Triggers of Subarachnoid Hemorrhage

Despite advances in medical and surgical treatments, subarachnoid hemorrhage (SAH) remains a serious disease, with most deaths resulting from the initial severe hemorrhage. Apart from cigarette smoking,¹ knowledge about predisposing risk factors for SAH is limited,² and the pathophysiological events that lead to the rupture of intracerebral aneurysms, the major cause of SAH, are poorly understood.³

The health benefits of regular exercise^4,5 are offset by the hazard of strenuous exertion, which is associated with a temporary increase in the risk of acute myocardial infarction, particularly among inactive individuals.^6,7 A similar effect of physical exertion may hold for SAH⁸ because it is commonly held that the acute event is often precipitated by a transient rise in blood pressure associated with sneezing, coitus, micturition or defecation, extreme physical activity, or a sudden emotional shock.^8,9 However, this perception is based mainly on anecdotal reports and studies of selected series of patients in which issues of bias and confounding have not always been addressed. Although a recent population-based study found the relative risk of sustaining SAH during vigorous activity to be 11.6 and 15.0 in case-control and case-crossover analyses, respectively, the small number of cases and complexities of analysis raise uncertainty about the estimates.¹⁰ To date, no other study has quantified the association between physical exertion (including sexual activity) and SAH or examined other factors that might trigger such an event, including heavy alcohol consumption or cigarette smoking.¹¹

The Australasian Cooperative Research on Subarachnoid Hemorrhage Study (ACROSS) was a large, prospective, multicenter, population-based study of the incidence, risk factors, and outcome of SAH.¹² We report here results of the nested case-crossover study, which aimed to estimate the relative risk of SAH after strenuous physical exertion, cigarette smoking, and binge (alcohol) drinking compared with periods free of, or only light, exposure to these factors.

Patients and Methods

Study Population

The design of ACROSS has been described elsewhere.¹² Briefly, ACROSS involved population-based registers of SAH in 4 major urban centers in Australia and New Zealand in 1995 to 1998. The total study population (age, ≥15 years) was ≈2.8 million according the 1996 census for each city. SAH was defined according to standard criteria as an abrupt onset of severe headache and/or loss of consciousness with or without focal neurological signs for which CT, necropsy, or lumbar puncture revealed focal or generalized blood in the subarachnoid space.¹³ Patients whose hemorrhage was found to originate from sources other than an intracranial aneurysm were excluded, but patients with proven hemorrhage in whom an aneurysm could not be identified either by cerebral angiography or at necropsy were included. Institutional ethics committees in each study center approved the protocol, and all participants (or next of kin for case subjects who were severely ill, unconscious, or deceased) provided written, informed consent.

Ascertainment of Exposure

ACROSS had a nested case-crossover study, an epidemiological technique used to assess the short-term effects of transient exposures on the risk of onset of acute events.^6,7,14–16 With this method, each subject serves as his or her own control, with bias and confounding resulting from age, sex, medical history, and other stable within-person factors thereby eliminated. Given that some patients may experience a warning headache¹⁷ resulting from early bleeding, we defined the time of onset of SAH as the time of the most severe symptoms.

As soon as possible after notification, the study nurses undertook face-to-face interviews with subjects or the partner or next of kin if the subject was deceased or disabled. A structured questionnaire was used to obtain information regarding demographics; clinical, medical, and family histories; and exposures of interest. Data were obtained on the intensity and timing of key exposures—physical activity, alcohol consumption, and cigarette smoking—over the 26 hours before the onset of SAH (ie, for the 2-hour hazard period immediately before and the preceding 24 hours). In the first year of the study, information on exposure was collected for each 6-hour interval in the preceding 26 hours, including the 2-hour hazard period immediately before the SAH. However, the questionnaire was refined for the second year of the study to allow the exposures to be recorded more precisely for each of the 2-hour intervals over the 26 hours before the SAH. Subjects were shown examples of types of physical activity on a card and asked to estimate the total minutes they spent engaged at various levels for a given type of activity. Physical activity was then assessed on a scale between 1 and 3 metabolic equivalents (METs) using generally accepted MET values⁶ and categorized as follows: sleeping (1 MET), sitting (2 MET), light exertion (3 to 4 MET), moderate exertion (5 MET), and extreme exertion (6 to 8 MET). Subjects were asked whether they smoked cigarettes or drank alcohol before the SAH; for each prespecified hazard period, data were recorded on the number of cigarettes smoked and the number of standard drinks of alcohol consumed in the preceding 26 hours.

In addition, subjects were asked questions to ascertain their usual frequency of low-intensity and high-intensity (moderate to extreme) physical activity during the 2 weeks preceding the event: (1) “During the course of your work, activities around the house, or recreational activities, did you have periods of 15 minutes or more in which you found yourself continuously walking or on the move,” and (2) “during the course of your work, activities around the house, or recreational activities, did you have periods of 15 minutes or more in which you found yourself sweating, and puffing and panting, as a result of your activity?” These questions were adapted from those used in health surveys conducted by the Australian Bureau of Statistics.¹⁸ Subjects were thus categorized as sedentary or active (on the basis of either high- or low-intensity activities at least 1 time per week). They were also questioned about their lifetime use of tobacco (never, current, or ex-smoker >12 months) and about their usual weekly use of alcohol in the previous year (number of standard drinks per week).

