Why an Early BLQ Result Needs Careful Interpretation
A concentration reported as below the limit of quantification (BLQ) is not a measured zero. It indicates that the assay could not quantify the concentration reliably at that sampling time; under the usual threshold interpretation, the underlying concentration is below the lower limit of quantification (LLOQ). That distinction matters before the first quantifiable sample: the early observations can contribute to the area under the concentration-time curve (AUC), even though their exact concentrations are unavailable.
Handling these observations can change the calculated early AUC. It can also affect peak timing when timing is estimated from a fitted concentration-time model. By contrast, if \(C_{\max}\) and \(T_{\max}\) are defined from the observed, quantifiable samples, early BLQ values below the LLOQ ordinarily cannot exceed a quantifiable peak. The right interpretation therefore depends both on the handling rule and on how the exposure metric is defined.
An observed BLQ sample is different from a missing sample: the former records that an assay result fell below quantification, while the latter has no usable result at that time. This tutorial concerns the former. As in the earlier material on BLQ values at the profile end, a handling convention should be specified in the analysis plan rather than selected after seeing which option produces a preferred result. End-of-profile rules are not automatically the right rules for the beginning of a profile.
How Handling Choices Enter the AUC Calculation
For adjacent sampling times \(t_i\) and \(t_{i+1}\), with concentrations \(C_i\) and \(C_{i+1}\), the linear trapezoidal area is
This calculation treats the concentration as changing linearly between the two recorded or assigned values. If a BLQ value is assigned a number, that number enters the trapezoids on either side of its sampling time. The calculation does not recover the unknown concentration; it produces an AUC conditional on the assigned value and the interpolation rule.
Common operational choices for early BLQ observations include assigning zero, assigning a fraction of the LLOQ such as LLOQ/2, or excluding the BLQ observations from the numerical calculation. A protocol may prescribe another method. These choices answer different practical questions: zero is a convenient boundary assignment, LLOQ/2 is a positive point imputation, and exclusion avoids assigning a concentration but can leave early area unrepresented. None should be described as revealing the true concentration.
| Handling choice | What is calculated | Key interpretation or risk |
|---|---|---|
| Assign zero | Trapezoids use zero at BLQ sampling times. | A practical convention, but zero is not the measured concentration. |
| Assign LLOQ/2 | Trapezoids use half the quantification threshold. | A positive point imputation; the actual concentration may be below or above that assigned value. |
| Omit early BLQ observations | The calculation starts at a later included time, or uses a separately specified rule for the early interval. | May leave early AUC out of the result; the resulting time window must be stated. |
| Use a censored-data model | The model uses the fact that the result is below LLOQ rather than treating it as an exact point value. | Requires a justified model and assumptions; it is not a simple trapezoidal substitution. |
A common noncompartmental convention is to assign zero to BLQ samples before the first quantifiable concentration. It is a convention, not a claim that the concentration was actually zero. The governing protocol or applicable guidance determines which convention is appropriate in a particular analysis.
Worked Example: Comparing Early AUC Under Three Rules
Consider hypothetical concentrations with an LLOQ of \(2.00\ \text{mg/L}\). Samples are collected at 0.5, 1, 2, 3, and 4 hours. The first two results are BLQ; the quantified concentrations are \(4.00\ \text{mg/L}\) at 2 hours, \(10.0\ \text{mg/L}\) at 3 hours, and \(6.00\ \text{mg/L}\) at 4 hours. Calculate the AUC from 0.5 to 4 hours under zero assignment, LLOQ/2 assignment, and omission of the early BLQ observations.
Zero assignment. Assign \(0\ \text{mg/L}\) at 0.5 and 1 hour. The 0.5-to-1-hour area is \(0\). From 1 to 2 hours the area is \((0+4.00)(2-1)/2=2.00\ \text{mg·h/L}\). The later areas are \((4.00+10.0)(3-2)/2=7.00\ \text{mg·h/L}\) and \((10.0+6.00)(4-3)/2=8.00\ \text{mg·h/L}\). Thus the assigned-rule AUC is \(0+2.00+7.00+8.00=17.0\ \text{mg·h/L}\).
LLOQ/2 assignment. Here LLOQ/2 is \(1.00\ \text{mg/L}\). The area from 0.5 to 1 hour is \((1.00+1.00)(0.5)/2=0.500\ \text{mg·h/L}\); from 1 to 2 hours it is \((1.00+4.00)(1)/2=2.50\ \text{mg·h/L}\). Adding the same later areas gives \(0.500+2.50+7.00+8.00=18.0\ \text{mg·h/L}\).
