Ideally, we would have two groups to compare: levodopa only v levodopa plus carbidopa. In the absence of this Tupelo3's CDC data is very interesting. There are many ways to interpret this. In addition to those that he mentions, I would add that the results could reflect the fact that progress on treatments for PD has, since the introduction of levodopa in the 1970's, been slower than for many other diseases. To see how this could affect the statistics, consider a person aged 60 who survives a heart attack, one which 20 years ago would have been fatal. He may go on to be diagnosed with PD, and ultimately to die with it.

As I see it, the argument in favour of the carbidopa effect would be stronger if the true PD mortality rates had a step change immediately after its introduction but, thereafter, remained constant.

Some more evidence against the carbidopa theory is offered by Morgan et al.[1]:

"Our data indicate that survival was no better or worse in patients starting levodopa within two years of symptom onset versus greater than two years after symptom onset."

I think it is likely that the mention of levodopa in this context actually means levodopa/carbidopa. So, those people who delayed taking levodopa also delayed taking carbidopa. If the carbidopa hypothesis were right this should have given the late administration group an advantage, which was not found.

Reference

[1] "Mortality in Levodopa-Treated Parkinson's Disease"
John C. Morgan,1 Lillian J. Currie,2 Madaline B. Harrison,2 James P. Bennett Jr.,3 Joel M. Trugman,4 and G. Frederick Wooten
Parkinson's Disease (2014)
http://www.hindawi.com/journals/pd/2014/426976/