I'm not sure if there is a procedure in medicine more maligned than the autopsy. Early modern European docs, by and large our modern forbears, were peeling recently deceased people apart since the very, very early stages of the renaissance (and, if Roy Porter is to be believed in his seminal-but-goddawfully-boringly-named history of medicine, The Greatest Benefit To Mankind, it may have happened even before then), and my guess is following every major advance in diagnostic medicine, somebody declared the need for the autopsy and research in the gross anatomy lab to be finished. This was followed by defenders of the practice pointing out that our faith in our diagnostics is grounded more in hubris than in reality; fortunately through the magic of the NY Times website I can even produce evidence of such a diatribe--from 1921! The opener: "The need for a change in attitude on the part of the public toward the autopsy and the fact that the very limited number of autopsies performed [in the US] acts as a handicap in the instruction of the [sic] interne and medical students are points which are being earnestly discussed in medical circles." Nearly 80 years later, the same complaints were being made.
Nevertheless, a skeptic of the need for modern autopsies could point to a few key technologies that really have changed the playing field of diagnosis in favor of the physician in the past two decades. When discussing it with patients I refer to this technology as "looking under the hood," and it consists mainly of CT scans, MRIs, and ultrasounds (including echocardiograms, which are just ultrasounds of the heart, yet nobody calls them that, no doubt contributing to the perplexity of laypeople on medical topics). Various maladies that used to be able to hide out in the chest, abdomen and pelvis--such as, for instance, a liver abscess--avoiding detection typically light up on various scans. When coupled with the dozens of biochemical tests we now have at our disposal, it's hard to assert that we don't know what's going on when people die. So the argument goes. "I mean, seriously, back when I was a med student you didn't really know why people were dying, and you wanted to know how good your tests were to see if you got the diagnosis right on someone who died," a very senior physician once told me when dismissing the need for routine autopsies. "But now, I'd say we know better than 95 percent of the time. Anyone who says otherwise is dreaming."
Like many physicians--probably the majority of physicians who harbor any opinions about autopsies--I generally agree with this sentiment, although I heard a presentation the other day that brought me up short. While discussing kidney disease occurring in the midst of HIV infection, a colleague posted some intriguing data that just recently was published in Kidney International. The group, led by Christina Wyatt at Mount Sinai, looked at a subset of HIV-infected patients who were part of a cohort called the Manhattan HIV Brain Bank, and had agreed to undergo autopsy following their deaths. These researchers, not interested in neurological pathology, exclusively looked at microscopic slides of the kidneys of 89 of these patients. And some of their findings struck me: 4 patients had the premortem diagnosis of glomerulonephritis (not a trivial condition), while 8 were found to have evidence of it at autopsy; 1 patient with the premortem diagnosis of interstitial nephritis, but 5 had it at autopsy; no patients were found to have chronic pyelonephritis (kidney infection) prior to autopsy, though 7 cases were discovered under the microscope. Overall, while 75 of the 89 patients were found to have some form of microscopic renal pathology at autopsy, only 27 carried the premortem clinical diagnosis of chronic kidney disease.
It isn't clear to me whether or not these microscopic findings bear relevance on the actual proximal and contributory causes of death for these patients, but it is hard to dismiss the idea that they might be. At the risk of sweeping generalizations, physicians often look for the simplest, most straightforward explanation of a medical event, but this study seems to indicate that, at the cellular level, there may be many pathological events, interacting with one another and creating feedback loops that lead to a medical event (like a heart attack, or abrupt kidney failure, or a blood clot in the lungs, or a stroke, or any number of other things). If so, some, or possibly many of our diagnoses may need to come with asterisks attached.
Two caveats: one is that this study doesn't have any impact on the debate about the gross anatomy autopsy (i.e. looking at a body with the naked eye); with the accuracy of cat scans, it is still almost impossible for abscesses and tumors to "hide." Two is that HIV is a particularly complicated disease, and researchers from bedside docs to bench scientists are still trying to understand how the disease progresses--so this was a patient population most likely to yield surprising findings like this. Would you find hidden kidney pathology in, say, a breast cancer cohort? Not sure. Nevertheless, with studies such as this, the days of autopsy are, I suspect, not yet finished.