Reply to Mendy.

TO THE EDITOR—We thank Mendy for highlighting the importance of appropriate control of covariates including visual impairment, child’s education level and stroke, when examining the association between herpesviruses and cognitive impairment [1]. As noted, herpes simplex virus type 1 (HSV-1) enters latency in the trigeminal ganglion but can travel along facial nerves to the eye and cause ocular diseases such as conjunctivitis and herpes keratitis, severe cases of which may lead to scarring of the cornea and subsequent vision impairment, including blindness [2]. Similarly, cytomegalovirus (CMV) can infect the eye, causing CMV retinitis, which, although primarily observed in immunocompromised populations, can also lead to blindness [3]. Mendy, therefore, draws attention to one biological mechanism by which herpesviruses may lead to poorer performance on visually based cognitive tasks [1]. The National Health and Nutrition Examination Survey (NHANES) III asked all participants whether they had “trouble seeing with one or both eyes, even when wearing glasses or contact lenses,” and asked adults whether they had “total blindness in one or both eyes.” Among those tested for HSV-1 and cognition, the weighted proportion reporting visual impairment among children and middle-aged adults was 13.3% and 10.8%, respectively, and the weighted proportion reporting blindness among middle-aged adults was <1.0%. HSV-1 but not CMV seropositivity was significantly associated with visual impairment among children in our sample (odds ratio, 1.69 [95% confidence interval {CI}, 1.03–2.76]), after controlling for age, sex, race/ethnicity, poverty income ratio (PIR), family education level, country of origin, and smoking exposure. While there was no association between visual impairment and reading score, visual impairment remained significantly associated with block design score (β, −0.62 [95% CI, −1.22 to −.02]), after adjusting for age, sex, race/ethnicity, PIR, family education level, country of origin, smoking exposure, and HSV-1 seropositivity. After additional adjustment for visual impairment, the associations between HSV-1 seropositivity and reading and block design scores were only slightly attenuated (ie, β, −0.67 [95% CI, −1.17 to −.16] and −0.78 [95% CI, −1.25 to −.31], respectively) and remained statistically significant. Among middle-aged adults, CMV and HSV-1 seropositivity were not significantly associated with visual impairment, nor was visual impairment significantly associated with serial digit substitution test (SDST) impairment in fully adjusted models. In addition, there was no association between HSV-1 or CMV seropositivity and blindness, nor was blindness associated with SDST impairment among middle-aged adults. Moreover, there was no substantial attenuation of the association between HSV-1 or