change of variable in a 2-loop integral given the 2 loop integral
$$ \int dq_{1} \int dq_{2}F(q1,q2) $$ (1)
then in dimension D=4 our integral will be a 8-dimensional integral
so why can not make a change of variable to 8-dimensional spherical coordinates? can we , the idea is to simplify the integral so we have
$$ \iint d\Omega\int_{0}^{\infty}r^{7}F(r,\Omega) $$
integration over angular variables $\omega $ can be obtained by numerical methods
 A: The question is not what is possible, but what is useful to obtain analytic results.
In multi-loop integrals one is often interested in analytic results, because it can be very hard to confirm numerical results to a level where you trust them completely.
Now the two-loop integrals usually involve propagators of the type
$$ F(q_1, q_2) = \frac{i}{(p - q_1 - q_2)^2 - m^2} = \frac{i}{p^2 + q_1^2 + q_2^2 - p\cdot q_1 - p \cdot q_2 - q_1 \cdot q_2 - m^2} $$
If the scalar products with the external momentum were not present, one could indeed go to 8D spherical coordinates (after an appropriate Wick rotation to Euclidean space). But as soon as you care about external momenta, things get more complicated.
There actually are books on techniques of dealing with two-loop integrals.
Edit to expand on why we don't want to do numerics:
First of all, numerics in multiloop calculations can be done and there are a great deal of people doing this. This is a very complicated research subject of its own. The main issue is not solving the integral numerically, but making sure you trust the answer.
In particle physics calculations we often have cancellations between different contributions. The terms that cancel can be much larger than the remaining contributions, see e.g. the GIM mechanism. Similar things can happen on a more basic level, just having two different bits of the integral canceling and leaving a smaller third bit over.
Making sure that everything that is supposed to cancel indeed does so to the required precision is hard. You have to apply your method tree times, to calculate both contributions and the final result. If the two contributions indeed cancel you can trust your result a bit. But then you have to perform a bajillion other tests on your numerics before you can be sure that 18.3922787 is really the number you were looking for.
Therefore, a lot of effort is spent on solving these integrals analytically, where cancellations are exact and you can understand in detail, where the result comes from. For these analytic calculations, 8D spherical coordinates would not do you any good, though.
