Why Follow a Citation to the Measurement?
A review or database can make a parameter easy to find, but its citation does not by itself establish that the cited paper measured that parameter. The reference may point to another review, a paper that used an earlier estimate, or a study whose measurement differs in population, assay, or reported summary. For PBPK work, these differences matter: a value can look plausible and still fail to support the specific input claim being documented.
The task here is a source audit: follow the citation path far enough to identify the report containing the relevant measurement, then check that the measurement supports the claim as stated. This is not a general ranking of evidence sources or a re-extraction of every assay detail. The earlier material on ranking evidence sources, reconstructing assay metadata, assessing measurement comparability, and preserving reported qualifiers provides the relevant background.
“Original” means the report that contains the measurement relevant to the claim, not necessarily the earliest paper ever to mention a number. A paper may be primary for one result and secondary for another. Likewise, a paper’s reference list is evidence about where its authors directed readers, not proof that the reference contains the measurement being sought.
A Reproducible Trace-and-Verify Workflow
Begin with the exact claim, not a broad search for a familiar parameter name. Record the value as presented, its qualifier, units, species or population, and the review or database version and location. Then follow the cited reference and inspect the cited source itself. If it points onward, continue until a report presenting the relevant measurement is located—or until the trail cannot be resolved.
Record the exact value and wording, source title and version or access date, page, table, figure, or database field, and the citation identifier attached to the claim. Keep qualifiers such as “approximately,” ranges, and population labels.
Use the reference list or database record to identify the cited work. Check that the bibliographic details match the work actually opened; similarly titled papers, duplicate records, and incorrect reference links can interrupt a trail.
Determine whether it reports a measurement, summarizes another source, or merely uses a value. Search its text, tables, figure legends, methods, supplements, and references for the parameter and its origin. Follow additional citations where needed.
Check the parameter identity, species or population, matrix or experimental system, measurement or summary being reported, units, and qualifier. The measurement must support the specific claim; a similar number in a different context is not a match.
Save enough bibliographic and location detail for another analyst to repeat the check. Mark what was verified and what was not. If a link or necessary evidence is missing, record the trail as incomplete rather than filling the gap by inference.
The decision can be expressed as a conjunction. Each condition matters: resolving the citation is not sufficient if the paper contains no measurement, and finding a matching-looking value is not sufficient if it concerns another population or assay.
This is an audit rule, not a claim that a verified measurement is free from bias or applicable to every PBPK use. It establishes that the documented source supports the stated claim in the checked context. Judging broader applicability or choosing between conflicting evidence requires additional assessment.
Worked Source Audits
Worked Example: Checking a Rounded Plasma Binding Value
Hypothetical case. A review states that the human plasma unbound fraction, \(f_{u,p}\), is 0.12 and cites reference 8. Reference 8 is a short experimental paper. Determine whether the cited measurement supports the review’s claim.
The review’s claim is recorded as: \(f_{u,p}=0.12\), human plasma, review table 2, reference 8. Opening reference 8 confirms that it reports an equilibrium-dialysis measurement in pooled human plasma. Its table gives 11.8% for the compound, with no “less than” qualifier. The cited paper therefore contains a measurement of the relevant parameter in the stated matrix and population.
The reported percentage as a fraction is \(11.8/100=0.118\). Rounded to two decimal places, \(0.118\) becomes \(0.12\). Thus the numerical claim is consistent with the reported value at the precision used by the review. The source trail is verified for this claim, with the original report’s table and assay context recorded. This check does not establish that the result represents every human population or every experimental condition.
Worked Example: Following a Citation Through an Intermediate Review
Hypothetical case. A database lists human liver microsomal intrinsic clearance as 42 µL/min/mg protein and cites a 2016 review. The review says “reported by Chen et al. (2011),” but its reference list identifies Chen et al. as a review that in turn cites a 2003 experimental paper by Rao et al. Is the measurement traceable?
First, the database record is captured with its field name, value, units, citation, and version. Next, the 2016 review is checked: it summarizes the clearance and directs the reader to Chen et al. Chen et al. is then inspected. It discusses the estimate but attributes the underlying measurement to Rao et al. The 2003 paper is opened and its table reports 41.6 µL/min/mg protein for the relevant compound in human liver microsomes. The study description and table identify the quantity as a measurement, rather than merely repeating a value without source.
At the reported precision, 41.6 rounds to 42 because the units are unchanged and the value is rounded to the nearest whole number. The path is database → 2016 review → Chen et al. (2011) → Rao et al. (2003). The database’s numerical claim is supported by a located measurement, and the intermediate review is documented rather than mistaken for the primary report. The audit does not silently add more decimal places to the database value or infer that this clearance applies to other experimental systems.
