# ISO 17025 Validity of Results

> Design and validate a quality assurance program per ISO/IEC 17025:2017 Clause 7.7. Covers internal QC techniques (CRMs, replicates, spikes, blanks), proficiency testing and inter-laboratory comparisons, control chart monitoring (Shewhart, CUSUM), out-of-control response procedures, and performance evaluation using z-scores and En numbers. Validates QC program completeness and PT/ILC coverage for all accredited methods.



Tags: ISO 17025, Quality Control, Proficiency Testing, ILC, Clause 7.7, Control Charts


## Example Prompts

- Design an internal QC program for our chemical testing laboratory
- Create a proficiency testing participation plan covering all our accredited methods
- Validate our quality assurance procedures for ISO 17025 compliance
- Set up control chart monitoring procedures with appropriate warning and action limits

URL: https://rakenne.app/skills/iso17025-validity-of-results/index.md

Try this skill: https://rakenne.app/a/?skill=iso17025-validity-of-results



## Overview

A method can be validated, equipment calibrated, and uncertainty estimated -- but none of that matters if the laboratory cannot demonstrate that its results remain valid day after day. ISO/IEC 17025:2017 Clause 7.7 requires a structured program for monitoring the ongoing validity of results through both internal quality control and external quality assurance.

This skill builds a comprehensive quality assurance program with two pillars. The internal QC program defines what checks are performed (CRMs, blanks, spikes, duplicates, inter-analyst comparisons), how often they run, what acceptance criteria apply, and how results are tracked on control charts with warning and action limits. The external quality assurance program maps every accredited method to a proficiency testing (PT) scheme or inter-laboratory comparison (ILC), defines participation frequency, sets performance evaluation thresholds using z-scores and En numbers, and establishes corrective action procedures for unsatisfactory results.

The two extension tools validate each pillar independently: the QC program validator checks for minimum QC types, frequency definitions, control chart procedures, and out-of-control response; the proficiency testing tracker verifies method coverage, provider selection criteria, and corrective action requirements.

### What It Covers

- **Internal QC types** -- certified reference materials (CRMs), method blanks, laboratory fortified blanks (spikes), duplicate analyses, blind samples, inter-analyst comparisons, and method comparisons
- **QC frequency** -- per-batch, per-run, daily, or periodic schedules tailored to method requirements
- **Acceptance criteria** -- control limits for each QC type, traceable to method requirements or statistical derivation
- **Control charts** -- Shewhart X-bar charts, moving range charts, CUSUM for trend detection; warning limits at 2s, action limits at 3s
- **Out-of-control response** -- halt testing, investigate root cause, implement corrective action, re-analyze affected samples
- **Trend analysis** -- monitoring for systematic shifts, drift, or bias that may not trigger action limits
- **Proficiency testing** -- PT scheme participation mapped to every accredited method, with ISO 17043 accredited providers preferred
- **Inter-laboratory comparisons** -- bilateral or multilateral ILCs as alternatives when PT schemes are unavailable
- **Performance evaluation** -- z-score thresholds (satisfactory |z| < 2, questionable 2 < |z| < 3, unsatisfactory |z| > 3), En number for calibration
- **Corrective action for PT** -- root cause investigation and documented corrective action for unsatisfactory results
- **Data review** -- approval workflow for QC data before releasing associated test results

