ASA Connect

Pacira Biosciences:When  are scientists legally liable for a bad statistical analysis?

A few years ago, I posted about a legal case called FTC v. Quincy  Bioscience Holding Co. According to the district court opinion, Quincy had claimed that its product was successful based on a post-hoc subgroup analysis which found success, unadjusted for multiple comparisonss, in a small number of subgroups. The FTC (the product was not subject to the FDA) had sued for fraud in making this claim.

At the time, I had posted on this board saying that I don't think the results of a statistical analysis whose data is accurate and whose methodology is disclosed  should be liable for things like fraud, even though I'd completely agree that a post-hoc analysis of this kind is worthless and the positives found likely false.

At the time, I got a number of messages disagreeing, and there are is a statistician or two in this commumity who completely stopped communicating with me after that post, perhaps believing I favor coddling fraudsters.

Two developments have happened since then that suggest giving this issue a second look. And perhaps the statisticians who stopped speaking to me might even want to reconsider.

The first development was in the Quincy case. The 2nd Circuit Court of Appeals said the district court had gotten the facts completely wrong. Quincy hadn't merely advertised success in the subgroups based on the post-hoc analysis. Instead, it had claimed that the product was successful in the entire population based on the aubgroup analysis. This misrepresentation of the subgroup analysis results was the basis of the fraud claim, not the mere fact that a post-hoc subgroup analysis was done. The 2nd Circuit reversed the dismissal of the fraud claim and sent the case back to the district court for trial.

This is a completely different statement of what happened than what the district court had said. Simce I have no personal knowledge of what happened and based my earlier post only on the district court's assessment of the facts, I also got the facts wrong.

A recent case explains why it might be advisable not to make it too easy to find scientists. The American Society of Anesthesiology published a meta-analysis in Anesthesiology, its peer-reviewed journal, that concluded that Pacira Biosciences main product, EXPAREL (liposomal bupivacaine), was not superior to much cheaper generic bupivacaine, together with an editorial suggesting EXPAREL was not worth the higher price.

Pacira Biosciences sued the ASA, journal editors, and paper authors for libel. It claimed the meta-analysis was scientifically flawed, pointing to methodological errors. Pacira Biosciences claimed that among other defects, the meta-analysis omited studies favorable to EXPAREL was based on "crude pooling" rather than a more sophisticated pooling method. , heterogeneity was not properly assessed, etc.

On February 4, the district court dismissed the case. It said that scientific disputes are immune from libel claims. "Scientific controversies must be settled by the methods of science rather than by the methods of litigation." Accordingly, "scientific conclusions are protected speech to the extent they are 'drawn…from non-fraudulent data, based on accurate descriptions of the data and methodology underlying these conclusions, on subjects about which there is legitimate ongoing scientific disagreement." And as a result, "absent an allegation that the author of a scientific article falsified the data from which she drew her conclusions, a plaintiff cannot sustain a claim for trade libel by alleging that some methodological flaw led to a scientifically 'incorrect' answer."

Applying these principles, the district court said that Pacira Biosciences was alleging nothing more than methodological flaws. Whether these alleged flaws invalidate the conclusions or not is a matter for scientists, not courts. It dismissed the libel claims and threw out the case.

This case suggests that it might be best not to make it so easy for people to sue scientists for allegedly flawed published papers. If well-endowed individuals and corporations get to hire lawyers and experts to persuade lay jurors that a paper is methodologically wrong, scientists and non-profit scientific institutions who publish research that aggrieves powers that be may quickly find themselves under siege, and they may find jurors favoring the side with the smoother talkers.

It was for this reason that I had thought the district court had been right to dismiss the fraud claim in Quincy Bioscience Holding Co. Under the district court's statement of the facts, the FTC was claiming that post-hoc subgroup analyses are inherently fraudulent. That's an overstatement. While such analyses are not conclusive, they can be valuable generators of hypotheses for future research, and prohibiting their use would diminish science. Moreover, it would be an assault on the basic principle that scientific disputes should be resolved by scientific means, not by judges.

And that principle is worth preserving even though it unquestionably allows bad science through, and not infrequently.

