IMO E.
Premises:
Of 2,500 people who survived a first heart attack, those who did not smoke had their first heart attack at a median age of 62.
However, of those 2,500, people who smoked two packs of cigarettes a day had their first heart attack at a median age of 51.
Conclusion:
nonsmokers tend to have a first heart attack eleven years later than do people who smoke two packs of cigarettes a day.

The premises speak about people who survived first heart Attack. The conclusion is a a generalized exaggerated statement about all people who has Heart Attack. [Including survivors and Dead]

Task is to find a statement that explains this flaw.

(A) the relative severity of heart attacks suffered by smokers and nonsmokers

What is the relative severity is given? Can we confirm the eleven year gap the conclusion is talking about? NO

(B) the nature of the different medical treatments that smokers and nonsmokers received after they had survived their first heart attack
Nature of the treatment has once again nothing to do with 11 years gap between S and NS
(C) how many of the 2,500 people studied suffered a second heart attack

This is irrelevant to our argument. We are concerned only with first HA people.

(D) the earliest age at which a person who smoked two packs a day had his or her first heart attack

What if we know this? We still need to know the earliest age at which non smoker had heart Attack. If we need to precisely tell the difference, we need to know earliest ages of both S and NS.

(E) data on people who did not survive a first heart attack.

This is correct. If we know this data, then we will know for sure the count of people who survived/did not survive, Smokers/non Smokers. We can calculate the 11 year gap easily.
Observe one more point- Option E says "data on people...". It does not confine itself to S data/NS data. So it means to say that we get data of both S and NS. Hence helpful.

Hope this helps!!

yes my ans is E too...

dont feel the need of explaination after such a good post by praveen...

D is irrelevant to the argument at hand...whenever verifying an option always check whether it is helping to answer the question asked...

Dutsch750

Another reason why D may not be the best choice is that the earliest age at which heart attack occurred may not be representative of the entire data set. What if only 1 person of of those who used to smoke 2 packs of ciggerates a day got a heart attack at the early age of 35. But all the rest got a heart attack at a much older age.

Median/mean are values representative of the entire data set in this case, not the least/highest value.