Compromised Blinding

Source material is catalogued in two companion resources:

• Protocol, Annotated — Blinding & Placebo Comparability — nine verbatim sources read side by side (NIEHS RFP Amendment 1992, Protocol v9, Protocol v10, TLC Group 1998 design paper, TLC Group 2000 Pediatric Research, Rogan 2001 NEJM, Dietrich 2004 Pediatrics, Chen 2005), with a cross-source contradiction map and scholarly commentary on each.
• knowledge/topics/blinding.md — full verbatim archive with provenance notes and key-date timeline (source-of-truth markdown file in the knowledge base).

This limitation page presents the argument; the two resources above hold the sources.

Succimer is a sulfur-based chelating compound with a strong, distinctive mercaptan odor and taste. In 2004, TLC Project Officer Walter Rogan, MD described this characteristic at an NIH Grand Rounds presentation (Rogan, NIH Grand Rounds, October 27, 2004):

“If I opened a bottle of succimer, down here, those of you in the last row would find it detectable in about 5 minutes, and intolerable in about 15…if you ever opened a rotten egg, that’s what it smells like… it’s a nasty drug.”

Any parent who opened a bottle of active drug would have been immediately aware that it contained something with a strong sulfurous smell — an obvious threat to double-blind integrity in a trial where parents were responsible for administering the medication.

The 1992 RFP Amendment explicitly recognized the blinding challenge:

“Can parents be made blind to treatment, if children given active drug smell like mercaptan? Possibly not. Offerors could consider proposing to query parents about whether they knew if the child was getting active drug, and see if they get it right. NIEHS knows of no agent that smells like a mercaptan and could serve as a placebo.”

The same RFP further stated:

“In general, the more open the design, the more likely it is that questions of bias will be raised about the results. Proposals that do not offer truly random treatment assignment and blind psychometric assessment are proposing in essence a different study than what was approved at NIEHS.”

Despite acknowledging that adequate blinding might not be achievable, and recommending that investigators query parents about treatment assignment, the trial proceeded as a “double-blind” study. No such assessment of parental awareness was ever reported.

TLC randomized participants from August 1994 to January 1997 (TLC Group 1998, p. 317). Two protocol versions describe incompatible blinding methods:

Protocol Version 9 (August 23, 1994) — Internal Contradiction:

• Section 4.6 stated: “a vented cylindrical plastic canister…will be filled with 100 mg of succimer and added to all bottles of study drug”
• Section 9.1.1 stated: “a two cm² piece of filter paper which has been soaked with Mucomyst (acetylcysteine) 20% solution [will be placed] into each bottle of placebo drug. An unsoaked piece of filter paper will be placed inside each bottle of succimer”

These two sections describe incompatible methods within the same protocol version.

Protocol Version 10 (November 4, 1997) — Changed Methods:

• Section 4.6: Changed to “200 mg of succimer” in canisters added to “all bottles”
• Section 9.1.1: Changed to “a small canister containing 200 mg of active drug into each bottle of placebo drug. A canister containing 200 mg of placebo will be placed inside each bottle of succimer”

If blinding was compromised, the direction of bias is not straightforward. Two opposing mechanisms are possible:

• Expectation bias (away from null): Parents who knew their child received succimer might have reported more favorable outcomes, which would bias results toward finding a treatment effect. This did not happen — the result was null.
• Adherence attenuation (toward null): Parents who knew their child was receiving placebo might have been less diligent about administration, sought treatment elsewhere, or dropped out — reducing the effective difference between groups and biasing results toward the null.

The absence of any blinding assessment means neither mechanism can be evaluated. CONSORT guidelines recommend assessing whether blinding was maintained, particularly when the intervention has distinctive sensory characteristics. This was not done.