BLQ Values at the Profile End

Why the Profile Endpoint Needs Care

A concentration profile may end with one or more samples reported as below the limit of quantification (BLQ). Those results matter, but they do not provide exact concentrations: they indicate that the assay could not quantify the analyte at those times. In noncompartmental analysis, the usual observed-AUC endpoint is therefore the time of the last quantifiable concentration, not the time of the last scheduled sample.

This distinction keeps two different ideas separate. Observed AUC summarizes the measured profile through its last quantifiable point. AUC extrapolated to infinity estimates the remaining area after that point using a terminal elimination-rate estimate. Later BLQ observations neither supply exact areas nor, by themselves, determine the terminal slope.

Definition: Let \(t_{\mathrm{last}}\) be the time of the last quantifiable concentration, and let \(C_{\mathrm{last}}\) be the concentration measured there. The observed area through that endpoint is denoted \(\mathrm{AUC}_{0-t_{\mathrm{last}}}\). A later BLQ sample is a result below the assay's lower limit of quantification (LLOQ); it is not an observed concentration equal to zero.

Observed AUC Ends at the Last Quantifiable Sample

Under the common, prespecified noncompartmental convention for end-of-profile BLQ results, later BLQ samples are not used to extend observed AUC. The area is calculated between quantifiable observations up to \(t_{\mathrm{last}}\), often with the linear trapezoidal rule for the illustration here. For adjacent quantifiable observations \((t_i,C_i)\) and \((t_{i+1},C_{i+1})\), that rule gives:

\mathrm{AUC}_{i\rightarrow i+1} = \frac{C_i+C_{i+1}}{2}(t_{i+1}-t_i)

This calculation assumes a straight-line concentration change between the two observed values for the purpose of numerical integration. It does not mean that the biological concentration truly follows a straight line between samples. The area has units of concentration multiplied by time, such as \(\mathrm{mg\cdot h/L}\).

Worked Example: Calculating Observed AUC Before Trailing BLQ Samples

Suppose a hypothetical profile has concentrations of 0, 2.0, 3.2, 1.2, and 0.55 mg/L at 0, 1, 2, 4, and 6 hours, respectively. Samples at 8 and 10 hours are reported BLQ. Find the observed AUC using the linear trapezoidal rule.

The last quantifiable observation is \(0.55\ \mathrm{mg/L}\) at 6 hours, so \(t_{\mathrm{last}}=6\ \mathrm{h}\). Calculate the four areas through that time:

\begin{aligned} \mathrm{AUC}_{0-1} &= \frac{0+2.0}{2}(1)=1.00,\\ \mathrm{AUC}_{1-2} &= \frac{2.0+3.2}{2}(1)=2.60,\\ \mathrm{AUC}_{2-4} &= \frac{3.2+1.2}{2}(2)=4.40,\\ \mathrm{AUC}_{4-6} &= \frac{1.2+0.55}{2}(2)=1.75. \end{aligned}

Thus \(\mathrm{AUC}_{0-6}=1.00+2.60+4.40+1.75=9.75\ \mathrm{mg\cdot h/L}\). The BLQ results at 8 and 10 hours do not extend this observed-AUC endpoint. If the 8-hour BLQ result were incorrectly replaced with zero and integrated from 6 to 8 hours, the added trapezoid would be \((0.55+0)/2\times2=0.55\ \mathrm{mg\cdot h/L}\). That calculation treats an unknown sub-LLOQ concentration as exactly zero, which the assay result does not justify.