Omission. If omission means starting the trapezoidal calculation at the first quantifiable sample, the included area is from 2 to 4 hours: \(7.00+8.00=15.0\ \text{mg·h/L}\). This is not an estimate of the same 0.5-to-4-hour AUC unless an additional rule supplies the omitted early area. The difference reflects the changed calculation window as well as the handling choice.
The three results are conditional on their stated rules. In particular, \(18.0\ \text{mg·h/L}\) does not show that the early concentrations were \(1.00\ \text{mg/L}\); it is the area obtained by using that imputed value.
What Can Be Inferred About the Missing Early Concentration?
A useful check is to ask what the BLQ result alone permits. Under the threshold interpretation and nonnegative concentrations, an unknown BLQ concentration \(C_b\) satisfies \(0 \leq C_b < L\), where \(L\) is the LLOQ. If the next concentration \(C_q\) is quantified at a time \(\Delta t\) later, the linear-trapezoidal area over that interval is
Because the coefficient of \(C_b\) is positive, the interval area increases as the unknown BLQ concentration increases. Substituting the permissible lower and upper limits gives
This is a bound on the trapezoidal area under those assumptions, not a confidence interval and not a claim that every value in the range is equally plausible. If a BLQ sample lies between two other sampled times, its assigned concentration affects both adjacent trapezoids; the same substitution logic applies to each interval.
Worked Example: Bounding the First Quantifiable Interval
Suppose the LLOQ is \(2.00\ \text{mg/L}\), a sample at 1 hour is BLQ, and the next sample at 2 hours is \(4.00\ \text{mg/L}\). Under linear interpolation, bound the 1-to-2-hour AUC contribution using only the BLQ threshold.
The time difference is \(\Delta t=1.00\ \text{h}\), and the unknown concentration \(C_b\) at 1 hour satisfies \(0\leq C_b<2.00\ \text{mg/L}\). Therefore
Zero assignment gives the lower-bound area, \(2.00\ \text{mg·h/L}\). LLOQ/2 assignment gives \((1.00+4.00)/2 \times 1.00=2.50\ \text{mg·h/L}\), an interior value under this bound. The BLQ result supports the range, but does not identify which of these values is the actual interval area.
Implications for Cmax and Peak Timing
For an observed \(C_{\max}\) and \(T_{\max}\), the usual definitions are the largest included concentration and its sampling time. If at least one sample is quantifiable, its concentration is at or above the LLOQ. Assigning zero or LLOQ/2 to earlier BLQ samples gives values below the LLOQ, so those assigned values cannot become the observed maximum. In ordinary circumstances, these assignments therefore change early AUC without changing the observed \(C_{\max}\) or \(T_{\max}\).
That conclusion depends on the definition and rule. If a BLQ value is assigned the full LLOQ, it is no longer assigned a value strictly below the quantification threshold. When the observed maximum is exactly the LLOQ, that assignment could create a tie with an earlier BLQ time; the reported \(T_{\max}\) could then depend on the tie-breaking convention. This is a reason not to silently replace BLQ with the LLOQ.
A model-estimated peak time is different from observed \(T_{\max}\). A fitted curve uses information across sampling times to estimate a continuous profile, and the early data can influence its fitted shape and peak time. Treating BLQ values as exact zeros can force a model toward an early concentration of zero; treating them as LLOQ/2 makes a different point-value assumption. A model that handles BLQ results as censored observations instead uses the event \(C<L\) as information. Each approach relies on its own assumptions, and a fitted peak time should be identified as model-based rather than presented as the observed sampling-time \(T_{\max}\).
Worked Example: Checking Whether Early BLQ Handling Changes Tmax
In a hypothetical profile, the LLOQ is \(1.00\ \text{mg/L}\). The results are BLQ at 0.5 and 1 hour, \(1.00\ \text{mg/L}\) at 2 hours, \(8.00\ \text{mg/L}\) at 3 hours, \(7.00\ \text{mg/L}\) at 4 hours, and \(2.00\ \text{mg/L}\) at 6 hours. Find observed \(C_{\max}\) and \(T_{\max}\) under zero and LLOQ/2 assignment, then assess a full-LLOQ assignment.
