Laboratory Data Package Review: A QA Checklist by Validation Stage

The EDD for SDG 25-1180 lands Tuesday afternoon: 14 monitoring wells, 62 analytes, 868 rows before the QC samples. The quarterly report goes to the client Friday. Before any of that becomes an exceedance table, somebody has to decide whether the package is sound enough to build on — and that decision is what laboratory data package review is for. This walks through the QA checklist an environmental project should run against a lab deliverable, organized by the validation stages EPA actually defines, plus the one adequacy check that is not on EPA’s list and decides compliance anyway.

What Laboratory Data Package Review Actually Decides

Two activities hide under the word “review,” and conflating them is where most small-firm workflows go sideways. Verification asks whether the lab delivered what the contract, the method, and the QAPP required — a completeness and compliance question answered against paperwork. Validation asks whether the results are usable for the decision you intend to make, and it produces qualifiers on individual results. EPA’s QA/G-8 guidance on data verification and validation draws this line explicitly.

The output of a review is not a stamp that says the package passed. It is a set of qualifiers applied to specific results, plus a usability narrative that tells the next reader — a project manager, a regulator, a plaintiff’s expert — what was checked and what was found. A package can be fully compliant and still yield data you cannot use for your purpose, and a package with three holding-time exceedances can still support a defensible report if the affected analytes are qualified and discussed.

How the Environmental QA Checklist for Laboratory Data Package Review Is Organized

EPA’s 2009 guidance for labeling externally validated laboratory analytical data defines five nested stages. Each builds on the one before it:

• Stage 1 — completeness and compliance of sample receipt conditions and reported results (checks 1–9)
• Stage 2A — Stage 1 plus sample-related QC (checks 10–16)
• Stage 2B — Stage 2A plus instrument-related QC (checks 17–23)
• Stage 3 — Stage 2B plus recalculation of results from instrument responses
• Stage 4 — Stage 3 plus review of the actual instrument outputs

The guidance also defines a label convention that almost nobody outside Superfund work uses, and probably should: S2AVM means Stage 2A Validation, Manual; S2AVE is the same stage performed electronically. Writing the label on the deliverable tells a reviewer in one token what was done, instead of leaving them to infer it from the length of the QA appendix.

Stage 1 — Did You Get What You Contracted For (Checks 1–9)

Stage 1 is paperwork, and it is the stage most often skipped because it feels clerical. Run it first, on every package, before anyone opens the EDD in Excel:

• The lab is identified, and every sample the project submitted appears in the package.
• The requested methods were the methods performed, and analysis dates are present.
• Target analyte results carry the lab’s original qualifiers and the qualifier definitions. A J-flag with no definition key is not a reportable result.
• Result units are reported. Do not infer them from the analyte.
• Reporting limits are present for all samples, with results at or below the RL clearly identified.
• Sampling dates and times, lab receipt date and time, and receipt condition — preservation, pH, temperature — are documented.
• Receipt conditions and sample characteristics are compared against the method and the contract.

The failure that costs the most time is a missing sample-specific RL. Labs sometimes report a single method-level reporting limit for a whole batch, which hides the dilutions. If well MW-7 ran at 10× dilution, its RL for every analyte in that fraction is ten times the batch value, and any non-detect in that sample means something different than the same non-detect in an undiluted sample.

Stage 2A — Holding Times and Sample-Related QC (Checks 10–16)

Stage 2A is where a review starts producing qualifiers. Its seven checks cover:

• Handling, preparation, cleanup, and analysis dates.
• Sample-related QC linked to the field samples it applies to — method blanks, surrogate and DMC recoveries, LCS recoveries, duplicates, matrix spike and MSD recoveries, serial dilutions, and post-digestion spikes.
• Whether the requested spike compounds were actually added.
• Holding-time evaluation.
• QC frequency, and comparison of both holding times and sample QC against the method and the applicable data validation guidance.