BLQ Results and Terminal Extrapolation

When a terminal log-linear phase is supported by quantifiable concentrations, the terminal elimination-rate constant, \(\lambda_z\), is estimated from the slope of the selected terminal observations after taking natural logarithms of concentration. In a simple regression, \(\ln C\) is related linearly to time, and \(\lambda_z\) is the negative of the fitted slope. The terminal extrapolated area is then estimated as \(C_{\mathrm{last}}/\lambda_z\), and the total area to infinity is the observed area plus that tail:

\mathrm{AUC}_{\mathrm{tail}}=\frac{C_{\mathrm{last}}}{\lambda_z}, \qquad \mathrm{AUC}_{0-\infty} = \mathrm{AUC}_{0-t_{\mathrm{last}}} + \mathrm{AUC}_{\mathrm{tail}}

This extrapolation relies on a positive, suitably estimated terminal rate constant and the assumption that the terminal decline can be represented by a log-linear phase. A BLQ observation cannot be entered into the log-concentration regression as though it were a quantified value: its concentration is not known exactly, and its logarithm is not available. A trailing BLQ result can be consistent with continued decline, but does not establish a numerical slope or prove that the concentration is zero.

Worked Example: Estimating the Terminal Tail

Continue the hypothetical profile above. For this arithmetic illustration, suppose the terminal slope has been selected as the decline between the quantifiable concentrations \(1.20\ \mathrm{mg/L}\) at 4 hours and \(0.55\ \mathrm{mg/L}\) at 6 hours. The two-point calculation is not, on its own, strong evidence that a terminal phase is well characterized; in an actual analysis, point selection and the adequacy of the terminal fit need to be assessed.

The estimated rate constant is the decline in log concentration per hour:

\lambda_z = \frac{\ln(1.20/0.55)}{6-4} = \frac{0.7801586}{2} = 0.3901\ \mathrm{h^{-1}}

Using the unrounded rate constant, the tail area is:

\mathrm{AUC}_{\mathrm{tail}} = \frac{0.55\ \mathrm{mg/L}}{0.3900793\ \mathrm{h^{-1}}} = 1.410\ \mathrm{mg\cdot h/L}

With the observed area of \(9.75\ \mathrm{mg\cdot h/L}\), the estimated total is \(9.75+1.410=11.16\ \mathrm{mg\cdot h/L}\). The tail is \(100(1.410/11.16)=12.63\%\) of that estimated total. This percentage describes the proportion of the estimated AUC supplied by extrapolation; it does not validate the terminal model or show that the BLQ samples were zero.

The calculation also shows why endpoint and slope decisions must be reported separately. The endpoint determines \(C_{\mathrm{last}}\) and the observed area. The selected terminal quantifiable observations determine \(\lambda_z\). A later BLQ sample does not replace \(C_{\mathrm{last}}\) in the tail formula, nor should it be used as a zero-concentration point to calculate \(\lambda_z\).

What a BLQ Result Can and Cannot Tell You

A BLQ result is informative as a threshold statement: the concentration was below the assay's LLOQ under the applicable reporting rules. It is not a precise concentration. Consequently, the actual area after the last quantifiable point is not known from the BLQ report alone. The standard observed-AUC convention avoids inventing that area; it does not claim that the true concentration became zero at \(t_{\mathrm{last}}\).

Worked Example: A BLQ Result Gives a Range, Not an Exact Area

Suppose the last quantifiable concentration is \(0.20\ \mathrm{mg/L}\) at 6 hours, and the 8-hour result is BLQ with an LLOQ of \(0.10\ \mathrm{mg/L}\). If one uses a linear trapezoid only to illustrate the uncertainty, the unobserved 8-hour concentration \(C_8\) is constrained to \(0\leq C_8<0.10\ \mathrm{mg/L}\), assuming a nonnegative concentration.

\mathrm{AUC}_{6-8} = \frac{0.20+C_8}{2}(2) = 0.20+C_8 \quad\Longrightarrow\quad 0.20\leq \mathrm{AUC}_{6-8}<0.30\ \mathrm{mg\cdot h/L}

The interval is an illustration under the stated interpolation rule, not an exact area obtained from the assay. Standard observed AUC stops at 6 hours in this example. Replacing the BLQ result with zero would choose the lower bound without evidence that the concentration was zero; replacing it with the LLOQ would choose a value at the assay threshold, also without evidence that this was the actual concentration.

