Peltier devices work best with Pure DC, Not a high frequency PWM.
Most important to try and keep the Hot side, Not Hot.
Pulsing it On and Off at a Slow Rate can work quite good.
Example: 5 to 10 seconds on, 5 to 10 seconds off, ETC.
So your advice is, "Don't do PWM, do PWM instead?"?
It's already been explained why any form of pulsing/PWM, regardless of the rate, is a bad idea. The heat transfer is proportional to the current but the resistive losses a proportional to the current squared (I^2*R), so feeding 2A at 50% duty cycle has the same heat transfer as DC 1A, but double the resistive losses. Given how significant the I^2*R losses are to the thermals of a peltier device, this is massively significant.
The OP is doing "filtered PWM", which is just a slightly odd way of wording "a buck converter". Which generates (practically) DC as its output. So please don't be derailing him/her with this incorrect advice.
Yes, there should be no reason why low frequency PWM is any better than high frequency PWM, apart from lower skin effect losses, which won't account for much unless yuou're talking about stupidly high frequencies.
Let's put some more figures into the equation.
Suppose the resistance, i.e. the part of the load which doesn't do any useful work, is 3Ohm and the full power current consumption at 12V is 2A.
Full power consumption:
P = I
2R = 2
2×8 = 12W
50% power consumption, using PWM.
P = 12*0.5 = 6W
50% power consumption, using half the DC current.
P = I
2R = 1
2×8 = 3W
Conclusion: reducing the current by half uses a half of the power of PWM at the full current.