatial and arithmetic abilities, tested during compulsory military service in young adulthood, predict coronary heart disease and stroke in men now aged over 60 years.

Ethics statement
The Helsinki Birth Cohort Study has been approved by the Ethics Committee of the National Public Health Institute, and military service data were linked with permission from the Finnish Defence Command. As only data from these registers were used, no individual consent was required.

Subjects
The study population comes from the Helsinki Birth Cohort which includes 8760 subjects (4630 men and 4130 women) who were born at Helsinki University Central Hospital between 1934 and 1944 and who attended child welfare clinics in Helsinki. The cohort has been described in detail elsewhere [20]. We obtained permission to link data collected within the Helsinki Birth Cohort Study (HBCS) with data from compulsory military service for these 4630 men, of whom 2786 (60.2%) had intellectual ability test data [1,16,21,22]. The main reason for data unavailability was that the test was gradually introduced in different army regiments from 1955 onwards and conscripts in some regiments did not undergo the test. We have previously reported a detailed nonparticipation analysis showing that availability of the ability test data was not related to father's occupational status, body size during military service, or maximum achieved education later in life [1].
We defined childhood socio-economic status based on the father's occupation as described [1]. Data for each subject's educational attainment, occupational and marital status were obtained from Statistics Finland for 5-yearly intervals between 1970 and 2000. Maximal educational attainment was grouped into four groups according to the UNESCO classification and maximum achieved occupational status into four groups based on a classification used by Statistics Finland [1]. Data for household taxable income came from Statistics Finland, based on 1980 census. We have previously described the associations between these socio-economic indicators and intellectual ability test scores [1].

Assessment of intellectual ability at the start of military service
The Finnish Defence Forces Basic Ability Test was developed at the Finnish Defence Forces Education Development Centre. This obligatory test was introduced from 1955 onwards and was given to new recruits during the first 2 weeks of service. It comprises verbal, visuospatial and arithmetic subtests, each composed of 40 multiple-choice questions increasing in difficulty, and has been described elsewhere [16,23]. Briefly, in the verbal reasoning subtest the subject chooses synonyms, antonyms or words belonging or not belonging to the same category as given words, and similar relationships between two word pairs. In the arithmetic reasoning subtest, the subject completes series of numbers, solves short problems, and computes simple arithmetic operations. The visuospatial reasoning task is analogous to the widely used Raven's Progressive Matrices [24] and consists of a set of matrices with one removed part. The subject is asked to decide which of the given single figures completes the matrix. We collected the test results for each subject from microfilms. The data also included the subject's height, weight and systolic blood pressure. BMI was calculated as kg/m 2 . Systolic blood pressure was missing for 165 men (5.9%), who were excluded from the regression models that included blood pressure.

Coronary heart disease and stroke
Using the personal identification number of each subject we identified deaths and hospital discharges for coronary heart disease and stroke during 1971-2003 from validated [25,26,27,28] national death and hospital discharge registers; the men were thus 26 to 69 years old during the follow-up. We used the first-ever coronary heart disease event and first-ever stroke event for each subject. For stroke subtypes (hemorrhagic/ischemic), however, we used the type of the first event to avoid uncertainty of diagnosis in subjects who later have a diagnosis of a different stroke subtype. Diagnoses were recorded according to the International Classification of Diseases (ICD). The codes for coronary heart disease were 410-414 in ICD-8 and 9 and I21--I25 in ICD 10 [20]. The codes for stroke were of 430-434 and 436-437 in ICD-8, 430-434 and 436-438 in ICD9; and I60-I69 in ICD-10. For thrombotic stroke, they were 432-434 and 436-437 in ICD 8, 433-434 and 436 in ICD-9, and I63-I64 in ICD 10; for hemorrhagic stroke 430-431 in ICD-8, 430-432 in ICD-9, and I60-I62 in ICD-10 [25]. We adjusted the analyses for hypertension diagnosed before first outcome event using the Social Insurance Institution's Register of people who receive special reimbursement for medication for the disease [29].

