As usual, it depends.  I plugged some stuff into an ICM calculator, based on payouts of 40%, 30%, 20%, 10%.  I briefly experimented with some other payout structures and it didn't make a huge difference.

No ante, roughly even stacks (besides you) — shove 17.6%, 22+ A2s+ ATo+ K9s+ KJo+ Q9s+ J9s+

3000 ante, roughly even stacks — 25.5%, 22+ A2s+ A8o+ K6s+ KTo+ Q8s+ QTo+ J8s+ JTo T8s+ 98s

No ante, 3 stacks your size, 3 stacks 552,000 — 17.8%, 33+ A2s+ ATo+ K9s+ KJo+ Q9s+ QJo JTs

No ante, 2 short stacks 100,000, 3 big stacks 65400 — 16.4%, 44+ A3s+ A9o+ KTs+ KJo+ QTs+ JTs

No ante, payouts, 4 tiny stacks 50,000, 1 huge stack — 8.4%, 99+ A9s+ ATo+ KJs+

3000, ante, payouts, 4 tiny stacks 50,000, 1 huge stack in SB — 7.4%, 88+ ATs+ AJo+ KQs

3000, ante, payouts, 4 tiny stacks 50,000, 1 huge stack in BB — 12.4%, 77+ A7s+ A5s-A4s ATo+ KTs+ KQo QTs+

3000, ante, payouts, 4 tiny stacks 50,000, 1 huge stack UTG — 38.9%, 33+ Ax+ K2s+ K6o+ Q5s+ Q8o+ J7s+ J9o+ T8s+

So assuming optimal play, we should be shoving 44 unless the stack sizes reward folding pretty drasitcally.  On a side note, it's interesting how much it affects optimal play to move the huge stack from the SB to BB.

Given your reads that your opponents were more likely to call, I think I would probably adjust to fold in situations where 44 is close to the border of your shoving range.  This is because I believe the added chip equity you gain by getting to be on the good side of a flip more often is countered by the $ev penalty of risk of going bust.