Why a sequence of concentrations is not automatically a trend
A concentration result is a snapshot taken under particular conditions. Across visits, a higher or lower value can reflect a change in exposure, but it can also reflect a different time since dose, a missed or delayed dose, a change in regimen, or measurement variability. A sequence of numbers is therefore not enough: interpreting a trend requires checking what was held comparable and what changed.
The aim is not to prove a cause from concentrations alone. It is to decide whether the observations support a sustained change in exposure, whether the apparent change has a plausible alternative explanation, or whether the available information is insufficient. The minimum metadata for interpreting a sample and the earlier material on within-patient variability provide useful context for that assessment.
Make the comparison interpretable
Before comparing values, list the factors that could change their expected level. At a minimum, check the dose and dosing interval, formulation and route, time of the last dose, time of sampling relative to that dose, and whether the patient took doses as prescribed in the days before sampling. Also check whether the clinical state, relevant interacting treatment, and laboratory method changed. For an oral drug, a sample taken during absorption is not directly interchangeable with one taken later in the dosing interval. For an infusion, the infusion and post-infusion timing matter.
A practical comparison uses the same sampling condition at each visit whenever possible—for example, a trough collected immediately before a scheduled dose under a consistent regimen. If collection conditions differ, do not silently treat the concentrations as equivalent. Document the difference and decide whether it can be interpreted with a suitable pharmacokinetic profile. If it cannot, label the comparison as uncertain rather than reporting a precise trend.
One useful descriptive quantity is the relative change from an earlier value \(C_1\) to a later value \(C_2\):
The concentration units cancel in this calculation, so the result is a percentage. The percentage describes the observed change; it does not by itself establish that exposure changed by that amount. When visits are separated by different lengths of time or there are several observations, a slope on the log-concentration scale can summarize the direction and rate of change.
Why consistent conditions matter: a simple model
The logic can be made explicit with a conceptual model. Let \(Y_i=\log(C_i)\) be the log of the observed concentration at visit \(i\), and let \(t_i\) be calendar time. A simplified representation is:
Here, \(\beta\) represents a time-related change under the model; \(\Delta\tau_i\) is the difference between the visit's sampling time relative to dose and a chosen reference time; \(A_i\) summarizes relevant adherence differences; and \(\epsilon_i\) includes residual biological and measurement variation. The coefficients \(\gamma\) and \(\delta\) represent how those factors affect log concentration in this simplified description. This is not a universal pharmacokinetic model: its purpose is to show that a time-related pattern can be confused with visit-condition effects if those effects are not controlled or accounted for.
If the regimen and sampling time are consistent and adherence is credible, the nuisance terms may be similar across visits, making a time-related comparison more informative. If those conditions differ, a raw slope can combine several effects. With only a few observations, it may not be possible to separate them. Even a visually smooth sequence does not resolve that identifiability problem.
Worked comparisons
Worked Example: A sustained rise under matched trough conditions
Suppose a hypothetical patient has trough concentrations of 10.0 mg/L, 11.0 mg/L, and 12.1 mg/L at visits one month apart. The dose, formulation, dosing interval, sampling procedure, and laboratory method are unchanged, and the patient reports taking the regimen consistently. Calculate the total relative change and a descriptive log-slope.
The first-to-last relative change is:
For the log-slope over the two-month interval:
The fitted first-to-last multiplicative change is \(\exp(0.0953\times2)=1.210\), or 21.0%. The intermediate observation also rises, so the values are consistent with a sustained upward pattern under the stated conditions. This supports investigating a real exposure change, but does not establish its cause. The trend could still reflect an unrecorded change in adherence, physiology, or another relevant factor.
Worked Example: Sampling time explains an apparent decline
Suppose two hypothetical samples were collected on the same unchanged regimen. At visit A, the concentration was 12.4 mg/L at 10 hours after a dose. At visit B, it was 9.4 mg/L at 12 hours after a dose. For illustration, assume an appropriate patient-specific profile predicts reference concentrations of 12.0 mg/L at 10 hours and 9.0 mg/L at 12 hours, with concentration proportional to the patient's overall exposure over this small comparison.
The unadjusted percentage change is:
That calculation alone suggests a decline, but the second sample was collected later in the dose interval. Compare each observation with the profile prediction for its actual timing:
Both observations are about 3–4% above their timing-specific predictions. Alternatively, express visit B on the 10-hour reference scale using the assumed profile ratio:
The timing-standardized visit B value, 12.53 mg/L, is close to the 12.4 mg/L observed at visit A. Under the stated assumptions, the apparent raw decline is largely explained by the later sampling time, and these data do not show a clear exposure decline. In practice, the timing adjustment is only as credible as the profile and its assumptions; without an appropriate profile, report the samples as non-comparable rather than applying an arbitrary correction.
Worked Example: A missed dose is a competing explanation
At two hypothetical visits on an otherwise unchanged regimen, a patient's comparable trough concentrations are 8.0 mg/L and 6.0 mg/L. The later visit's history indicates that the patient missed a dose shortly before sampling. The unadjusted relative change is:
This is a 25.0% lower measured concentration at the later visit, but the missed dose makes it unsafe to interpret that difference as a sustained decline under the prescribed regimen. There is not enough information here to calculate what the later concentration would have been without the missed dose; that would require a suitable pharmacokinetic model and reliable dose history. The sound conclusion is that the observed result is lower and adherence offers a plausible explanation, not that the patient's underlying exposure has definitively fallen by 25.0%.