Worked Example: Recording an Incomplete Trail
Hypothetical case. A database lists a tissue-to-plasma partition coefficient of 0.73 for a compound and cites “Miller, 2009.” The database provides no page or table. A search of the available 2009 paper finds the compound name and a discussion of distribution, but no measurement table, methods for this coefficient, or citation attached to the value. The paper’s bibliography includes another distribution review, but that review cannot be accessed.
The audit can verify only that the database makes the claim, that it links to the 2009 paper, and that the 2009 paper discusses distribution. It cannot verify that the paper measured or reported 0.73, that the inaccessible review contains that value, or that the value corresponds to a particular measurement context. The trail therefore stops before the measurement is located.
The appropriate record is not “verified from Miller, 2009.” It is “unverified: database cites Miller (2009); no matching measurement located in the accessible paper; that paper points to an unavailable secondary source; primary measurement not identified.” The value may be retained as a provisional database claim if the modeling workflow allows it, but its provenance must not be described as verified. No missing value, table, or citation should be invented to complete the chain.
What to Record and How to Diagnose a Broken Trail
A useful audit record makes the chain reproducible without requiring the next analyst to repeat the same search from scratch. Include both the claim and the evidence supporting the audit decision. A compact record can contain:
- The parameter name and exact secondary-source claim, including units and qualifiers.
- The secondary source’s identifier, version or access date, and precise location of the claim.
- Each citation link followed, in order, with enough bibliographic detail to identify the source and its role.
- The primary report’s location for the relevant measurement, such as a table, figure, page, or supplement, and the reported value and context.
- An explicit status and reason: verified, partly traceable, or unverified because a specific link or measurement could not be checked.
Common diagnostic signs of a source-trail problem include a secondary source citing only a review, a cited paper that uses rather than measures the value, a mismatch between the parameter named in the text and the table entry, a value appearing only in a later source with no traceable attribution, or a reference that cannot be matched to the cited work. Check supplementary files and the reference list before concluding that a measurement is absent. If access remains unavailable, distinguish “not found in the material inspected” from “does not exist.”
A citation chain can also resolve successfully while the claim fails its match check. For example, a paper may report the parameter for a different species, assay system, or statistic. In that case, report the trail as resolved but the claim as unsupported in its stated context. Do not call it verified merely because the paper contains the same compound name or a numerically similar value.
Common Pitfalls and Alternative Routes
| Approach or mistake | Why it is insufficient | Better practice |
|---|---|---|
| Stop at the first cited review | The cited work may repeat a claim rather than contain the measurement. | Inspect the cited work’s attribution and follow the trail to a report presenting the measurement. |
| Search for a matching number and treat it as confirmation | The number may refer to a different parameter, population, unit, or experiment. | Match the measurement context and reported quantity as well as the numerical value. |
| Use a direct primary-paper search instead of following the citation | A search may find a plausible paper but not establish that it supports the particular secondary claim. | Use search as a discovery aid, then verify the link and measurement in the source itself. |
| Call an inaccessible or missing link “the primary study” | An assumed source cannot substantiate a claim. | Describe exactly where the accessible trail ends and mark the measurement unverified. |
Following citations and searching independently are complementary, not competing, methods. Citation tracing preserves the path by which the secondary source attributes the claim. Independent searches can reveal a primary report missed by a broken citation, but a search result alone does not prove that it is the source intended or that it supports the claim. If an independently located paper appears relevant, document how the match was established and keep the original citation problem visible.
The main limitation is that an audit establishes traceability and support, not scientific validity in every use. The primary report may have methodological limitations, restricted applicability, or a result that conflicts with other evidence. A verified trail should therefore not be confused with a final decision that the value is the best input. The earlier material on measurement comparability is relevant when deciding whether an otherwise verified result applies to the intended PBPK context.
Check Your Understanding
Answer each question, then compare with the brief solutions below.
- A database cites a review for a parameter. The review says the value came from an experimental paper, but the experimental paper is not yet checked. What can be marked as verified?
- A primary report gives 0.246, while a review reports 0.25. The parameter, context, and units match, and the review presents two decimal places. Is the numerical claim consistent? Show the rounding.
- A cited paper reports the same compound’s parameter in rat liver, while the secondary source labels the value as human liver. Has the claim been verified? Why or why not?
- A primary report cannot be accessed after the secondary citation has been confirmed. What should the source record say, and what should it avoid saying?
Solutions. (1) Only the database claim, its citation to the review, and the review’s attribution to the experimental paper are verified. The measurement and its support for the claim remain unchecked. (2) Yes, the values are numerically consistent at the stated precision: \(0.246\) rounded to two decimal places is \(0.25\). This does not by itself verify other aspects of the measurement context. (3) No. The citation may resolve to a measurement, but the species does not match the human-liver claim, so that claim is unsupported by the located measurement as stated. (4) Record that the trail reaches the identified primary report but the report could not be inspected; state that the measurement and claim–measurement match remain unverified. Do not describe the value as measured or confirmed from that inaccessible report.