Data analysis
We converted the ability test results to z scores (SD units) so that the mean z score of the study cohort was zero. Correlation coefficient between the verbal and visuospatial z scores was 0.62, between verbal and arithmetic 0.67, and between visuospatial and arithmetic 0.68. To illustrate the associations with the visuospatial z scores in relation to the two other scores, we included all scores simultaneously in the regression model. To illustrate general intelligence, we calculated the mean of all z scores. We examined these z scores as predictors of coronary heart disease and stroke using Cox proportional hazards model, stratified according to year of birth. As in our previous analyses [1], we adjusted all hazard ratios for year and the subject's age at testing. In line with previous work from this [1] and other cohorts [2,3,4,5,8,9,10,13], we performed additional adjustments for markers of childhood environment (father's occupational status and the subject's height and BMI at the start of military service) and adult socio-economic status (education, occupational status and income). We tested the proportionality assumption of the Cox model by using a timevarying indicator variable, determining whether the hazard ratio for intellectual ability subscore was the same at different ages. None of these time-varying indicators were statistically significant (p.0.08). We divided the ability test scores into fourths to illustrate the graded nature of the relationships; for trends of hazard ratios, the scores were used as continuous variables. We assessed interactions between the effects of socio-economic status and intellectual ability on stroke by including a product term in the model.

Results
Of the 2786 men, 281 (10.1%) had experienced a coronary heart disease event. Stroke had been diagnosed in 131 (4.7%) men, of whom 39 had hemorrhagic and 81 thrombotic stroke (one subject had both diagnoses), and 12 an unspecified type of stroke. Mean ages at diagnoses and numbers or deaths are shown in Table 1.
Verbal, visuospatial and arithmetic reasoning subscores as predictors of coronary heart disease and stroke Table 2 shows that men who had lower scores in intellectual ability tests were more likely to get coronary heart disease. The risk associated with each of the three subtests was similar. Further shown in Table 2 is that the risk of stroke was predicted by low scores in the visuospatial reasoning subtest, whereas verbal and arithmetic reasoning scores were unrelated to the risk of stroke.
Visuospatial reasoning subscores in relation to the other subscores Table 3 (Model 1) illustrates the specific risk associated with low visuospatial reasoning scores in relation to low scores in the two other subtests. Visuospatial scores adjusted for verbal and arithmetic scores predicted stroke (hazard ratio for one SD lower score 1.40; 95% CI 1.10 to 1.79), with similar hazard ratios for hemorrhagic (1.49; 0.94 to 2.36) and thrombotic (1.31; 0.96 to 1.78) stroke, but did not predict coronary heart disease (1.15; 0.97 to 1.36). However, the mean intellectual ability test score predicted coronary heart disease (1.35; 1.18 to 1.54), but not stroke (1.10; 0.90 to 1.35).

Linearity
The hazard ratios across fourths of each subscore (Table 2) showed no evidence of non-linear associations, and no quadratic trend was observed (p values .0.2). An exception was the association of visuospatial reasoning scores with stroke, which suggested a threshold effect with a lower risk in the highest fourth as compared with the three other fourths. We therefore assessed the linearity of this association in more detail. There was no evidence for a quadratic (p = 0.5), third-(p = 0.1) or fourth-degree (p = 0.5) polynomial association. For visuospatial reasoning scores adjusted for the two other scores, hazard ratios from the lowest to highest fourths were 2.0 (95% CI 1.2 to 3.4), 1.3 (0.8 to 2.3), 1.6 (0.9 to 2.7) and 1.0 (referent).

Socio-economic status
In line with previous studies in this and other cohorts [1,2,3,4,5,8,9,10,13], we assessed markers of socio-economic status in childhood and in adulthood as covariates. These factors were showed stronger correlations with mean intellectual ability than with the visuospatial score adjusted for the two other scores. For men who grew up in manual workers', lower middle or upper middle class families, mean z scores were 20.18, 0.23 and 0.45, respectively (r = 0.28; p,0.0001). Corresponding values for visuospatial z score adjusted for the two other scores were 20.02, 0.02 and 0.04 (r = 0.03; p = 0.4). Mean z scores for people who as adults had achieved basic, upper secondary, lower tertiary and upper tertiary education were 20.41, 20.16, 0.48 and 0.82, respectively (r = 0.51; p,0.0001). Corresponding values for visuospatial reasoning z scores adjusted for the other scores were 20.11, 20.02, 0.16 and 0.02 (r = 0.08; p,0.0001), A similar pattern was seen for adult occupational status (r = 0.50; p,0.0001 for mean score and r = 0.11; p,0.0001 for adjusted visuospatial score) and income (r = 0.26; p,0.0001 and r = 0.04; p,0.06). The associations between mean intellectual ability tests scores and coronary heart disease were slightly attenuated when adjusted for father's occupational status; (Table 3, Model 3). They were further attenuated and approached null when adjusted for educational attainment, occupational status and income in adult life. The association of visuospatial reasoning scores adjusted for verbal and arithmetic scores, with stroke was, however, unaffected even after adjustment for all of these socio-economic indicators ( Table 3, Models 4 to 6). Because we have previously shown that the association between intellectual ability and all-cause mortality is stronger among men who grew up in families of low socioeconomic status [1], we also tested for interactions with socioeconomic status. There was no interaction between the effects of any of the socio-economic indicators in childhood or adult life and intellectual ability mean score or the visuospatial score adjusted for the two other scores, except between adult occupational status and the visuospatial score (p = 0.05): among high officials, hazard ratio per one SD higher score was 0.9; 95% CI 0.6 to 1.4; among the remaining occupational groups combined it was 1.7; 1.2 to 2.3.  Table 3. The effect of adjustment for socio-economic status in childhood and adulthood and blood pressure/hypertension on hazard ratios (95% confidence intervals) for coronary heart disease and stroke, per one unit lower z score.