Under zero assignment, the early values are \(0\ \text{mg/L}\); under LLOQ/2, they are \(0.500\ \text{mg/L}\). In both cases the largest included value is \(8.00\ \text{mg/L}\) at 3 hours. Thus \(C_{\max}=8.00\ \text{mg/L}\) and \(T_{\max}=3\ \text{h}\) under either rule.
If the two BLQ values are instead assigned \(1.00\ \text{mg/L}\), they tie the quantified value at 2 hours but remain below the \(8.00\ \text{mg/L}\) peak. The reported peak time still remains 3 hours in this profile. In a different profile whose maximum quantified value were exactly \(1.00\ \text{mg/L}\), however, assigning the full LLOQ to an earlier BLQ sample could create a tie and change the earliest time reported as \(T_{\max}\). The result would depend on the tie rule as well as the inappropriate point assignment.
Diagnostics, Assumptions, and Limitations
Before calculating or interpreting exposure metrics, check the assay and the calculation rule rather than treating every BLQ pattern alike. A defensible analysis should make clear which samples were BLQ, the LLOQ that applied, how early observations were handled, and the time window represented by the reported AUC. It should also distinguish observed peak metrics from peak timing obtained by model fitting.
- Check the time order. Confirm that the observations called “before the first quantifiable sample” precede the first quantified result for that profile.
- Check units and threshold. The LLOQ and quantified concentrations must use the same concentration units. AUC units are concentration multiplied by time.
- Check the integration window. If early values are omitted, identify whether the reported area starts at the first quantifiable time or whether another rule accounts for the earlier interval.
- Check consistency. Apply the prespecified handling rule consistently across profiles and subjects. A different rule can make comparisons reflect data processing as well as concentration patterns.
- Inspect sensitivity. If early intervals are long or many early samples are BLQ, compare results under reasonable prespecified alternatives. Large differences indicate that early AUC depends strongly on the handling choice.
- Check model assumptions. Censored-data modeling requires an appropriate model for concentrations and the censoring threshold. A trapezoidal bound does not validate a model, and a model estimate does not remove uncertainty about BLQ concentrations.
The trapezoidal bound assumes nonnegative concentration, a threshold interpretation of BLQ, and linear interpolation over the interval. The actual concentration-time path need not be linear, so the bound describes the trapezoidal calculation conditional on its interpolation rule, not an exact bound on the biological integral without further assumptions. Sparse sampling and uncertainty in the assay are additional issues; they should not be resolved by treating a censored observation as exact.
Common Pitfall
Correction: BLQ means the concentration was not quantified below a stated threshold. Zero assignment may be a prespecified calculation convention, but it is not an assay measurement of zero. Describe the rule, report the resulting metric with its units, and consider whether the early AUC is sensitive to plausible alternative handling.
A related mistake is to claim that any early BLQ handling must change observed \(T_{\max}\). If the peak is selected among quantifiable samples and BLQ substitutions remain below the LLOQ, the observed maximum and its time are unchanged. Potential changes to a model-estimated peak time are a separate issue and should be discussed as such.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- An early sample is BLQ with an LLOQ of \(3.00\ \text{mg/L}\). What does the result establish, and what does it not establish?
- A BLQ sample at 1 hour precedes a quantified concentration of \(6.00\ \text{mg/L}\) at 3 hours. Under linear interpolation and an LLOQ of \(2.00\ \text{mg/L}\), calculate the range of possible trapezoidal AUC contributions for 1 to 3 hours.
- Why can replacing an early BLQ result with zero change early AUC but leave observed \(T_{\max}\) unchanged?
- If early BLQ values are omitted and the calculation starts at the first quantifiable sample, what must be stated when reporting the resulting AUC?
Solutions. (1) Under the threshold interpretation, the concentration is below \(3.00\ \text{mg/L}\); it is not established to equal zero or any particular positive value below the threshold. (2) Here \(\Delta t=2.00\ \text{h}\), \(C_q=6.00\ \text{mg/L}\), and \(L=2.00\ \text{mg/L}\). Thus \((2.00/2)(6.00)\leq A<(2.00/2)(2.00+6.00)\), so \(6.00\leq A<8.00\ \text{mg·h/L}\). (3) Zero assignment changes the concentration used in the early trapezoids, but a quantified peak is at or above the LLOQ and remains larger than assigned early values below that threshold; the observed maximum time therefore remains the same. (4) State that the calculation starts at the first quantifiable sample and give the AUC time window, because the omitted early interval is not included unless another specified rule supplies it.