Worked Example: Is a small ratio distinguishable from analytical variability?
Suppose two hypothetical concentrations are 10.0 mg/L and 11.5 mg/L, measured under matched conditions. Assume, for illustration, that each assay result has an independent analytical coefficient of variation of 8%, and ignore other sources of variation. The observed ratio is \(11.5/10.0=1.15\), a 15.0% increase.
For a lognormal measurement error with coefficient of variation \(CV=0.08\), the log-scale standard deviation is:
If the two analytical errors are independent and have the same variance, the standard deviation of the log ratio is:
Under a normal approximation on the log scale, a range covering about 95% of ratios expected from analytical error alone, when the true ratio is 1, is:
The observed ratio of 1.15 lies within this illustrative range. Therefore, under these assumptions, analytical variability alone could plausibly account for a difference of this size. This calculation does not include within-patient biological variability and is not a universal decision threshold; it only shows why a modest difference should not automatically be called a trend.
Interpreting several visits and checking for problems
With three or more comparable observations, examine both direction and consistency. A steady sequence is more compatible with a sustained trend than one isolated high or low value surrounded by similar concentrations. A log-slope can summarize the overall direction when visit times are not equally spaced, but the slope can hide reversals and outliers. Inspect the individual points as well as the summary. If a fitted log-linear description is used, its interpretation depends on the approximate adequacy of that description over the time span; it does not imply that concentrations must change linearly or exponentially indefinitely.
The following checks help identify when a trend interpretation may be failing:
- Unequal sampling times: verify actual dose and collection times rather than relying only on the intended schedule. A small timing difference may matter greatly near a rapidly changing part of a concentration profile.
- Regimen or formulation changes: mark dose changes, missed doses, delayed doses, and formulation or route changes. A concentration change after one of these events may be a response to the changed regimen, not a trend under stable treatment.
- Isolated discordant results: check sample identity, collection procedure, contamination risk, and analytical uncertainty before allowing one point to dominate the interpretation.
- Changing clinical context: consider whether illness, organ function, interacting treatment, or other relevant conditions changed between visits. A trend may be real while its explanation remains uncertain.
- Too few or too noisy observations: do not infer a sustained trajectory from a pair of results when visit conditions differ or plausible variability is of similar size to the observed change.
Choosing a comparison approach
| Approach | Useful when | Main limitation |
|---|---|---|
| Compare matched concentrations directly | Dose, timing, and other key conditions are consistent across visits. | Differences in unrecorded adherence or biological variability can still affect results. |
| Standardize for sampling time with an appropriate profile | Collection times differ and a suitable pharmacokinetic profile and assumptions are available. | An incorrect profile or unsupported adjustment can create false precision. |
| Compare AUC estimates | Exposure over a specified interval is the relevant measure and appropriate samples support estimation. | AUC estimation has its own sampling and model limitations; it is not interchangeable with every single concentration. |
The simplest defensible approach is usually to compare samples collected under the same conditions. Timing standardization or AUC-based comparison can be helpful when the clinical question calls for it and the data support it. Neither approach removes the need to check adherence, regimen changes, and measurement uncertainty.
Common mistake and key limitation
Wrong reasoning: “The concentration fell from 12.4 to 9.4 mg/L, so the patient's exposure fell by 24.2%.” This treats two values collected at different times after dose as if they represented the same sampling condition and treats the observed difference as a confirmed exposure change.
Correction: First establish whether the samples are comparable. In the worked timing example, the later collection time explained most of the raw difference under the stated profile. If conditions differ and cannot be credibly standardized, report the numerical change but qualify its interpretation. Concentration trends can guide clinical review, but observational patterns alone do not identify their cause.
Check Your Understanding
Answer each question, then check the brief solutions below.
- Matched hypothetical troughs rise from 5.0 mg/L to 5.8 mg/L. What is the observed relative change, and what does it establish?
- Why might a concentration sampled 2 hours after a dose not be directly compared with a concentration sampled 8 hours after a dose?
- A later result is lower after a patient reports missing a dose. What conclusion is justified, and what conclusion is not?
- For a simple log-linear trend with slope \(0.05\ \text{month}^{-1}\), what multiplicative change does the model imply over 2 months?
- Name two conditions, other than concentration itself, that should be checked before interpreting a serial pattern.
Solutions. 1. The relative change is \((5.8-5.0)/5.0\times100\%=16.0\%\). It establishes a 16.0% difference between the observed values, not by itself a sustained exposure trend or its cause. 2. Concentration may change with time since dose, so the samples may represent different points on the concentration profile. 3. The later measured concentration is lower and the missed dose is a plausible explanation; the data do not establish a sustained decline under the prescribed regimen. 4. The multiplicative change is \(\exp(0.05\times2)=\exp(0.10)=1.105\), approximately a 10.5% increase. 5. Examples include dose and dosing interval, actual time since dose, adherence, formulation, laboratory method, or relevant clinical changes.