This is different from government agencies imposing rules for acceptability of studies for particular societal purposes, e.g. clinical trials for drug approval, environmental, toxicology, and engineering studies, etc.

https://storage.courtlistener.com/recap/gov.uscourts.njd.469581/gov.uscourts.njd.469581.92.0.pdf

• https://www.nature.com/articles/502017a.pdf 
• https://schachtmanlaw.com/2014/07/24/subgroups-subpar-statistical-practice-versus-fraud/ (Nathan Schactman blog)
• https://hdsr.mitpress.mit.edu/pub/bd5k4gzf/release/3 (Deborah Mayo in the Harvard Data Science Review)
• https://www.bizjournals.com/sanfrancisco/news/2021/01/20/trump-pardon-scott-harkonen-actimmune-ipf.html (San Francisco Business)

Pacira Biosciences:When  are scientists legally liable for a bad statistical analysis?

A few years ago, I posted about a legal case called FTC v. Quincy  Bioscience Holding Co. According to the district court opinion, Quincy had claimed that its product was successful based on a post-hoc subgroup analysis which found success, unadjusted for multiple comparisonss, in a small number of subgroups. The FTC (the product was not subject to the FDA) had sued for fraud in making this claim.

At the time, I had posted on this board saying that I don't think the results of a statistical analysis whose data is accurate and whose methodology is disclosed  should be liable for things like fraud, even though I'd completely agree that a post-hoc analysis of this kind is worthless and the positives found likely false.

At the time, I got a number of messages disagreeing, and there are is a statistician or two in this commumity who completely stopped communicating with me after that post, perhaps believing I favor coddling fraudsters.

Two developments have happened since then that suggest giving this issue a second look. And perhaps the statisticians who stopped speaking to me might even want to reconsider.

The first development was in the Quincy case. The 2nd Circuit Court of Appeals said the district court had gotten the facts completely wrong. Quincy hadn't merely advertised success in the subgroups based on the post-hoc analysis. Instead, it had claimed that the product was successful in the entire population based on the aubgroup analysis. This misrepresentation of the subgroup analysis results was the basis of the fraud claim, not the mere fact that a post-hoc subgroup analysis was done. The 2nd Circuit reversed the dismissal of the fraud claim and sent the case back to the district court for trial.

This is a completely different statement of what happened than what the district court had said. Simce I have no personal knowledge of what happened and based my earlier post only on the district court's assessment of the facts, I also got the facts wrong.

A recent case explains why it might be advisable not to make it too easy to find scientists. The American Society of Anesthesiology published a meta-analysis in Anesthesiology, its peer-reviewed journal, that concluded that Pacira Biosciences main product, EXPAREL (liposomal bupivacaine), was not superior to much cheaper generic bupivacaine, together with an editorial suggesting EXPAREL was not worth the higher price.

Pacira Biosciences sued the ASA, journal editors, and paper authors for libel. It claimed the meta-analysis was scientifically flawed, pointing to methodological errors. Pacira Biosciences claimed that among other defects, the meta-analysis omited studies favorable to EXPAREL was based on "crude pooling" rather than a more sophisticated pooling method. , heterogeneity was not properly assessed, etc.

On February 4, the district court dismissed the case. It said that scientific disputes are immune from libel claims. "Scientific controversies must be settled by the methods of science rather than by the methods of litigation." Accordingly, "scientific conclusions are protected speech to the extent they are 'drawn…from non-fraudulent data, based on accurate descriptions of the data and methodology underlying these conclusions, on subjects about which there is legitimate ongoing scientific disagreement." And as a result, "absent an allegation that the author of a scientific article falsified the data from which she drew her conclusions, a plaintiff cannot sustain a claim for trade libel by alleging that some methodological flaw led to a scientifically 'incorrect' answer."

Applying these principles, the district court said that Pacira Biosciences was alleging nothing more than methodological flaws. Whether these alleged flaws invalidate the conclusions or not is a matter for scientists, not courts. It dismissed the libel claims and threw out the case.

This case suggests that it might be best not to make it so easy for people to sue scientists for allegedly flawed published papers. If well-endowed individuals and corporations get to hire lawyers and experts to persuade lay jurors that a paper is methodologically wrong, scientists and non-profit scientific institutions who publish research that aggrieves powers that be may quickly find themselves under siege, and they may find jurors favoring the side with the smoother talkers.

It was for this reason that I had thought the district court had been right to dismiss the fraud claim in Quincy Bioscience Holding Co. Under the district court's statement of the facts, the FTC was claiming that post-hoc subgroup analyses are inherently fraudulent. That's an overstatement. While such analyses are not conclusive, they can be valuable generators of hypotheses for future research, and prohibiting their use would diminish science. Moreover, it would be an assault on the basic principle that scientific disputes should be resolved by scientific means, not by judges.