Statistical Analyses

The hypotheses to be tested were that the following exposures in the predefined 2-hour hazard interval were each associated with an increased risk of SAH: (1) a peak exertion estimated to be at least 6 MET (a level that is routinely used in studies of coronary events), (2) consumption of ≥4 standard drinks (ie, ≥40 g ethanol), or (3) smoking ≥4 cigarettes. In regard to physical exertion, however, too few cases had undertaken activities at ≥6 MET to allow meaningful statistical analysis, so the exposure level was subsequently changed to moderate to extreme, defined as from 5 to 8 MET. The primary statistical analysis used the pair-matched interval method of the case-crossover technique (model 1)^14,15 in which each person’s exposure in the defined 2-hour hazard period immediately before the SAH was contrasted with their exposure during the comparable 2-hour period at the same time on the day before the event (the Figure). These data were checked for robustness with a supplementary approach to the case-crossover technique that uses an alternative source of control data. In the nonparametric multiple intervals method (model 2), exposure in the hazard period was contrasted with each of the 2-hour control periods in the 26 hours preceding the SAH (the Figure). This model takes into account circadian variation in SAH¹⁹ and includes time of day as a potential confounding variable. However, this method was used only for the subset of the study population that had complete data on all 2-hour control periods in the 26 hours preceding the SAH.

• Download figure
• Download PowerPoint

The sample size of 432 provided 80% power and 2-tailed significance (α=0.05 for detecting relative risks of 2.0 attendant on exposures in 20% of the population). Analyses were undertaken with SAS 6.12,²⁰ S-PLUS,²¹ and LogXact software.²² Differences between groups were examined with the χ² test for categorical variables and independent-sample t tests for continuous variables. Odds ratios (ORs) as estimates of relative risks and 95% confidence intervals (CIs) were calculated with the Mantel-Haenszel approach²³ and conditional logistic regression analysis.²⁴ Modifications of risks were assessed by comparing different subgroups, with particular attention given to known risk factors for SAH. Population-attributable risk,²⁵ a measure of the proportion of total cases of SAH associated with a risk factor in the total population, and corresponding 95% CIs²⁶ were also calculated.

Results

Study Population

Table 1 gives the characteristics of the 432 patients with first-ever SAH (62% women; mean±SD age, 56.5±17 years) registered in ACROSS. SAH was verified by CT in 390 (90%) or necropsy alone in 39 (9%), and 330 patients (76%) had an aneurysmal origin of the SAH diagnosed by angiography, surgery, or autopsy. Interviews were completed for patients or proxies (in 63% of cases) at a median of 9 (interquartile range, 4 to 34) days after onset of SAH. Proxies were interviewed for patients who experience sudden or early death or disability; these patients had a significantly worse prognosis, were older, and were more likely to be married, to be more physically active, and less likely to drink alcohol than patients who survived and were able to be interviewed. However, there was no difference between the groups (direct- versus proxy-interviewed patients) in regard to the type of SAH (aneurysmal versus nonaneurysmal) or smoking status. Proxies were a spouse, partner, or other close relative such as a parent, child, or sibling in 79% of cases. Although the information obtained was cross-checked for accuracy against medical records, the research nurses rated the reliability of interview as good or high quality in 98% and judged the person to have no difficulty responding to the questions in 73% of cases. Use of medical records alone for exposure data occurred in only 12% of cases.

Physical Exertion

A definite time of onset of SAH was established for 393 patients (91%), and information on the levels of physical activity around this time was obtained for 338 (78%). Table 2 shows the level of activity reported by these 338 patients, categorized as sleeping (13%), sitting (36%), light activity (31%), and moderate to extreme (strenuous) exertion (19%). In the primary pair-matched interval analysis (model 1), there were 304 patients (70%) for whom information was available on activities in the 2 relevant time periods. Of these patients, 58 (19%) reported moderate to extreme exertion only during the 2-hour period before onset compared with 27 (9%) who reported such exertion only during the control period (the same 2-hour period on the previous day) and 10 (3%) who reported moderate to extreme exertion at both times. This analysis yielded almost a tripling in risk of SAH for those who engaged in moderate to extreme exertion during the hazard period (OR, 2.7; 95% CI, 1.6 to 4.6).

The alternative nonparametric multiple intervals method of analysis (model 2) was restricted to the 158 patients for whom we had information on activities over the full 26 hours before the onset of SAH. In this analysis, adjusted for time of day, moderate to extreme exertion within the 2-hour hazard period was associated with a similar risk of SAH (OR, 2.7; 95% CI, 1.2, 4.6).

