Background: Central aortic systolic blood pressure (cSBP) has been shown to be a better marker of physiological and pathological changes associated with central hemodynamics as compared to brachial systolic pressure (bSBP). However, the noninvasive estimation of cSBP requires the registration of the radial or brachial blood pressure waveform. This study assessed methods of estimating cSBP without the peripheral pulse waveform. Method: Three statistical learning methods: (i) multiple linear regression (LR), (ii) artificial neural network (ANN) and (iii) support vector regression (SVR) were investigated for the estimation of cSBP directly from cuff-based oscillometric brachial systolic, diastolic and mean pressure without the peripheral pressure waveform. These models were established using a development group of normal subjects (n = 77) and validated by a noninvasive validation group (n = 312) and by an invasive validation group (n = 62, subjects undergoing catheterization and nitroglycerin infusion). Results: Noninvasive validation showed that the cSBP estimated by these three models agreed well with the cSBP measured by the SphygmoCor (AtCor, Australia) using a peripheral waveform (LR: r2=0.95, ANN: 0.95, SVR: 0.95, all p < 0.001; mean difference = 0.4 ± 3.7, 0.5 ± 3.6 and 0.6 ± 3.7 mm Hg respectively; all p < 0.001). Invasive validation showed the cSBP estimated by the SVR model achieved the smallest mean difference (1.3 ± 5.5 mm Hg, p < 0.05) relative to invasive cSBP as compared with the LR model (3.3 ± 4.7 mm Hg, p < 0.001) and ANN model (6.8 ± 5.3 mm Hg, p < 0.001), respectively. Conclusion: The study demonstrated that statistical learning methods, especially the SVR, have a potential application for the simplification of noninvasive measurement of central blood pressure measurement in clinical practice when the peripheral pulse waveform is not available.