Analysis of Below Detectable Biomarker Lab Values

Options for analyzing below detectable lab values

Simple options include:

  1. If the % of below detectable (BD) is very high, analyze as a binary variable (detectable vs. undetectable)
  2. If %BD is moderate (<25%), create discrete categories (like tertiles, quartiles, or quintiles)
  3. If %BD is rare (<5%), and values are highly skewed, replace BD values with the LOD/2, where LOD = limit of detection
  4. If %BD is rare (<5%), and values are less skewed, replace BD values with the LOD/sqrt(2)

Options 3 and 4 can introduce bias and reduce power; if the %BD is extremely rare, this should have little effect on the results.

More complex options:

   1. If the lab value is the outcome, use Tobit regression to handle BD values (note that lab value may first need to be normalized, depending on the software used):

In Stata:

In SAS: use proc lifereg or proc qlim:

   2. Use multiple imputation to replace BD values with imputed values. The missing at random assumption is violated, so a modified approach using a Gibbs sampler is needed:

   3. Use propensity score weighting (model probability of being BD, then incorporate the result in the model using detectable values).



Baker SG, Fitzmaurice GM, Freedman LS, Kramer BS. Simple adjustments for randomized trials with nonrandomly missing or censored outcomes arising from informative covariates. Biostatistics. 2006;7(1):29–40. [PubMed]

Hughes MD. Analysis and design issues for studies using censored biomarker measurements with an example of viral load measurements in HIV clinical trials. Stat Med. 2000;19:3171–3191. [PubMed]

Lambert D, Peterson B, Terpenning I. Nondetects, detection limits, and the probability of detection. J Am Stat Assoc. 1991;86:266–277.

Lubin JH, Colt JS, Camann D, Davis S, Cerhan JR, Severson RK, et al. Epidemiologic evaluation of measurement data in the presence of detection limits. Environ Health Perspect. 2004;112:1691–1696. [PMC free article] [PubMed]