Friday, September 21, 2012

When a chemist is too productive

Courtesy of Boston-area news tweep and chemist Brian D'Amico, I was alerted to the story of the shutdown of the Massachusetts state drug testing laboratory in August due to irregularities. But another tweet from @FreeRadical1 noted the political fallout from it, with respect to Governor Deval Patrick:
“We’re dealing with, by all ­accounts, a rogue chemist who for many years, going back to 2003 or 2004, has not done her job, and that’s gone undetected for a long time,” Patrick said. “I know there are those who want to say this has to do with the budget challenges of the last two years, but that’s just not borne out by the evidence.”
I gotta say, a "rogue chemist" being blamed for troubles is pretty funny. Of course, the rogue chemist in mind, Ms. Annie Khan (or Dookhan) has gotten the Masschusetts criminal justice system in a good bit of trouble:
In a letter sent to members of the Massachusetts Association of Criminal Defense Attorneys, association president Max D. Stern said his group, along with federal and state public defenders, were given more details on the extent of the scandal involving the former state chemist. 
During the meeting, Stern wrote, the defense bar was given some insight into why the State Police have notified prosecutors that 64,000 drug samples — representing 34,000 criminal cases — may now be tainted as a matter of law and as a matter of science. 
“The lab analyst in question had unsupervised access to the drug safe and evidence room, and tampered with evidence bags, altered the actual weight of the drugs, did not calibrate machines correctly, and altered samples so that they would test as drugs when they were not,’’ Stern wrote in the letter.
What I found most remarkable about Ms. Khan/Dookhan's troubles is how they had not been caught:
On Thursday, top Patrick administration officials said that lab director Dr. Linda L. Han had resigned and director of analytical chemistry Julie ­Nassif had been fired. Disciplinary proceedings are now also underway against Dookhan’s direct super­visor, the officials said.
Administration officials said Dookhan’s supervisors missed obvious signs of problems. 
In 2004, for example, Dookhan processed 9,239 samples while her peers tested an average of 2,938 samples. (emphasis mine)
When someone at an analytical laboratory is 3 times as productive as their colleagues, people should be concerned about both quality and pencil-whipping. 


  1. The streets of Boston are going to look a lot more interesting once all of these guys get out because their cases are overturned.

  2. "When someone at an analytical laboratory is 3 times as productive as their colleagues, people should be concerned about both quality"

    I agree with this.

    Wouldn't this also be an issue in drug discovery? I recall working for a, I assume, typical biotech in which chemists were evaluated by the number of compounds registered. To be fair, the compounds were QCed before being allowed into system, but this seemed to me a great incentive to pursue easier chemistry, and to register every single intermediate. Oddly, that company never produced anything of value and lost a ton of money for its VC backers.

  3. "When someone at an analytical laboratory is 3 times as productive as their colleagues, people should be concerned about both quality and pencil-whipping."

    Or maybe not. I'm recalling the Melendez-Dias v. Massachusetts case that went to the Supreme Court. In that case, the report from the crime lab was one line: "The substance was found to contain: Cocaine."

    The case was not about the quality of the report, but about whether or not this report was considered "testimonial" and therefore whether or not Melendez-Dias could bring the analyst into the court room for cross-examination. The court decided that analytical reports such as this are testimony and subject to cross-examination - a wise decision if you ask me - but still beside the point here. At no point in the testimony did anyone complain about the shortness of the report. It was implicitly assumed to be a good quality report. If a crime lab is not being gigged (and still weren't even after this decision) for 1-line reports, why change? Maybe these "rogue" chemists were writing 1-liners, while everyone else was writing 3-liners. Would the quality really be any better?

    Furthermore, we all know that all samples are not the same. If someone is just running FTIR's, they can go through samples much faster than someone running a DSC to look for melting point or a PCR to match DNA. Did everyone run all the instruments, or did they specialize?

    Obviously, there is a lot more going on here than is being reported, and the resignations certainly point to guilt. My point is simply that looking at the sample thoughput numbers is insufficient evidence to pass judgement on anyone.

    1. John, I think that's a fair critique, which is why I said that people should be concerned, not "people should suspect." I think you're right on the instrumentation issue, but one hopes that they were comparing apples to apples, etc.

      Presumably, the lack of following protocols, etc., tampering w/evidence, etc. is a bigger cause for concern. But ceteris paribus, I would have found the productivity numbers to be a pink flag, if not a red one.

    2. Having now worked in two GMP-overseen analytical labs, I can attest there are people who are faster than others, and a 3:1 output difference isn't necessarily unheard of, especially as John pointed out, for different assays. Running 100 appearance or pH tests in a day is a heck of a lot faster than 100 HPLC assays.

      However, the key part of this story (that's been pointed out by most everyone here as well) is that the review of data was the missing component. That and the free access to the drug cabinet. Considering how tight the access controls are to the Schedule I cabinets at my company I'm amazed there wasn't even a simple countersignature requirement there.

      As a friend of a Suffolk county (Boston) DA, I don't envy her workload stemming from this right now.

    3. Working 40 hr/wk and 50 wk/yr, 9239 samples equates 1 sample processed every 13 minutes. I've never worked in this area, but that seems like impressive productivity regardless of the instrument. It's even more impressive with the substantial documentation requirement I imagine they're working under. Comparatively, her colleagues run samples at an average rate at once every 40 minutes.

    4. 13 minutes per sample is not implausible. Say, I run a GC analysis with a program that takes 30 minutes. I can run 2 samples simultaneously. I prepare 46 samples, analyze results of the last batch of samples, check and prepare the instrument. I run the sequence overnight, resulting in about 10.5 minutes spent per sample. And that's if I need 30 minutes. I imagine that if my samples are obvious meth and I analyze only for meth, I can run it much, much faster.