Prevention Opportunities for Oral Contraceptive–Associated Ischemic Stroke
Background and Purpose—Literature suggests a small increased risk of ischemic stroke with oral contraception (OC) use. We evaluated the association of stroke and OC use in women on the basis of whether they recalled being advised by a physician not to use OC or to discontinue OC use because of the presence of stroke risk modifiers, and whether such advice resulted in behavioral change.
Methods—A total of 572 women (224 strokes and 348 controls) aged 15 to 49 years were interviewed about OC use and risk modifiers, including cigarette smoking and headaches, among others.
Results—The adjusted odds ratio for OC use and stroke was 2.00 (95% confidence interval, 1.29–3.09). The association of OC use with stroke was stronger in women that reported receiving doctor’s advice against OC use because of the presence of other stroke risk modifiers (odds ratio, 3.12; 95% confidence interval, 1.62–6.00) than in women who did not recall receiving such advice (odds ratio, 1.31; 95%confidence interval, 0.71–2.43). Of 256 women who recalled being advised by their doctor not to start OC or to discontinue OC use because of the presence of other stroke risk modifiers, 24% were still on OC at the time of stroke or interview.
Conclusions—We confirm that certain medical conditions increase the risk of stroke during OC use and demonstrate the importance of physician counseling in those using OC in the setting of concurrent high-risk conditions and the need for improved patient compliance with such counseling.
Oral contraception (OC) has been associated with an increased risk of ischemic stroke.1 We sought to determine whether the elevated stroke risk in women using OC was restricted to women with other concurrent conditions (risk modifiers); whether women with these conditions recalled being advised by their physician not to begin using OC or to stop using OC; and whether this advice resulted in behavioral change.
Material and Methods
The Stroke Prevention in Young Women Study is a population-based case–control study consisting of 514 cases aged 15 to 49 years at the time of stroke and 617 controls matched by age, region of residence, and ethnicity; the details of study design have been published previously.2 Exclusions included missing data on OC use (n=17); unknown menstrual status (n=3); postmenopausal or currently pregnant, pregnant within past 2 months, or currently nursing (n=283); prior surgery to remove the uterus, ovaries, both, or a tubal ligation (n=256), resulting in a final study cohort of 224 cases and 348 controls.
Demographic variables, OC use, and the presence of stroke risk modifiers were determined through a standardized face-to-face interview. Current OC use was defined as use within 1 month before the stroke with a comparable reference time for matched controls. Participants were asked whether they had been told by a physician not to start OC or to discontinue OC use because of any of the following conditions: cigarette smoking, headaches, high blood pressure, diabetes mellitus, sickle cell disease or sickle cell trait, prior transient ischemic attack, chest pain or myocardial infarction, blood clots in legs or lungs, or a family history of heart problems.
Cases and controls were compared using t tests for continuous variables and Mantel–Haenszel χ2 tests for categorical variables (SAS Institute, Inc, version 9.2; Cary, NC). Unadjusted associations between ischemic stroke and risk modifiers were then examined within predefined subgroups using χ2 or Fisher exact tests. We additionally categorized subjects according to whether they had received physician advice not to start using OC or to discontinue OC use because of the presence of ≥1 risk modifiers versus those who had not received such advice. We estimated the association between stroke and OC use in those receiving physician advice against OC use versus those who did not receive such advice using logistic regression adjusting for age and ethnicity (self-report). Two-tailed P values <0.05 were considered statistically significant.
Table 1 demonstrates participant characteristics. There was no statistically significant difference between the percentages of cases and controls that were told not to start or to stop OCs because of the following risk conditions: current smoking, headache, hypertension, chest pain or myocardial infarction, or history of blood clots in the legs or lungs. Because of the paucity of participants with diabetes mellitus, sickle cell disease or sickle cell trait, transient ischemic attack, and family history of heart problems, these variables were excluded from further analyses.
Evaluating the entire population, OC use was significantly associated with stroke (adjusted odds ratio [OR], 2.00; 95% confidence interval [CI], 1.29–3.09) as indicated in Table 2. Table 2 further demonstrates the association of OC use with stroke stratified by the absence or presence of doctor’s advice against OC use because of the presence of ≥1 risk modifiers. OC use was highly associated with stroke in women receiving physician advice not to use OC because of the presence of risk modifiers (OR, 3.12; 95% CI, 1.62–6.00), although the association was substantially attenuated and no longer statistically significant in women who did not receive such advice (OR, 1.31; 95% CI, 0.71–2.43). When we considered individual conditions that resulted in physician advice against OC use, the 2 most common reasons for receiving such advice were current smoking (106 of 256; 42%) and headaches (84 of 256; 33%). When stratified by these individual conditions, OC use was strongly associated with stroke among current smokers (OR, 4.29; 95% CI, 1.51–12.16), but much less so among women who did not currently smoke (OR, 1.70; 95% CI, 1.04–2.79). Similarly, OC use was strongly associated with stroke in women reporting headaches (OR, 3.82; 95% CI, 1.27–11.56), but only borderline significant in women reporting headache free (OR, 1.63; 95% CI, 1.01–2.66). Stratified analyses evaluating OC use–associated stroke risk in the presence or absence of hypertension, chest pain or myocardial infarction, or a history of blood clots in the legs or lungs demonstrated no significant associations (results not shown). Among women who received physician advice against OC use (either not to start or to discontinue), 34% of cases and 18% of controls were still on OC at the time of stroke or interview, respectively (P=0.03).
