A Regulatory Perspective on Threats to the Integrity of

Analgesic Clinical Trial Efficacy DataSharon Hertz, MDDivision Director

Division of Anesthesia, Analgesia, and Addiction Products

FDA/CDER

Disclaimer

The content of this talk does not necessarily reflect the views of the FDA, and is entirely based on my own observations and viewpoints.

I have no potential conflicts of interest to report.

2

TIACTED

Errors in the design, the conduct, the data collection process, and the analysis of a randomized trial have the potential to affect not only the safety of the patients in the trial, but also, through the introduction of bias, the safety of future patients.*

* Colin Baigent, Frank E Harrell, Marc Buyse, Jonathan R Emberson and Douglas G Altman, Ensuring trial validity by data quality assurance and diversification of monitoring methods, Clinical Trials 2008; 5: 49–55

3

TIACTED

Beyond the potential to affect safety, threats to clinical trial data integrity affect the ability to demonstrate efficacy and can substantially increase the time to get new products to market.

4

What Are TIACTED?• Inadequate study design• Sloppy study conduct

–Poor training/supervision of clinical site staff, patients

–Protocol violations by clinical site staff, patients

–Unverifiable data/audit trail

5

What Are TIACTED?

• Intentional actions that negatively affect clinical trial data integrity– Deceptive subjects– Fraudulent data– Intentional failure to adhere to protocol– Improper handing of data– Deviation from prespecified analyses

6

Example 1 - Investigator Fraud?

Study Study 1 Study 2 Study 3

Treatment Study Drug

Placebo Study Drug

Placebo Study Drug

Placebo

PID, VAS 24 hours 72 hours 72 hours

Mean ± SD -46 ± 22 -13. ± 13 -57 ± 16 -20 ± 12 -31 ± 21 -31 ± 21

Difference from placebo

LS Mean (95% C.I.)

-32 (-37, -28) -36 (-40, -32) -0.7 (-5, 3)

p-value <0.0001 <0.0001 0.76

Study Location Non-US Non-US US

7

• 3 clinical efficacy trials with very similar design: 2 successful, 1 failed

Example 1

• Successful studies– Large effect size, larger than expected – Higher baseline pain intensity– Less use of rescue, non-drug treatment– Lower placebo response

• Smaller change PI, 0% placebo reported onset of meaningful PR

8

Example 1

• What could explain the difference?– Demographics mostly similar– Looked for treatment by site effect – results

not driven by one site– Compared to other similar product trial results

including US and other non-US trials – no other studies with similar placebo response or effect size

9

Example 1• Site Inspections of 3 sites common to both non-US

studies, based on high enrollment numbers– 1 site

• Study nurse transcribed PI notes “to be legible”, destroyed original documents

• 21 subjects enrolled in both studies, 14 of whom injured and enrolled on same day, for both studies

• 17 of 55 in study 1, 6 of 35 in study 2 - part of a pair or triplet with same surname and/or address and many with same day of injury

• Site excluded from analysis

10

Example 1• Findings from site 1 led to evaluation for similar patterns

of enrollment from other sites– All sites had some same-day enrollment from related

subjects or subjects sharing an address, and multiple subjects enrolled in both studies

– Applicant explained that multiple members of the same family or household could sustain the same injury, on the same day, repeatedly, because people in this country more active than US

– Unable to verify identity of any subjects based on local privacy laws

11

Example 2 – Failure to Follow Protocol

• 2 clinical efficacy trials, one single site for both studies

• Inspection findings– Failure to record safety variables, investigator

felt protocol required too frequent recording of vital signs (although research assistant present to record dosing)

– No automated blood pressure machine available for baseline measures

12

Example 2

• Additional data requested• Possible safety problems, data insufficient to

adequately characterize safety

13

Example 3 – Improper Handling of Data

• 1st review cycle – routine inspections –– Protocol deviations that could impact the

validity, reliability, and integrity of data– Applicant failed to report protocol violations in

final study report– Accidental unblinding at several sites

14

Example 3• 2nd review cycle - pivotal study repeated, routine

inspections of 2 clinical sites and applicant– Statisticians extracted data for SAS datasets

with unblinded treatment assignment field prior to database lock

– Variable subsequently blinded– Datasets (not actual data) deleted– Applicant claimed statisticians either did not

view data or had no interaction with sites or the critical outcome data

15

Example 3

• Applicant failed to notify FDA when event occurred, even though unblinding contributed to initial CR

• Failed to maintain audit trails for the deletion of datasets

• FDA unable to confirm attestations of statisticians, no longer with company

16

End Note• Three examples demonstrated importance

of early identification of data integrity problems, corrections may salvage study

• Better clinical trial monitoring may help identify problems earlier

• Important to notify FDA, may be able to help

• Best approach – avoid these problems

17