Discussion
Our longitudinal follow-up of male military conscripts had two main findings. First, the pattern of low intellectual abilities that predicted stroke was different from the pattern that predicted coronary heart disease, although both disorders share common risk factors including hypertension and atherosclerosis. While lower scores on verbal, visuospatial and arithmetic reasoning subtests were equally strong predictors of coronary heart disease, stroke was predicted by a profile characterized by lower visuospatial reasoning subtest scores in relation to scores in the two other subtests. Second, while adjustment for indicators of socio-economic status in childhood and adulthood explained a major part of the associations with coronary heart disease, the associations with stroke were little affected by adjustment for socioeconomic variables.

Comparisons with previous studies
The association we found between lower premorbid intellectual abilities and higher rates of coronary heart disease is consistent with previous studies [8,9,10,11,12,13,14]. The hazard ratio of 1.35 for one SD lower mean test score was also comparable to previously reported hazard ratios which vary between 1.25 and 1.4, at least before the age of 65 years [8,9,10,11,12,13,14]. Also consistent with previous studies was that these associations were substantially attenuated when adjusted for socio-economic variables in childhood and adulthood, suggesting that low intellectual ability and low socio-economic status are related to coronary heart disease through similar processes.
A novel finding in our study was that stroke was predicted by visuospatial but not by verbal or arithmetic reasoning abilities. Five previous studies have assessed the relationship between premorbid cognitive abilities and stroke: only one includes subscores assessing different abilities. Two of the studies were based on Scottish data and used the Moray House Test [14] or parts thereof [10]. That test contains diverse items with a preponderance of verbal reasoning but produces a total score without differentiating between different abilities [14,30]. In one of these Scottish studies, the Aberdeen Study with a follow-up until approximately 50 years of age, the test score based on only verbal ability at age 11 years predicted stroke with a hazard ratio of 1.32 for one SD lower scores in men [10]. In the other study based on the whole test battery, the hazard ratio was larger, 1.47, but it was observed only among people diagnosed before 65 years of age when the number of events was relatively small. Among 1.1 million Swedish male conscripts followed up until 28 to 53 years of age, a general intelligence test predicted stroke [15]; transferred to comparable units, the hazard ratio in that study was 1.18 (95% CI 1.16 to 1.21) for one SD lower total IQ score. This was higher than that in the present study although within its confidence interval (1.09; 0.89 to 1.33, for mean score of the three subtests, with corresponding adjustments). In a study of a smaller, partially overlapping Swedish cohort, associations with stroke were similar than in the larger Swedish study and weaker than those for coronary heart disease [8]. Neither of the Swedish studies differentiated between subscores of the test. The only study that did so followed up Danish men who at age 12 underwent a Hä rnqvist test producing verbal, spatial and inductive intelligence scores. Whereas each of these scores predicted coronary heart disease, none of them predicted stroke [9]. The follow-up, however, was restricted to 47 years. While the comparison of previous studies is difficult be differences in intellectual ability tests used and other cohort characteristics including dissimilar ages during follow-up, it is nevertheless safe to conclude from previous work that stroke is predicted by lower intellectual abilities. Our study adds significantly to this conclusion by demonstrating that visuospatial reasoning, as assessed by a Raven's progressive matrices-type test, may be one of the a core elements in this prediction.