## Example Conversation

> **You:** We need a QC program and PT plan for LabTech Analytical. We run EPA 200.8 (metals by ICP-MS), EPA 300.0 (anions by IC), SM 4500-H+ B (pH), SM 2510 B (conductivity), and EPA 180.1 (turbidity). Our A2LA assessor is coming in 4 months and we basically have nothing formal -- our analysts run QC but there's no written program.
>
> **Agent:** Four months is tight but doable for documentation. Let me understand what your analysts actually do now, so I can formalize it rather than building from scratch. For your ICP-MS metals work:
>
> 1. Do you run a calibration verification standard (CCV) during the run?
> 2. Do you include blanks?
> 3. Do you run any CRMs or matrix spikes?
> 4. How often do you run duplicate analyses?
>
> **You:** Yes -- for EPA 200.8 we run an ICV at the start, CCV every 10 samples, CCB (blank) every 10 samples, one matrix spike and matrix spike duplicate per batch of 20 samples, and a method blank per batch. We have NIST 1643f that we run once a week as a check standard. For the other methods it's less formal -- pH and conductivity just check against a buffer or standard before each batch.
>
> **Agent:** That's actually a solid QC foundation for metals -- EPA 200.8 QC requirements are built into the method. I'll formalize all of that. For pH and conductivity, the buffer/standard check is the verification, but Clause 7.7 expects additional monitoring. I'll add:
>
> - For pH: daily check with 2 buffer solutions (pH 4.00 and 7.00), acceptance +/- 0.05 pH units; weekly duplicate analysis; NIST-traceable buffer as CRM
> - For conductivity: daily check with NIST-traceable KCl standard, acceptance +/- 5%; weekly duplicate
>
> For control charts -- are you plotting any of this data over time?
>
> **You:** Honestly, no. We keep the data in our LIMS and look at it if something seems off, but we don't have formal control charts.
>
> **Agent:** Control charts are a hard requirement for Clause 7.7. I'll set them up for your most critical QC checks:
>
> - **NIST 1643f weekly CRM results** for each element -- Shewhart X-bar chart with centerline = certified value, warning at +/- 2s, action at +/- 3s
> - **Matrix spike recoveries** -- Shewhart chart with centerline at 100% recovery, limits derived from your first 20 data points
> - **Method blank results** -- chart to detect contamination trends
>
> The mean and standard deviation for setting control limits should come from at least 20 data points collected under intermediate precision conditions (different days, different analysts). How much historical data do you have for the NIST 1643f check standard?
>
> **You:** We've been running it weekly for about 8 months, so roughly 35 data points per element. That should be enough, right?
>
> **Agent:** Perfect -- 35 points is more than adequate. You can calculate the mean and SD from those results to set your initial control limits.
>
> Now for proficiency testing. Which PT programs are you currently participating in?
>
> **You:** ERA water PT studies for metals and anions -- we do two rounds per year. We just started USGS standard reference samples for wet chemistry parameters. We haven't done PT for turbidity.
>
> **Agent:** Turbidity is a gap. A2LA will expect PT coverage for every accredited method. Options:
> 1. ERA offers a turbidity PT sample in their WP study -- you may already be enrolled but not running it
> 2. If no commercial PT is available, you could arrange a bilateral ILC with another accredited lab
>
> I'll flag this in the plan with action needed.
>
> I've drafted both documents. Running validators...
>
> `qc_program_validator` -- **PASS**. Found 5 QC types (CRM, spike, duplicate, blank, inter-analyst via NIST 1643f comparison), frequency defined, control charts defined, out-of-control response included.
>
> `proficiency_testing_tracker` -- **REVIEW**:
> - WARNING: No guidance for methods where PT schemes are not available
>
> I'll add a section on alternative approaches: bilateral ILC with a partner lab, or CRM verification studies when no PT scheme exists, with justification documented. Running again after update -- **PASS**.
>
> Both documents saved to `output/`.

## Sample Output Excerpt

The skill produces a quality assurance procedure, a QC program, and a PT/ILC plan. Here is a representative excerpt from the internal QC program.

---

## 3. Internal Quality Control Program -- EPA 200.8 (Metals by ICP-MS)

### 3.1 QC Elements per Analytical Batch

| QC Type | Frequency | Acceptance Criteria | Action if Failed |
|---------|-----------|--------------------|-----------------|
| Initial calibration verification (ICV) | Start of each run | Recovery 90--110% for all elements | Recalibrate and re-run ICV |
| Continuing calibration verification (CCV) | Every 10 samples and end of run | Recovery 90--110% | Re-run CCV; if still failing, recalibrate and re-analyze samples since last passing CCV |
| Calibration blank (CCB) | Every 10 samples | Less than or equal to MDL for each element | Investigate contamination source; re-analyze affected samples |
| Method blank | 1 per batch (max 20 samples) | Less than or equal to MDL for each element | Identify contamination; re-prepare batch if blank is elevated |
| Matrix spike (MS) | 1 per batch (max 20 samples) | Recovery 75--125% (element-specific per EPA 200.8 Table 4) | Evaluate matrix effect; report with flag if recovery outside limits |
| Matrix spike duplicate (MSD) | 1 per batch (max 20 samples) | RPD less than or equal to 20% | Investigate precision issue; re-analyze if RPD exceeds limit |
| Laboratory fortified blank (LFB) | 1 per batch | Recovery 85--115% | Investigate calibration or preparation error |
| CRM verification (NIST 1643f) | Weekly | Within +/- 2s of certified value on control chart | If outside 2s warning: investigate; if outside 3s action: halt and investigate |

