IQ is standardized to have a mean (average) of 100 and a standard deviation of 15. It has been extensively shown that it follows approximately a normal distribution. Hence, the probability of having an IQ over X equals P and we have E als the expected number of people in the world with this IQ or higher (assuming 7 billion inhabitants):
x=160, p=0.000032, e=221 699
x=170, p=0.0000015, e=10 714
x=180, p=0.000000048, e=337
x=190, p=0.00000000099, e=6.9
x=200, p=0.000000000013, e=0.092
x=210, p=0.00000000000011, e=0.00079

As for measurement, that's a completely different discussion. Even for regular IQ tests, there is quite some measurement error and people can "study" for them somewhat so they are certainly flawed a least a little bit.
Secondly, for proper calibration, you also need a reference population. As you can see from the calculations above, the number of people with a high IQ is simply way too small to accurately and precisely determine a reference above 160 which is why good IQ tests to measure those extreme IQ's are rare.
Thus, I completely agree with what Yeon states, "there are no credible source that claims to be able to measure anything that high." because if there was any source claiming it, it would probably not be credible. ;-)

Edited 12/25/2013 10:04:27

how do they even get the math result for this?
x=160, p=0.000032, e=221 699
x=170, p=0.0000015, e=10 714
x=180, p=0.000000048, e=337
x=190, p=0.00000000099, e=6.9
x=200, p=0.000000000013, e=0.092
x=210, p=0.00000000000011, e=0.00079

???

You calculate the cumulative probability under a normal distribution with mean 100 and standard deviation 15. Most statistical software (I used R) has a built in function to do this. (See for example http://en.wikipedia.org/wiki/Intelligence_quotient#Current_tests for a brief explanation on this, more extensive, scientific explanations can be found on the web.)

Everyone can check these results in Excel and/or OpenOffice Calc with the command 1-NORMDIST(160,100,15,1) although rounding error may apply there (see http://support.microsoft.com/kb/827371 for the use of NORMDIST in Excel and an explanation on the precision).

The expected numbers are simply calculated as e=p*7 000 000 000