A 2 layer
microstrip shares air on one side (Dk=1.01 to 1.04 (?) depending on humidity) and a dielectric substrate Dk= X in between the bottom side ground plane. At low f < 1MHz the effective Dk (or er eff.) might be 60% to 70% typ. of the substrate depending which. Meanwhile, when buried between conductors layers or
stripline it is 100% of Dk. Microstrip will also radiate more like an antenna patch but not as effective.
Since the microwave skin effect is very sensitive, at 20 GHz the
conductor thickness, roughness and material are just as critical as the
low-loss tangent dielectric e.g. PTFE, Rogers (e.g. Duroid 5880) or a custom ceramic substrate. This will affect your stripline forward and return loss and vary significantly up to 20 GHz as well as the length if mismatched.
As
@barry mentioned using 3 (or 4 layer) allows thinner traces because Dk is now 100%,
Essentially is the ratio of the square root of L/C where L depends on a log of the length/width ratio and C depends on Dk times a ratio of the area per unit length over the gap (Dk*a/g) times some constant for units in metric or not. Thus for a given length, Zo = 50 Ohms can be a thin trace and even thinner with a thinner substrate.
Dielectric tolerances are often 10% so electrical testing is recommended where the Mfg uses a test coupon outside your layout and verifies by TDR or a time-domain reflectometer to your tolerance specs. Copper thickness can also vary the impedance another 10% but this is predictable until you get to 20 GHz then the surface roughness increases the length and inductance per unit length more than the capacitance. Conductor losses are also a factor and some go to the trouble to gold electroplate the traces.
Find out what your tolerance expectations are and how you will verify it.