This is more of a opinion question than a information question. I am designing some boards that will be working near decent sized actuators, so I ask you, what is more resistant a 0805 SMD or a 7mm thru-hole resistor?
The only vibration (SAE J1455) SMT problems I've ever seen for common components are failures for large aluminum-electrolytics. The solution there is just to anchor them down with a gob of silicone.
An 0805 resistor will not fall off from pure vibration unless there is a tremendous amount of board warping going on (then it may fracture), or unless you are going to expose it to several thousand g routinely (in which case you have bigger things to worry about).
An 0805 resistor weighs about 4 milligrams, and the pound of force or so I just put on one (on a PCB on my desk) with my fingernail did nothing, so that's equivalent to about 113,000 times earth's gravity?
Are you up against twisting forces at all? Or just vibration? SMD component has a bit less mass, so for a given amount of vibration it'll put less stress on it's joints. On the other hand, the through-hole component won't care nearly as much if the board is flexing because it's got wire leads that should let it move around a bit.
A friend of mine who does control boards for windturbines (read: high-vibration, high-reliability) swears by SMD, specifically BGA, even QFP have too long pins and will suffer from fatigue too soon in that application.
The shorter the pins, the stiffer the mount and the higher the reliability.
You must protect the board from warping, though, because there is no give in those stiff connections and BGAs will crack before jumping off the board.
for design in previous role for defence and aerospace, we would subject our boards and enclosures to large amounts of shock and vibration in order to comply with the required standards.
From a construction perspective, any boards with larger components (or where possible) would be fitted with anti vibration (AV) mounts of some sort.
Generally we never had issues with surface mount components. Large electrolytic capacitors, inductors, transformer and so on would normally need some extra adhesive. Poly-sulphide was used for this, not a pleasant chemical to work with, but it would meet the stringent environmental requirements of the defence and aerospace industries, and stop the large components from vibrating off.
Mike is right: less mass is better, since the part will exert less force on the soldering when vibrating. So the SMT will be better, even when PTH has a larger soldering contact. (An 0402 resistor weighs only 1 milli-gram).
When I was in college we learned that in very high-vibration environments they would use wire-wrapping instead of soldering. But that's many moons ago, and I guess soldering techniques have improved since. The space shuttle didn't exist yet (talking about high-vibration)