One of the key milestone of a successful cell therapy development (e.g. CAR-T) is the determination of optimal dose from the first-in-human (FIH) study. Commonly used dose finding designs, e.g., 3+3 and BLRM, utilize safety data only assuming toxicity and clinical benefit increase with dose so the maximum tolerated dose (MTD) has the most promising efficacy. However, the sensitivity of infused modified T cells and variations of final T-cell product among individuals make such assumption in doubt. New paradigm of designs to identify a tolerable dose with acceptable efficacy is imperative in cell therapy. Clinical meaningful biomarkers can be used as surrogate efficacy outcomes. We reviewed and evaluated the current developed escalation designs and presented the limitations. To improve the performance and benefit patients, we proposed novel model-assisted designs TEPI-2 and U-BOIN, flexible of dose-safety/efficacy relationship. Both designs generate transparent dose escalation decision tables before the trial and determine the optimal dose retrospectively, and demonstrate more accuracy in determining optimal dose and are more safe to patients.