Table 3 shows that according to model 1, the risks of SAH associated with moderate to extreme exertion were consistent among subgroups of patients with different characteristics. Older patients and those with a history of hypertension were at a somewhat higher risk when engaged in moderate to extreme exertion, although the associations were not significant. Usual level of activity did not significantly alter the risk associated with acute moderate to extreme exertion. Similarly consistent results were derived with the second approach. However, the number of cases who reported engaging in moderate to extreme activity on a regular basis (activities involving puffing undertaken >3 times per week in the previous 2-week period) was only 17%. Thus, the point prevalence of involvement in such activity in any single 2-hour period is only 0.1%, and the corresponding population-attributable risk is only 2% (95% CI, −4.9 to 8.4).

Information on cigarette smoking was available for 348 patients in the 2 relevant time periods (model 1). Overall, 15 patients (4%) had smoked ≥4 cigarettes over the 2-hour hazard period, and 14 (4%) had smoked a similar amount during the same period on the day before onset. This analysis yielded an OR of 1.1 (95% CI, 0.4 to 3.7). There were 340 patients for whom information was available on alcohol intake in the 2 relevant time periods (model 1). Three patients (1%) had drunk ≥4 standard drinks of alcohol during the hazard period, and 5 (1%) had drunk a similar amount during the same period on the day before onset, yielding an OR of 0.4 (95% CI, −∞ to 5.3).

Discussion

We found that moderate to extreme physical exertion was associated with a tripling of the risk of SAH in the subsequent 2 hours compared with periods of light or no activity and that usual levels of activity did not alter this transient risk association. Heavy cigarette smoking and binge drinking were not associated with an increased risk of SAH, but these analyses suffered from the problem of sparse data in a subset of the population.

The strengths this study include the prospective ascertainment of cases from a large population base (limiting referral bias), uniform and strict diagnostic criteria (limiting misclassification of SAH), and use of the case-crossover design. However, the study did have several limitations, including the use of proxy respondents to obtain information on exposures in a high proportion of the cases who were unable to be interviewed because of early death or disability, but this situation is not uncommon in epidemiological studies of SAH.^11,27 Even though studies have reported a high level of agreement between index and proxy respondents in regard to personal habits and medical conditions^28,29 such as those considered in these analyses, it is possible that some of our cases may have underreported exposures such as recent alcohol intake, whereas proxies might have overestimated them, leading to a bias. To some extent, though, any such error is likely to be nondifferential because the case-crossover design permits a self-matched analysis in a subject who serves as his or her own control. On the other hand, differential error resulting from missing information is hard to avoid in studies of SAH because of its high mortality and morbidity. It is possible, therefore, that the lifestyle patterns of survivors were different from those of patients who died early, leading to bias. Another problem is that the measure of physical exertion was limited compared with other more objective measures of physical activity. Moreover, the questionnaire was updated during the course of the study, and the prespecified criteria for heavy exertion, often used in studies of coronary artery disease, were not appropriate for this analysis because of sparse data in this category. For all these reasons, the magnitude of the relative risks associated with the transient exposures investigated should be viewed as estimates.

The role of physical activity in the etiology of SAH has not been clearly defined,² although surveys of activities preceding aneurysmal SAH have shown that it is common for patients to be engaged in at least moderate activity before the event.^8,9 The 1 previous case-crossover¹⁰ investigation suggested that vigorous physical activity increased the risk of SAH 15-fold. However, this study was based on a small series of patients, and the methods used for the case-crossover analysis were not standard because assumptions were made about patterns of exertion in the reference period. Our results suggest that moderate to extreme physical exertion (≥5 MET) can trigger the onset of SAH, which mirrors the results of case-crossover analyses of triggers of myocardial infarction^6,7,30 and sudden death.³¹ It is important to note, however, that our study differs from other case-crossover investigations of physical activity because, for reasons given earlier, the cutoff point used for exertion here was 5 MET, whereas most of the other studies in cardiac patients have used a cutoff of 6 MET.

Transient elevations in blood pressure would seem to be an obvious explanatory mechanism linking physical activity to SAH. This hypothesis is supported by the circadian pattern in the incidence of SAH,³² intracerebral hemorrhage,³³ and stroke in general,³⁴ being highest in the morning when elevations in blood pressure and heart rate occur in parallel with the commencement of activity.

Case-crossover studies have identified strenuous physical exertion as a trigger for myocardial infarction, particularly among sedentary people,^6,7 whereas we found no association between habitual exercise habits and risk of SAH. There are several potential explanations for this difference. First, chronic exposures, including physical activity, may be less relevant to SAH and/or the formation and rupture of cerebral aneurysms. Second, regular dynamic recreational exercise may produce only modest reductions in 24-hour levels of blood pressure,³⁵ so levels achieved during moderate to extreme exertion may remain independently high enough to precipitate SAH. Third, usual activity levels may have been incorrectly estimated or misclassified, producing a null effect.

In conclusion, if the relationship between moderate to extreme physical exertion and SAH were determined to be causal, this finding should not deter people from being physically active because the absolute number of cases of SAH caused by this activity is extremely low.

Footnotes