The data sheet for the ADuCM360 indicates that the Integral Nonlinearity of the ADC is +- 10ppm of FSR (Gain = 1)

If my FSR is 2.5V that means the Integral Nonlinearity can cause a 2.5mV error.

Is this the worst case error at the at the mid point (FSR /2)?

How does the Integral Nonlinearity error affect measurements at the lower and upper ends of the FSR?

I'm calculating it to be 0.025mV not 2.5mV. (10ppm is 0.001%)

Yes typically the worst case would be around the midpoint.

At the lower and upper ends the offset and gain would have a bigger effect respectively.