And that principle is worth preserving even though it unquestionably allows bad science through, and not infrequently.

This is different from government agencies imposing rules for acceptability of studies for particular societal purposes, e.g. clinical trials for drug approval, environmental, toxicology, and engineering studies, etc.

https://storage.courtlistener.com/recap/gov.uscourts.njd.469581/gov.uscourts.njd.469581.92.0.pdf

Pacira Biosciences:When  are scientists legally liable for a bad statistical analysis?

A few years ago, I posted about a legal case called FTC v. Quincy  Bioscience Holding Co. According to the district court opinion, Quincy had claimed that its product was successful based on a post-hoc subgroup analysis which found success, unadjusted for multiple comparisonss, in a small number of subgroups. The FTC (the product was not subject to the FDA) had sued for fraud in making this claim.

At the time, I had posted on this board saying that I don't think the results of a statistical analysis whose data is accurate and whose methodology is disclosed  should be liable for things like fraud, even though I'd completely agree that a post-hoc analysis of this kind is worthless and the positives found likely false.

At the time, I got a number of messages disagreeing, and there are is a statistician or two in this commumity who completely stopped communicating with me after that post, perhaps believing I favor coddling fraudsters.

Two developments have happened since then that suggest giving this issue a second look. And perhaps the statisticians who stopped speaking to me might even want to reconsider.

The first development was in the Quincy case. The 2nd Circuit Court of Appeals said the district court had gotten the facts completely wrong. Quincy hadn't merely advertised success in the subgroups based on the post-hoc analysis. Instead, it had claimed that the product was successful in the entire population based on the aubgroup analysis. This misrepresentation of the subgroup analysis results was the basis of the fraud claim, not the mere fact that a post-hoc subgroup analysis was done. The 2nd Circuit reversed the dismissal of the fraud claim and sent the case back to the district court for trial.

This is a completely different statement of what happened than what the district court had said. Simce I have no personal knowledge of what happened and based my earlier post only on the district court's assessment of the facts, I also got the facts wrong.

A recent case explains why it might be advisable not to make it too easy to find scientists. The American Society of Anesthesiology published a meta-analysis in Anesthesiology, its peer-reviewed journal, that concluded that Pacira Biosciences main product, EXPAREL (liposomal bupivacaine), was not superior to much cheaper generic bupivacaine, together with an editorial suggesting EXPAREL was not worth the higher price.

Pacira Biosciences sued the ASA, journal editors, and paper authors for libel. It claimed the meta-analysis was scientifically flawed, pointing to methodological errors. Pacira Biosciences claimed that among other defects, the meta-analysis omited studies favorable to EXPAREL was based on "crude pooling" rather than a more sophisticated pooling method. , heterogeneity was not properly assessed, etc.

On February 4, the district court dismissed the case. It said that scientific disputes are immune from libel claims. "Scientific controversies must be settled by the methods of science rather than by the methods of litigation." Accordingly, "scientific conclusions are protected speech to the extent they are 'drawn…from non-fraudulent data, based on accurate descriptions of the data and methodology underlying these conclusions, on subjects about which there is legitimate ongoing scientific disagreement." And as a result, "absent an allegation that the author of a scientific article falsified the data from which she drew her conclusions, a plaintiff cannot sustain a claim for trade libel by alleging that some methodological flaw led to a scientifically 'incorrect' answer."

Applying these principles, the district court said that Pacira Biosciences was alleging nothing more than methodological flaws. Whether these alleged flaws invalidate the conclusions or not is a matter for scientists, not courts. It dismissed the libel claims and threw out the case.

This case suggests that it might be best not to make it so easy for people to sue scientists for allegedly flawed published papers. If well-endowed individuals and corporations get to hire lawyers and experts to persuade lay jurors that a paper is methodologically wrong, scientists and non-profit scientific institutions who publish research that aggrieves powers that be may quickly find themselves under siege, and they may find jurors favoring the side with the smoother talkers.

It was for this reason that I had thought the district court had been right to dismiss the fraud claim in Quincy Bioscience Holding Co. Under the district court's statement of the facts, the FTC was claiming that post-hoc subgroup analyses are inherently fraudulent. That's an overstatement. While such analyses are not conclusive, they can be valuable generators of hypotheses for future research, and prohibiting their use would diminish science. Moreover, it would be an assault on the basic principle that scientific disputes should be resolved by scientific means, not by judges.

And that principle is worth preserving even though it unquestionably allows bad science through, and not infrequently.

