This page has the following table:
So, if the 22uF is a 20% tolerance part such that is may actually be 17.6uF (-20%), and it is really 15uF (-15%) due to heating, it may be reduced to ~10.5uF (-30%) near its tolerance or ~14uF (-5%) near its operating point. To reduce the effect, use a higher breakdown value capacitor.
Although this webpage tests much higher breakdown voltage ceramics, the results are illustrative:
This part, the 08056D226MAT2A by AVX, states that it meets its tolerance tested at 0.5 VRMS, not close to its breakdown:
For Cap > 10 μF, 0.5Vrms @ 120Hz
It also mentions that capacitance may change up to 12.5% over its load life, 7.5% due to soldering thermal shock, and 12% due to flexure fissures (cracks).
Info tidbit: The term X5R means it is composed of a class 2 dielectric (ie: ceramic) that will maintain its capacitance to within 15% (18.7uF - 25.3uF) over a temperature range -55oC to 85oC.
Something that may interest you, though it has more to do with the final application than the question -- the great 'pedia also mentions:
Due to its piezoelectric properties, they are subject to microphonics.
... And the previously linked page:
High-K ceramic capacitors can show significant piezoelectric effects; if you tap them they will produce a voltage spike. This is caused by the barium titanate, the main material in high K ceramics. The higher the K, the stronger this affect.
Assume 10nH total inductance. Given V = L * dI/dT, or conversely,
I = 1/L * integral (V * dT),
the peak I = 1/10nH * integral (100V * 100nS) = 10^+8 * 100 * 1e-7 = 1,000 amps.
If your total inductance (caps plus solder pads plus GND inductance of vias + vias to a shared high current bus) is 100nH, the peak current is only 100 amps.
Best Answer
You must pick your capacitor technology bearing in mind your intended use.
X5R (Or worse Y5V) is not really the smart choice for filters (Which is what you have if you care about phase shift thru the cap), apart from anything else you will have built a voltage controlled phase shift network as applied DC across that cap will cause SIGNIFICANT changes in value (This gets much worse with smaller case sizes).
C0G is generally ok for this sort of thing, but you will struggle with finding more then about 100nF or so. The other option is film or maybe a bipolar electrolytic (If going here make the value LARGE, so the phase shift is negligible over any reasonable amount of drift).