Limitations

Succimer achieved ~11 µg/dL separation at week 1, collapsing to 4.5 at 6 months, 2.7 at 12 months, and zero by the 36- and 60-month IQ measurements. The trial tested whether a transient, past reduction produces lasting cognitive benefit years after the drug effect vanished.

Children were chelated and returned to lead-contaminated homes. Mean peak BLL post-randomization was 30.6 µg/dL (Chen et al. 2005), exceeding the 26.2 baseline. Dust lead levels rebounded to near-baseline within 24 months (Farfel et al. 2000). Protocol Section 4.7 acknowledged re-exposure as a competing explanation; the analysis to distinguish it was never published.

Ferritin was reported once, as arithmetic mean, in a growth paper (Peterson et al. 2004), with SDs nearly equaling the means — indicating a skewed distribution that obscures depletion prevalence. No publication analyzed iron status as an effect modifier of the treatment effect on IQ. Iron supplementation was withheld during chelation.

628 of 780 participants (81%) and 149 siblings ingested vitamins containing 32–36 µg Pb/tablet for up to 3 months during the follow-up period when cognitive outcomes were being measured. No sensitivity analysis excluding exposed children from the primary IQ outcome was performed.

Retreatment rates were reported as five different values (75–85%) across six sources for the same completed dataset. The vitamin recall analysis excluded 26 of 104 unexposed children without explanation.

The post-hoc 96% power claim (Rogan et al. 2001) requires an unreported covariate R² ≈ 0.45, nearly triple the 0.16 assumed in the design paper. The 3 IQ point detectable difference was at the upper bound of what contemporaneous meta-analyses supported for this disadvantaged, high-BLL population.

The retreatment threshold of 15 µg/dL was based on an assumed 10 µg/dL separation. The nonlinear dose-response relationship (Lanphear et al. 2005) means a reduction from 26 to 16 µg/dL predicts only ~1.5 IQ points of benefit — well below the 3-point difference the trial was powered to detect.

Succimer has a strong mercaptan odor that the 1992 RFP acknowledged might make blinding impossible. Blinding methods changed between Protocol Versions 9 and 10 with internal contradictions. No blinding assessment was ever reported.

MEMS electronic adherence data was collected at the Cincinnati site but never reported. All published adherence figures are based on pill counts alone — a method known to overestimate true adherence.

Four sites with different environmental interventions, undefined sampling frames, and varied recruitment pathways. No site × treatment interaction was tested. If Baltimore (with better environmental control) showed a treatment effect, pooling would have masked it.

Each limitation independently biased results toward the null, introduced uncontrolled confounding, or compromised validity. Their combined effect may have precluded detection of a true treatment effect.