This is different from government agencies imposing rules for acceptability of studies for particular societal purposes, e.g. clinical trials for drug approval, environmental, toxicology, and engineering studies, etc.

https://storage.courtlistener.com/recap/gov.uscourts.njd.469581/gov.uscourts.njd.469581.92.0.pdf

Pacira Biosciences:When  are scientists legally liable for a bad statistical analysis?

A few years ago, I posted about a legal case called FTC v. Quincy  Bioscience Holding Co. According to the district court opinion, Quincy had claimed that its product was successful based on a post-hoc subgroup analysis which found success, unadjusted for multiple comparisonss, in a small number of subgroups. The FTC (the product was not subject to the FDA) had sued for fraud in making this claim.

At the time, I had posted on this board saying that I don't think the results of a statistical analysis whose data is accurate and whose methodology is disclosed  should be liable for things like fraud, even though I'd completely agree that a post-hoc analysis of this kind is worthless and the positives found likely false.

At the time, I got a number of messages disagreeing, and there are is a statistician or two in this commumity who completely stopped communicating with me after that post, perhaps believing I favor coddling fraudsters.

Two developments have happened since then that suggest giving this issue a second look. And perhaps the statisticians who stopped speaking to me might even want to reconsider.

The first development was in the Quincy case. The 2nd Circuit Court of Appeals said the district court had gotten the facts completely wrong. Quincy hadn't merely advertised success in the subgroups based on the post-hoc analysis. Instead, it had claimed that the product was successful in the entire population based on the aubgroup analysis. This misrepresentation of the subgroup analysis results was the basis of the fraud claim, not the mere fact that a post-hoc subgroup analysis was done. The 2nd Circuit reversed the dismissal of the fraud claim and sent the case back to the district court for trial.

This is a completely different statement of what happened than what the district court had said. Simce I have no personal knowledge of what happened and based my earlier post only on the district court's assessment of the facts, I also got the facts wrong.

A recent case explains why it might be advisable not to make it too easy to find scientists. The American Society of Anesthesiology published a meta-analysis in Anesthesiology, its peer-reviewed journal, that concluded that Pacira Biosciences main product, EXPAREL (liposomal bupivacaine), was not superior to much cheaper generic bupivacaine, together with an editorial suggesting EXPAREL was not worth the higher price.

Pacira Biosciences sued the ASA, journal editors, and paper authors for libel. It claimed the meta-analysis was scientifically flawed, pointing to methodological errors. Pacira Biosciences claimed that among other defects, the meta-analysis omited studies favorable to EXPAREL was based on "crude pooling" rather than a more sophisticated pooling method. , heterogeneity was not properly assessed, etc.

On February 4, the district court dismissed the case. It said that scientific disputes are immune from libel claims. "Scientific controversies must be settled by the methods of science rather than by the methods of litigation." Accordingly, "scientific conclusions are protected speech to the extent they are 'drawn…from non-fraudulent data, based on accurate descriptions of the data and methodology underlying these conclusions, on subjects about which there is legitimate ongoing scientific disagreement." And as a result, "absent an allegation that the author of a scientific article falsified the data from which she drew her conclusions, a plaintiff cannot sustain a claim for trade libel by alleging that some methodological flaw led to a scientifically 'incorrect' answer."

Applying these principles, the district court said that Pacira Biosciences was alleging nothing more than methodological flaws. Whether these alleged flaws invalidate the conclusions or not is a matter for scientists, not courts. It dismissed the libel claims and threw out the case.

This case suggests that it might be best not to make it so easy for people to sue scientists for allegedly flawed published papers. If well-endowed individuals and corporations get to hire lawyers and experts to persuade lay jurors that a paper is methodologically wrong, scientists and non-profit scientific institutions who publish research that aggrieves powers that be may quickly find themselves under siege, and they may find jurors favoring the side with the smoother talkers.

It was for this reason that I had thought the district court had been right to dismiss the fraud claim in Quincy Bioscience Holding Co. Under the district court's statement of the facts, the FTC was claiming that post-hoc subgroup analyses are inherently fraudulent. That's an overstatement. While such analyses are not conclusive, they can be valuable generators of hypotheses for future research, and prohibiting their use would diminish science. Moreover, it would be an assault on the basic principle that scientific disputes should be resolved by scientific means, not by judges.

And that principle is worth preserving even though it unquestionably allows bad science through, and not infrequently.

