Eight-month-old Matthew Eappen collapsed in the care of his British nanny.
He had a large blood clot (subdural hematoma) between the brain and a tough membrane that lies between the brain and skull.
Police and prosecutors believed Matthew's nanny shook him and slammed his head against something to stop his crying. She was tried for murder in the high-profile "nanny trial."
The trial became a battle of experts as forensic and neuropathologists argued about whether Matthew's injuries were inflicted shortly before his collapse or were from an earlier accidental injury and had started to bleed again.
In medical school, I was told that surgeons know nothing and do everything, internists know everything and do nothing and pathologists know everything and do everything - too late. Sadly, pathologists' ability to answer some important questions is limited.
Estimates of the time between injury and death are based on the appearance of injured tissue under the microscope, but reported ranges are broad, leading to bitter disputes.
Because only tiny pieces are looked at under the microscope, pathologists select areas for study from large organs or tissues. The selection is, in part, random, but there are guidelines (pieces from the edge and from the middle; pieces from anyplace that looks different).
How can we be sure the pieces examined fairly and completely represent the whole? Why were only so many pieces selected? Why pieces from here and not from there? Why was the remainder of Matthew's tissue discarded, preventing defense experts from examining more?
Such were the arguments about Matthew Eappen.
Questions about certainty and selection bias aren't specific to the Eappen case or to forensic practice. When organs or sizeable pieces of tissue are removed surgically, the pathologist looks at only tiny bits.
If cancer is seen, fine. If not, does that prove there's none elsewhere? If the cancer looks indolent, fine. But might an aggressive kind requiring more or different therapy be lurking somewhere in the specimen? If the cancer hasn't invaded the part examined, surgical removal without chemotherapy should be adequate. But how can we know it's not invasive half an inch away?
All pathologists are aware of the possibility of selection bias, but collective experience suggests normal sampling techniques work. If tumors diagnosed as benign harbored cancers, those cancers would come back. When cancer does recur after a favorable diagnosis, it's almost always because the original sections were misinterpreted, not because of sampling.
If injury dating is based on microscopic changes in pieces that are not representative, additional pieces (perhaps taken by a defense expert) should show something different. That's unlikely, but not impossible.
Despite everyday experience, should the undeniable possibility of selection bias constitute reasonable doubt?
The jury didn't think so. They convicted Matthew's nanny of second-degree murder. The judge wasn't so sure. He sentenced her to time served and ordered her released.
Dr. Carol J. Huser, a forensic pathologist, has served as La Plata County coroner since January 2003.