(A) Suggested exposure margin for marketed VEGF-R inhibitors based on post-marketing incidence of hypertension in correlation with plasma exposure (VEGF-R IC50/free Cmax). The proposed 10 times margin represents clear separation of VEGF-R inhibitors with and without significant increase in hypertension with the only exception of nintedanib. (B) FAERS reports of small molecule kinase inhibitors with VEGF-R2 inhibition show an increased incidence of hypertension reports only in case their exposure margin is less than 13. The label of drugs with high incidence of hypertension in FAERS lists this side effect, while none of those drugs that have low incidence carry the label. *p-value of association between drug and hypertension <0.001. Counts (N), expected counts (E), and an often-used disproportionality measure (EB05) based on the FDA’s FAERS database of spontaneous reports of suspected drug adverse drug reactions are provided. The values of E are the expected number of patients reporting vascular hypertensive disorder after taking each drug if the drug reports and the reports of the event were independent within the database, conditional on the patients age and gender. The ratio N/E is a measure of disproportionality of report counts of each particular drug-event combination. The value EB05 (empirical Bayes 5% lower bound of a 90% credible interval) is a conservative estimate of the true reporting disproportionality that uses estimated overall prevalence of drug-ADR associations throughout the database. The value of EB05 is less than N/E and has the effect of correcting the simple ratio for sampling variance and multiple comparisons bias. See literature (DuMouchel, 1999; DuMouchel and Pregibon, 2001; Szarfman et al., 2002; Almenoff et al., 2007) for details and discussion of the FAERS database and the use of disproportionality analyses within spontaneous report databases. The values of EB05 for the first three drugs indicate 95% confidence that reports of those three drug-event combinations are reported about three or four times as often as would be expected if they were independent, while the values of EB05 <1 in the final three drugs in the table indicate no evidence for higher than expected reporting rates. More detailed results from Bayesian analysis are available in Supplementary file 3. †Significant increase.