This is different from government agencies imposing rules for acceptability of studies for particular societal purposes, e.g. clinical trials for drug approval, environmental, toxicology, and engineering studies, etc.

https://storage.courtlistener.com/recap/gov.uscourts.njd.469581/gov.uscourts.njd.469581.92.0.pdf

Pacira Biosciences:When  are scientists legally liable for a bad statistical analysis?

A few years ago, I posted about a legal case called FTC v. Quincy  Bioscience Holding Co. According to the district court opinion, Quincy had claimed that its product was successful based on a post-hoc subgroup analysis which found success, unadjusted for multiple comparisonss, in a small number of subgroups. The FTC (the product was not subject to the FDA) had sued for fraud in making this claim.

At the time, I had posted on this board saying that I don't think the results of a statistical analysis whose data is accurate and whose methodology is disclosed  should be liable for things like fraud, even though I'd completely agree that a post-hoc analysis of this kind is worthless and the positives found likely false.

At the time, I got a number of messages disagreeing, and there are is a statistician or two in this commumity who completely stopped communicating with me after that post, perhaps believing I favor coddling fraudsters.

Two developments have happened since then that suggest giving this issue a second look. And perhaps the statisticians who stopped speaking to me might even want to reconsider.

The first development was in the Quincy case. The 2nd Circuit Court of Appeals said the district court had gotten the facts completely wrong. Quincy hadn't merely advertised success in the subgroups based on the post-hoc analysis. Instead, it had claimed that the product was successful in the entire population based on the aubgroup analysis. This misrepresentation of the subgroup analysis results was the basis of the fraud claim, not the mere fact that a post-hoc subgroup analysis was done. The 2nd Circuit reversed the dismissal of the fraud claim and sent the case back to the district court for trial.

This is a completely different statement of what happened than what the district court had said. Simce I have no personal knowledge of what happened and based my earlier post only on the district court's assessment of the facts, I also got the facts wrong.

A recent case explains why it might be advisable not to make it too easy to find scientists. The American Society of Anesthesiology published a meta-analysis in Anesthesiology, its peer-reviewed journal, that concluded that Pacira Biosciences main product, EXPAREL (liposomal bupivacaine), was not superior to much cheaper generic bupivacaine, together with an editorial suggesting EXPAREL was not worth the higher price.

Pacira Biosciences sued the ASA, journal editors, and paper authors for libel. It claimed the meta-analysis was scientifically flawed, pointing to methodological errors. Pacira Biosciences claimed that among other defects, the meta-analysis omited studies favorable to EXPAREL was based on "crude pooling" rather than a more sophisticated pooling method. , heterogeneity was not properly assessed, etc.

On February 4, the district court dismissed the case. It said that scientific disputes are immune from libel claims. "Scientific controversies must be settled by the methods of science rather than by the methods of litigation." Accordingly, "scientific conclusions are protected speech to the extent they are 'drawn…from non-fraudulent data, based on accurate descriptions of the data and methodology underlying these conclusions, on subjects about which there is legitimate ongoing scientific disagreement." And as a result, "absent an allegation that the author of a scientific article falsified the data from which she drew her conclusions, a plaintiff cannot sustain a claim for trade libel by alleging that some methodological flaw led to a scientifically 'incorrect' answer."

Applying these principles, the district court said that Pacira Biosciences was alleging nothing more than methodological flaws. Whether these alleged flaws invalidate the conclusions or not is a matter for scientists, not courts. It dismissed the libel claims and threw out the case.

This case suggests that it might be best not to make it so easy for people to sue scientists for allegedly flawed published papers. If well-endowed individuals and corporations get to hire lawyers and experts to persuade lay jurors that a paper is methodologically wrong, scientists and non-profit scientific institutions who publish research that aggrieves powers that be may quickly find themselves under siege, and they may find jurors favoring the side with the smoother talkers.

It was for this reason that I had thought the district court had been right to dismiss the fraud claim in Quincy Bioscience Holding Co. Under the district court's statement of the facts, the FTC was claiming that post-hoc subgroup analyses are inherently fraudulent. That's an overstatement. While such analyses are not conclusive, they can be valuable generators of hypotheses for future research, and prohibiting their use would diminish science. Moreover, it would be an assault on the basic principle that scientific disputes should be resolved by scientific means, not by judges.

And that principle is worth preserving even though it unquestionably allows bad science through, and not infrequently.

This is different from government agencies imposing rules for acceptability of studies for particular societal purposes, e.g. clinical trials for drug approval, environmental, toxicology, and engineering studies, etc.

https://storage.courtlistener.com/recap/gov.uscourts.njd.469581/gov.uscourts.njd.469581.92.0.pdf

Pacira Biosciences:When  are scientists legally liable for a bad statistical analysis?