Two things here reliably get missed. First, holding time is two clocks, not one — the guidance specifies evaluating sampling date to preparation and preparation to analysis separately. A package can meet the total holding time and still have blown the extraction window. Second, QC frequency is a check in its own right: the guidance’s own example is one LCS per twenty samples in a preparation batch, and a package can contain a perfectly acceptable LCS recovery that covers thirty-five samples.

Holding times for Clean Water Act methods come from 40 CFR Part 136 Table II, which is the controlling list for non-potable water — wastewater, surface water, and groundwater. Metals other than mercury and hexavalent chromium run 6 months preserved to pH < 2 with nitric acid; mercury by cold-vapor runs 28 days. Check the table rather than working from memory, because the preservation condition changes the number for several analytes and the Method Update Rules have moved some of them.

Stage 2B and Up — When Instrument QC Is Worth Paying For (Checks 17–23)

Stage 2B adds initial and continuing calibration data, calibration verification and blank recoveries, whether reported samples are bracketed by CCV and CCB standards, method-specific instrument performance checks, and instrument QC frequency — the guidance cites GC-MS tunes every 12 hours as an example. Stages 3 and 4 add recalculation from instrument responses and review of raw instrument output. Both are expensive, and both are usually performed on a defined percentage of results rather than all of them.

Which stage a project needs should be specified in the QAPP before samples go out, not negotiated after the EDD arrives. When the QAPP is silent — common on small commercial due-diligence work — that silence is the finding, and it is worth raising before the sampling event rather than after. Litigation-facing data, sole-basis closure decisions, and anything supporting a no-further-action determination are the situations where firms most often regret having bought Stage 1. The National Functional Guidelines for data review are the criteria most external validators apply once you get past Stage 2A; DoD projects use the EDQW guidelines instead, and the two do not assign qualifiers identically.

The Check That Is Not on EPA’s List: Reporting Limit vs. Applicable Standard

Every Stage 1 and 2A check asks whether the reporting limits are present and identified. None of them asks whether the reporting limits are low enough to answer your question. That check belongs to you, and it fails quietly.

This is a third state, distinct from both compliant and exceeding, and it needs to be visible in the table and named in the narrative. Practically, catch it by joining the RL column to your standards table during screening and flagging any row where RL ≥ standard. Catching it before the samples ship is better still: specify the required detection limits in the lab request, because a reanalysis after the fact is only possible if the sample is still within holding time. The same join logic that produces the flag also feeds the qualifier handling described in substitution methods and censored statistics for non-detects, and it belongs in the same pass of the field-to-report data workflow rather than as a separate manual sweep.

When a Check Fails: Qualify, Reanalyze, Resample, or Reject

A failed check is not automatically a rejected result. The ordering that usually applies:

1. Qualify. Most QC exceedances produce a J (estimated) on affected analytes, with the direction of bias stated when it is known — a low LCS recovery biases results low, which matters differently for a non-detect than for a hit.
2. Reanalyze. Available only if the sample is within holding time and enough volume remains. Worth asking about immediately, because that window closes while the review is in progress.
3. Resample. The expensive option, and the one to escalate to the client before the report is drafted rather than after.
4. Reject. R-flagged data is unusable and should not appear in a compliance table at all, including as a non-detect.

Whichever path you take, the narrative has to say so. A reviewer who sees a J-flag with no explanation cannot tell whether it came from a marginal surrogate recovery or a blown holding time, and those two carry different weight in a groundwater monitoring dataset that will be trended over years. Record the stage label on the deliverable, keep the checklist you ran, and note who ran it. The stage-and-narrative convention exists so that the next person to touch the data — often a regulator, sometimes an expert on the other side — does not have to guess.

This checklist is a starting point for scoping a review, not a substitute for a qualified data validator or for the project-specific criteria in your QAPP. Other posts in the water quality compliance archive cover the downstream steps once the package is accepted.