Our data demonstrates that young AAs are more likely to experience a lacunar stroke than young EAs, as explained by an increased incidence of HTN. Younger-young individuals, those less than 40 years old, are more likely to experience a cardioembolic stroke, while individuals in the 40–50 year old age group are more likely to experience a lacunar or large vessel stroke. In general, current smokers are more likely to experience a large artery stroke, while individuals with HTN are more likely to experience a lacunar or large artery stroke. Of note, our two analyses which utilize differing reference groups produced similar results less that Analysis 2 (cryptogenic stroke as reference) did not demonstrate that patients below age 40 were more at risk of cardioembolic stroke; this may relate to a decreased reference sample size.
Our results add to the growing literature demonstrating ethnic differences in stroke subtype proportions [1, 12, 13], further inferring on these relationships in a younger-onset population. In 2012 Song et al. retrospectively evaluated 350 acute ischemic stroke cases (mean age of 63) on the basis of TOAST classification. In contrast to our findings, their older population demonstrated similar proportions of lacunar strokes in the AA and EA cohorts. As consistent with our findings, similar proportions of cardioembolic stroke were reported in their AA and EA cohorts. In another study , a cohort of 511 patients between 18 and 49 years of age (mean age of 39.8) demonstrated no significant sex-based differences in the proportion of small- and large-vessel disease, and stroke of undetermined etiology, although cardioembolism (and substance abuse) predominated in men as compared with women. In contrast, and limiting potential comparisons to our study, 44 % of the young stroke patients (and almost 60 % of the women) had nontraditional etiologies for stroke (i.e. prothrombotic states, migraine-related conditions, substance abuse, cervical artery dissection, cerebral venous thrombosis, inflammatory and miscellaneous vasculopathies, and pathological conditions related to pregnancy, postpartum, fibromuscular dysplasia or Moyamoya syndrome) . Although sex differences have also been demonstrated in other studies with men experiencing more strokes than women [15, 16], the precise mechanisms for these differences remains uncertain. In our study, among the lacunar, large-artery and cardioembolic subtypes we did not find any significant sex differences in our young-onset population, although by percentage, females were more likely to have a stroke of other determined etiology.
Other prior studies have demonstrated differences in stroke incidence between ethnic groups at young ages. In the Northern Manhattan Stroke Study young AA aged 20–44 were found to be 2.4 times more likely to have a stroke than similarly aged EA . Other more recent studies have demonstrated that the incidence of stroke in the young is on the rise. For example, an analysis of temporal trends of stroke in the Greater Cincinnati/Northern Kentucky region demonstrated that the incidence of ischemic stroke in adults below age 55 has risen from 12.9 % in 1993/1994 to 18.6 % in 2005 . In our data, when controlling for sex and ethnicity, we found that younger adults below age 40 were 1.62 times more likely to experience cardioembolic stroke than adults between ages 40–50, whereas adults in the 40–50 age group were 2.97 and 2.77 times more likely to experience lacunar and large artery strokes, respectively.
While ethnicity, age, and sex are major non-modifiable risk factors for stroke, the modifiable risk factors remain a critical intervention point for physicians working to reduce future stroke risk. Smoking and hypertension have long been associated with a higher risk of stroke. Smoking increases stroke risk through multiple mechanisms  and does so in a dose-dependent type fashion . Ethnic differences in stroke risk as associated with smoking remain uncertain. In contrast, evidence exists demonstrating a differing ethnic response to hypertension as associated with stroke risk, with a greater risk among AA as compared to EA . Many of these traditional risk factors develop during young-adulthood. One study examined the risk factor profiles of ischemic stroke under the ASCO classification  in patients age 16–54 and found that smoking, diabetes, hypertension and low HDL-cholesterol are significant risk factors for atherothrombosis, and that hypertension is a significant risk factor for small vessel disease in young adults . While few studies exist that comprehensively evaluated ethnic and risk factors differences in young-onset stroke, some studies have evaluated such strata individually. For example, a Greek study evaluating the incidence of ischemic stroke in young adults aged 15–45 demonstrated that smoking was the most frequent risk factor, with 59.3 % of the young strokes actively smoking . In our young-onset stroke study, we found that HTN increases the risk of lacunar stroke by 2.03 times and large artery stroke by 1.71 times, when controlling for sex, ethnicity, and age. Furthermore, our data show that the increased risk for lacunar stroke in AA compared to EA is mediated by HTN. Despite the fact that we did not find smoking to mediate the ethnic differences in stroke occurrence, we did find that current smokers were 1.81 times more likely to have a large artery stroke than non-smokers.
Another important point to consider is that a significant proportion of the patients in our study had cryptogenic stroke (51.3 %). While some prior studies have shown similar percentages of cryptogenic stroke [11, 12], this is somewhat higher than other studies evaluating young-onset ischemic stroke [24, 25]. This distribution not only reduced the sample sizes of the other subtypes, the heterogeneity of this group limited our ability to determine which risk factors were primarily responsible for the stroke.
Our study has several limitations. One primary limitation was the small sample sizes in several of the ethnically-stratified TOAST-subtype cells, thereby precluding investigations of several common traditional risk factors (i.e. DM, prior MI). Furthermore, the effects of dyslipidemia could not be evaluated secondary to inconsistent lipid data collection. Additionally, our study only includes those of EA and AA descent; other ethnicities present in the US population (e.g. Hispanics, Asians) could not be explored. Again, as based on limitations of ethnic- and subtype-specific sample sizes we were unable to analyze other stroke etiologies (dissection, hematological, etc.) and other risk factors, including drug abuse, which are commonly seen in younger stroke populations. However, we did not exclude such cases, which may also influence our findings. For example, 8 cases with acute cocaine use (within 24 h) were included in our analyses; their TOAST subtypes were: Cardioembolic n = 1, Lacunar n = 1, Other Determined n = 1 and Cryptogenic n = 5. Our study also only evaluated a single US geographic region, specifically the areas including and surrounding the Baltimore/Washington-DC metro region limiting the application of our results to other populations; within the US the distribution of traditional and non-traditional risk factors can differ vastly by geographical region. As such, similar studies conducted across the US and world-wide are needed to verify our findings. Also as mentioned previously, TOAST classifies strokes to one of five categories with either probable or possible certainty, with patients whose strokes that may have resulted via several possible etiologies being included in the category of undetermined etiology. Perhaps in the future improved subtyping methodologies such as the Causative Classification Algorithm (https://ccs.mgh.harvard.edu/ccs_title.php) [26–28] will allow patients to be better categorized into distinct stroke subtypes, thereby allowing for more refined classifications and analyses.
While our study was able to identify several groups predisposed to specific stroke subtypes and the risk factors driving these relationships, additional studies are needed to further identify other high risk ethnic-specific groups and refine risk estimates in other non-traditional risk factors, such as migraine headache and evolving risk factors such as e-cigarettes.