A few years ago, I posted about a legal case called FTC v. Quincy  Bioscience Holding Co. According to the district court opinion, Quincy had claimed that its product was successful based on a post-hoc subgroup analysis which found success, unadjusted for multiple comparisonss, in a small number of subgroups. The FTC (the product was not subject to the FDA) had sued for fraud in making this claim.

At the time, I had posted on this board saying that I don't think the results of a statistical analysis whose data is accurate and whose methodology is disclosed  should be liable for things like fraud, even though I'd completely agree that a post-hoc analysis of this kind is worthless and the positives found likely false.

At the time, I got a number of messages disagreeing, and there are is a statistician or two in this commumity who completely stopped communicating with me after that post, perhaps believing I favor coddling fraudsters.

Two developments have happened since then that suggest giving this issue a second look. And perhaps the statisticians who stopped speaking to me might even want to reconsider.

The first development was in the Quincy case. The 2nd Circuit Court of Appeals said the district court had gotten the facts completely wrong. Quincy hadn't merely advertised success in the subgroups based on the post-hoc analysis. Instead, it had claimed that the product was successful in the entire population based on the aubgroup analysis. This misrepresentation of the subgroup analysis results was the basis of the fraud claim, not the mere fact that a post-hoc subgroup analysis was done. The 2nd Circuit reversed the dismissal of the fraud claim and sent the case back to the district court for trial.

This is a completely different statement of what happened than what the district court had said. Simce I have no personal knowledge of what happened and based my earlier post only on the district court's assessment of the facts, I also got the facts wrong.

A recent case explains why it might be advisable not to make it too easy to find scientists. The American Society of Anesthesiology published a meta-analysis in Anesthesiology, its peer-reviewed journal, that concluded that Pacira Biosciences main product, EXPAREL (liposomal bupivacaine), was not superior to much cheaper generic bupivacaine, together with an editorial suggesting EXPAREL was not worth the higher price.

Pacira Biosciences sued the ASA, journal editors, and paper authors for libel. It claimed the meta-analysis was scientifically flawed, pointing to methodological errors. Pacira Biosciences claimed that among other defects, the meta-analysis omited studies favorable to EXPAREL was based on "crude pooling" rather than a more sophisticated pooling method. , heterogeneity was not properly assessed, etc.

On February 4, the district court dismissed the case. It said that scientific disputes are immune from libel claims. "Scientific controversies must be settled by the methods of science rather than by the methods of litigation." Accordingly, "scientific conclusions are protected speech to the extent they are 'drawn…from non-fraudulent data, based on accurate descriptions of the data and methodology underlying these conclusions, on subjects about which there is legitimate ongoing scientific disagreement." And as a result, "absent an allegation that the author of a scientific article falsified the data from which she drew her conclusions, a plaintiff cannot sustain a claim for trade libel by alleging that some methodological flaw led to a scientifically 'incorrect' answer."

Applying these principles, the district court said that Pacira Biosciences was alleging nothing more than methodological flaws. Whether these alleged flaws invalidate the conclusions or not is a matter for scientists, not courts. It dismissed the libel claims and threw out the case.

This case suggests that it might be best not to make it so easy for people to sue scientists for allegedly flawed published papers. If well-endowed individuals and corporations get to hire lawyers and experts to persuade lay jurors that a paper is methodologically wrong, scientists and non-profit scientific institutions who publish research that aggrieves powers that be may quickly find themselves under siege, and they may find jurors favoring the side with the smoother talkers.

It was for this reason that I had thought the district court had been right to dismiss the fraud claim in Quincy Bioscience Holding Co. Under the district court's statement of the facts, the FTC was claiming that post-hoc subgroup analyses are inherently fraudulent. That's an overstatement. While such analyses are not conclusive, they can be valuable generators of hypotheses for future research, and prohibiting their use would diminish science. Moreover, it would be an assault on the basic principle that scientific disputes should be resolved by scientific means, not by judges.

And that principle is worth preserving even though it unquestionably allows bad science through, and not infrequently.

This is different from government agencies imposing rules for acceptability of studies for particular societal purposes, e.g. clinical trials for drug approval, environmental, toxicology, and engineering studies, etc.

https://storage.courtlistener.com/recap/gov.uscourts.njd.469581/gov.uscourts.njd.469581.92.0.pdf