Duration is a measure that provides an approximation of the effect of a small parallel shift in the interest rate term structure on a bond's price. However, it does not provide a good approximation when the change in the bond yield arises from a non-parallel shift in the interest rate term structure or when the change being considered is large. This is because duration assumes that the yield curve shifts in a parallel manner, which is not always the case in real-world scenarios. Additionally, duration is a linear measure, which means it does not accurately capture the effects of large changes in interest rates. For large changes, the relationship between bond prices and yields is better captured by convexity, which considers the curvature of the price-yield relationship.

Choice B is incorrect. Duration actually provides a good approximation of the effect of a small parallel shift in the interest rate term structure. This is one of its primary uses, and it performs well under these conditions.

Choice C is incorrect. Contrary to this statement, duration is most effective when changes in interest rates are small. When there are large changes in interest rates, duration may not provide an accurate estimate due to convexity effects.

Choice D is incorrect. As explained above, options B and C do not accurately describe the limitations of using duration as a measure of interest rate risk.