Possible mechanisms
Mechanisms that link premorbid intellectual abilities with chronic adult disease have been suggested to include adult socioeconomic advantage, a healthier lifestyle, and a theoretical concept of ''body system integrity'' underlying both intellectual abilities and disease [1,4,13]. Consistent with previous studies [8,11,13,14], we found that much of the relationship with coronary heart disease is explained by adult socio-economic indicators. These relationships have been widely discussed elsewhere [8,11,13,14,31] and we here focus on our findings regarding stroke which we believe reflect the concept of integrity of the body, in particular of the central nervous system.
Among different methods to measure intelligence, Raven's progressive matrices [24] and analogous tests are considered to be closest to the ''g'' factor of intelligence, a concept used to describe the common underlying component of different intellectual abilities [17]. Such tests show stronger correlations with brain volume than other types of tests [32] and are tightly correlated with neurophysiological measures [17] reflecting the speed and reliability of neural transmission and the integrity of central nervous system. Our finding that stroke was predicted exclusively by low scores in this type of test, in particular in relation to the two other subtests, whereas coronary heart disease was equally predicted by low scores assessing different abilities, suggests that the association of intelligence with stroke is likely to be mediated by mechanisms related to the integrity of the central nervous system rather than by mechanisms related to common risk factors of stroke and coronary heart disease such as hypertension and atherosclerosis. Consistent with this, the visuospatial reasoning score of the Finnish Defence Forces test is a strong predictor of another disorder or the central nervous system, schizophrenia, whereas the other two scores are not [23].
Intellectual abilities and stroke could have common underlying risk factors operating during early life, possibly during fetal development. For example, a meta-analysis of 212 studies showed that, although heritability of intellectual abilities was high (around 48%), fetal environment accounted for 20% of the correlation of intelligence between identical twins [33]. In the present cohort, lower visuospatial abilities are predicted by thinness and small head circumference at birth [16]. In the whole source cohort of 13,345 individuals, stroke was also predicted by thinness at birth, although not by head circumference [34].
Men of this cohort who grew up in families of low socioeconomic status, or who as adults had low socio-economic status, had on average lower intellectual ability scores (including visuospatial reasoning) [1]. They also had higher rates of stroke [34]. However, the association we found between visuospatial reasoning in relation to the two other subscores and stroke remained strikingly similar after adjustment for multiple indicators of socio-economic status throughout life. Interaction analyses showed that this association was similar across all socio-economic groups in childhood and in adult life, with the exception that no association was present among men who attained the highest occupational status. We would, however, urge caution in interpreting this finding because of the relatively large number of interaction analysis made. Our findings are paralleled by a recent study showing that whereas low fluid and crystallized intelligence measured at age 70 both predict higher mortality rates, adjustment for socio-economic status attenuates the association with crystallized but not with fluid intelligence [35]. Together these findings again argue for the ''body and central nervous system integrity hypothesis''. In addition, the association between stroke and indicators of early life events, such as exposure to maternal pre-eclampsia [36] or reduced early growth [34], are not affected by adjustment for socio-economic status. This suggests that socio-economic status is unlikely to be a significant common underlying factor or a mediator of the link between low visuospatial reasoning and stroke.

Strengths and limitations
The strengths of this study include the intellectual ability test performed during compulsory military service in early adulthood and the near-comprehensive availability of later-life data through national registers. In validation studies the hospital and death registers show an over 85% agreement for coronary heart disease and stroke against individual hospital records [25,26,27,28]. We have previously discussed the limitations of the Helsinki Birth Cohort Study [20]. The present study was limited to men. In addition, not all men in the HBCS underwent the intellectual ability test. Although a detailed non-participation analysis raised little concern over participation bias [1], the smaller number of participants resulted in decreased power to detect small to moderate effects. For example, while adjustment for socioeconomic status produced little change in the main effects of the hazard ratios, the confidence intervals of visuospatial reasoning subscores predicting stroke overlap one after this adjustment, whereas those of visuospatial reasoning adjusted for the two other subscores remain statistically significant. Moreover, we lack data for some potential confounders such as smoking and use of alcohol. Whereas we have adjusted for blood pressure, a major risk factor for stroke, our data consist of systolic blood pressure measured during military service and special reimbursement for antihypertensive medication, which capture only a part of variation in blood pressure. People with untreated hypertension may be at a similar or even higher risk for stroke than those who receive medication.

Conclusions
We found that coronary heart disease was equally predicted by low scores in subtests measuring different intellectual abilities, while stroke was predicted a profile characterized by lower visuospatial reasoning scores in relation to the two other subscores. The association with stroke was independent of socio-economic status in childhood and over the course of adult life. We suggest that the association with stroke may be mediated through a common underlying factor, possibly an early impairment in brain development.