Recall $\rm\ (p,lcm(q_1,q_2)) = lcm((p,q_1),(p,q_2)),\, $ i.e. gcd distributes over lcm. Hence
$\rm\ \ ((p_1,p_2),lcm(q_1,q_2)) = lcm((\color{#c00}{p_1},p_2,\color{#c00}{q_1}),(p_1,\color{#0a0}{p_2,q_2})) = lcm(\color{#C00}1,\color{#0a0}1) = 1\ $
since we know, by hypothesis $\rm\ \color{#c00}{(p_1,q_1)=1},\ \color{#0a0}{(p_2,q_2)=1}.\ \ $ QED
Remark $\ $ Using only gcds and lcms (vs. full-blown prime factorizations), this method generalizes to any domain where gcds exist, i.e. any gcd domain (which need not have any primes, e.g. the ring of all algebraic integers). Moroever, if one desires to compute the value of an expression of this form, it is much more efficient to compute gcds (Euclidean algorithm) than prime factorizations. Generally, one can use gcds to compute factorizations into coprimes (vs. primes), which often suffices to solve many problems.