Diagnostics, Alternatives, and Limitations

Before interpreting \(\mathrm{AUC}_{0-\infty}\), check whether the terminal estimate is credible. Inspect the quantifiable observations used to estimate \(\lambda_z\), their timing, and the log-concentration pattern. A sparse or curved terminal pattern, a slope that changes markedly when one point is added or removed, or a nonpositive fitted decline undermines the tail estimate. Trailing BLQ values can prompt scrutiny of the profile and assay sensitivity, but they cannot repair a weak terminal fit.

Also report or inspect the extrapolated fraction. A large fraction means that the total AUC depends substantially on a model-based tail rather than measured concentrations; the percentage alone is not a universal pass/fail rule. Interpretation depends on the study's analysis plan and applicable standards. The terminal rate estimate and extrapolated fraction should be considered alongside sampling design and the quality of the terminal observations.

There are alternatives to the usual endpoint convention, but they answer different questions and require explicit justification. Substituting zero, LLOQ, or LLOQ/2 assigns an exact value that the result did not provide and can alter area calculations. A censored-data or pharmacokinetic model may use the information that observations fall below a threshold, but its estimates depend on the model and assumptions; it is not the same as standard observed AUC. A more sensitive assay or later quantifiable samples may provide additional measured information. The chosen approach should be prespecified and applied consistently, rather than selected after seeing which value gives a preferred result.

Key takeaway: For the common noncompartmental convention, observed AUC ends at the last quantifiable concentration. Later BLQ results are neither exact zeros nor usable log-concentrations. Terminal extrapolation uses the last quantifiable concentration and a defensible \(\lambda_z\) estimate, and its uncertainty should be considered separately from the BLQ endpoint convention.

Common Pitfalls

Extending observed AUC to the final scheduled time. A scheduled sample time is not automatically an AUC endpoint. Use the last quantifiable concentration under the stated analysis convention.
Treating BLQ as zero. BLQ means below the quantification limit, not measured absence of analyte. Do not create an exact terminal trapezoid by substituting zero without a justified, prespecified method.
Using BLQ in a log-linear slope fit. A BLQ value has no known concentration to log-transform. Estimate \(\lambda_z\) from suitable quantifiable terminal observations.
Assuming a BLQ tail makes extrapolation reliable. The result is compatible with low concentrations, but does not establish a terminal slope or make a poor fit adequate.
Reporting total AUC without its extrapolated contribution. AUC to infinity includes a model-based tail. State the observed endpoint and assess the fraction extrapolated.

Check Your Understanding

Answer each question, then compare your reasoning with the solutions.

A profile has a quantifiable result at 10 hours followed by BLQ samples at 12 and 14 hours. Under the common observed-AUC convention, what is the endpoint?
Why is a BLQ result not appropriate as a zero-valued point in a terminal log-linear regression?
If \(C_{\mathrm{last}}=0.40\ \mathrm{mg/L}\) and \(\lambda_z=0.20\ \mathrm{h^{-1}}\), calculate the extrapolated tail area.
In one sentence, distinguish observed AUC from AUC to infinity when trailing samples are BLQ.
What does a large extrapolated fraction indicate, and what does it not establish?

Solutions. (1) The endpoint is 10 hours, the time of the last quantifiable concentration; the later BLQ times do not extend observed AUC under that convention. (2) BLQ does not provide an exact concentration, so its logarithm and its position on a log-linear concentration-time line are unknown. (3) \(\mathrm{AUC}_{\mathrm{tail}}=0.40/0.20=2.00\ \mathrm{mg\cdot h/L}\). (4) Observed AUC integrates through the last quantifiable point, whereas AUC to infinity adds a terminally extrapolated tail to that observed area. (5) It indicates that a substantial share of estimated total AUC comes from extrapolation rather than measured concentrations; by itself, it does not prove the terminal model is valid or invalid.

What You'll Learn