This is rather a comment to Steven Landsburg/ Vaserstein answer, but too long for a comment.

Finding a quotient of $\mathbf Z[X]$ with $E_2 \neq \mathrm{SL}_2$ is easy. Actually, $\mathbf Z[X]$ will do. For example, Cohn proved that the matrix :
$$\begin{bmatrix}
1+2x & 4 \cr
 -x^2 & 1-2x
\end{bmatrix} $$
is in $\mathrm{SL}_2$ but not in $E_2$.

If one prefers an example in the spirit of the one suggested by Vaserstein, then Cohn (again!) shjowed that in the ring of integers of $\mathbf Q[\sqrt{-19}]$, with $\theta= \frac{1+\sqrt{-19}}{2}$, the matrix 
$$ 
\begin{bmatrix}
3- \theta & 2+ \theta \cr
-3-2\theta & 5-2\theta
\end{bmatrix}
$$
is in $\mathrm{SL}_2$ but not in $E_2$.

(I took these two examples of matrices from the book of T.Y. Lam on "Serre's problem on projective module, Chapter I.9.)

With these examples, it should be straightforward to make explicit a unimodular row, but I have to admit that I don't understand Vaserstein's argument. (Is it well known that $\mathrm{sr}(A)= 2 $ implies $\mathrm{SL}_2(A)= E_2(A)$?)