### 3.2 Control Chart Procedures

**Shewhart X-bar Chart for NIST 1643f (per element)**

- **Centerline**: Mean of at least 20 historical results (calculated from 8 months of weekly data)
- **Warning limits (+/- 2s)**: Mean +/- 2 * standard deviation of historical results
- **Action limits (+/- 3s)**: Mean +/- 3 * standard deviation of historical results
- **Plotted by**: Analyst who ran the CRM
- **Reviewed by**: Department Lead, weekly

**Westgard Rules Applied:**

| Rule | Trigger | Response |
|------|---------|----------|
| 1-3s | Single result exceeds 3s | Halt analysis; investigate; re-analyze affected batch |
| 2-2s | Two consecutive results exceed 2s (same side) | Investigate systematic shift; increase monitoring |
| R-4s | Range between two consecutive results exceeds 4s | Investigate random error source |
| 10-x | Ten consecutive results on same side of mean | Investigate systematic bias; recalibrate |

### 3.3 Out-of-Control Response Procedure

1. **Halt**: Stop analysis for the affected method immediately
2. **Quarantine**: Flag all results generated since the last passing QC check
3. **Investigate**: Identify root cause (contamination, calibration drift, reagent expiry, equipment malfunction)
4. **Correct**: Implement corrective action and document in CAPA log
5. **Re-analyze**: Re-analyze quarantined samples after corrective action is confirmed effective
6. **Resume**: Resume routine analysis only after passing QC is demonstrated
7. **Record**: Document the entire event in the out-of-control log with root cause, corrective action, and analyst/supervisor signatures

<!-- /excerpt -->

## Extension Tools

### `qc_program_validator`

Validates the internal quality control program for Clause 7.7 completeness.

| Check | What It Validates | Severity |
|-------|-------------------|----------|
| **QC types (min 3)** | At least 3 different QC types used (CRM, spike, duplicate, blank, blind, inter-analyst, method comparison) | INSUFFICIENT if fewer than 3 |
| **QC frequency** | Per-batch, per-run, or periodic schedule defined | MISSING if absent |
| **Acceptance criteria** | Control limits or tolerance defined for QC checks | MISSING if absent |
| **Control charts** | Shewhart, CUSUM, EWMA, or equivalent chart procedures | MISSING if absent |
| **Warning/action limits** | 2s warning and 3s action limits (or equivalent) specified | MISSING if absent |
| **Out-of-control response** | Investigation and corrective action procedure for exceedances | MISSING if absent |
| **Trend analysis** | Approach for detecting systematic shifts or drift | MISSING if absent |
| **Data review** | Review or approval process for QC results | WARNING if absent |

### `proficiency_testing_tracker`

Validates the proficiency testing and inter-laboratory comparison program for completeness.

| Check | What It Validates | Severity |
|-------|-------------------|----------|
| **PT/ILC program** | Proficiency testing or ILC program defined | MISSING if absent |
| **Method coverage** | All accredited methods mapped to PT schemes or ILCs | MISSING if no coverage mapping |
| **Provider criteria** | PT provider selection requirements (ISO 17043 preferred) | MISSING if absent |
| **Frequency** | Participation frequency defined (at least annual per method) | MISSING if absent |
| **Performance evaluation** | z-score or En number thresholds defined | MISSING if absent |
| **Corrective action** | Process for unsatisfactory results (root cause, investigation) | MISSING if absent |
| **Participation records** | Requirements for maintaining PT/ILC history | MISSING if absent |
| **Exclusion justification** | Guidance when PT schemes are unavailable (alternatives documented) | WARNING if absent |

## Getting Started

Start by activating the *ISO 17025 Validity of Results* skill. The agent will ask about your current QC practices and PT participation, then formalize and expand them into a complete Clause 7.7 program.

Have this information ready:
- All test methods in your accreditation scope
- Current QC practices per method (what QC types you run, how often, what limits you use)
- CRMs and check standards you have on hand (NIST SRMs, ERA standards, etc.)
- Historical QC data -- at least 20 data points for setting control chart limits
- Current PT program participation (providers, schemes, frequency, recent results)
- Any methods where PT schemes are not available
- How QC failures are currently handled (investigation, re-analysis, documentation)
- Your LIMS capabilities for QC tracking and control chart generation

If you have already completed the **ISO 17025 Laboratory Profile** and **ISO 17025 Method Validation** skills, the agent will use your method scope and validation performance data to tailor QC acceptance criteria and ensure PT coverage is complete for all methods.



---

Back to [Skill Library](https://rakenne.app/skills/index.md)