#conditional-probability-table

In a Bayesian network, $x_4$ has parents $x_1$ and $x_2$ (all binary). Its table so far reads $P(x_4 = 1 \mid x_1 = 0, x_2 = 0) = 0.8$, $P(x_4 = 1 \mid x_1 = 0, x_2 = 1) = 0.4$ and $P(x_4 = 1 \mid x_1 = 1, x_2 = 0) = 0.8$. What must $P(x_4 = 1 \mid x_1 = 1, x_2 = 1)$ be for the edge $x_1 \to x_4$ to be removable without misrepresenting the distribution?