Physician Advice (Not to Start or Discontinue OC) and Patient Compliance
There were 256 women with ≥1 risk modifiers (99 cases; 157 controls). Among these women, only 38 (15%) recalled being advised not to start OC on the basis of their pre-existing risk-modifier profile; a similar percentage of cases (13%) and controls (16%) recalled such physician counseling. Of note, 9 (24%) of these 38 women were taking OC at the time of stroke or interview, despite being advised not to start OC.
Of the 256 women with ≥1 risk modifiers, 93 (36%) recalled being told to discontinue OC (40 cases; 53 controls) on the basis of their risk-modifier profile. A similar percentage of cases (40%) and controls (34%) recalled such physician counseling. Of note, 14 (15%) of the 93 were taking OC at the time of stroke or interview, despite being told to discontinue OC use.
Several conditions, termed risk modifiers, have been shown to increase stroke risk in the setting of OC use, with smoking3 and headache4 being those most cited. We confirm these findings demonstrating a significantly increased risk of stroke in the setting of OC use in those with ≥1 concomitant risk modifiers (OR, 3.12; 95% CI, 1.62–6.00), but no significantly increased risk in those free of such conditions (OR, 1.31; 95% CI, 0.71–2.43). In further agreement with the existing literature, those at highest risk were current smokers (OR, 4.29; 95% CI, 1.51–12.16) and those with headaches (OR, 3.82; 95% CI, 1.27–11.56). Combinations of these risk-modifying conditions in the setting of OC use have also been shown to act synergistically, thereby elevating risk in a greater than additive fashion. Of these, the combination of smoking and migraine headache in the setting of concurrent OC use has been demonstrated to be particularly deleterious.5 Although our sample size precluded a detailed evaluation of this combination, interestingly, of the 22 study participants who were smokers with headaches, 8 of the 13 cases were on OC, but none of the 9 control subjects were so. As such, our data seem to agree with a markedly elevated risk of stroke in this setting.5
Our study not only acts to confirm previously recognized high-risk subgroups, but also evaluates whether patients with such conditions recalled being told by a physician not start or to stop OC as based on the existence of these conditions and whether they did so or not. Currently, there exists scant literature on the physician–patient encounter in this topic area, thereby making our study unique. As described in the Results, evaluation of our prestroke physician counseling data demonstrated that only ≈15% (38/256) of patients recalled being told not to start OC on the basis of their risk-modifier profile. However, once patients were taking OC physician counseling improved, with ≈36% (93/256) of the participants with ≥1 high-risk conditions being told to discontinue OC use. Furthermore, we also found that patients were not optimally compliant with physician instructions when they were provided, with 24% (9/38) of the women with ≥1 high-risk conditions who were advised not to start OC remaining on OC at the time of their stroke or interview. These results indicate that in the setting of OC use and concomitant risk modifiers, improved physician counseling and improved patient compliance could potentially reduce ischemic stroke rates.
Our study benefited from the rigorous exclusion criteria designed to limit the confounding effects of hormonal fluctuations associated with pregnancy, the postpartum period, and lactation. Unfortunately, these same criteria also acted to reduce our sample size, thereby limiting our ability to evaluate the risk modifiers by stroke subtype, headache type (nonmigraine versus migraine±aura), OC by formulation (estrogen versus progestin versus combination), OC dose, and OC delivery method. Our sample size also limited our ability to control for other vascular risk factors (eg, diabetes mellitus and thrombophilia), which potentially may have produced unmeasured confounding. Furthermore, our data were collected retrospectively; as such, our results could be influenced by recall bias, particularly in the sense that those who had a stroke might be more likely to remember whether they did or did not receive specific physician instructions relating to OC use. Finally, there is a chance that subjects may have misunderstood or incorrectly answered our questions.
We confirm that certain medical conditions increase the risk of stroke during OC use and demonstrate the importance of physician counseling in those using OC in the setting of concurrent high-risk conditions and the need for improved patient compliance with such counseling.
Sources of Funding
This work was supported by the Department of Veterans Affairs and the National Institutes of Health.
- Received October 31, 2013.
- Revision received December 10, 2013.
- Accepted December 13, 2013.
- © 2014 American Heart Association, Inc.
- MacClellan LR,
- Howard TD,
- Cole JW,
- Stine OC,
- Giles WH,
- O’Connell JR,
- et al
- Bousser MG