Data Validation Requirements for Low Level Alpha Spectroscopy

---

Overview

Data validation is the process of proving that analytical results are accurate. This process usually involves several facets, each derived from expert knowledge about the type of samples being measured, the method of preparation, the instrumentation and the data analysis technique. By associating validation data with the unknown results, a laboratory can document the proper processing of the sample(s).

Data validation is not unique to alpha spectroscopy. The concept exists for almost every analytical method, including sample analysis using gamma spectroscopy. However, the nature of alpha particles dictates more complex sample preparation and mounting processes. Long count times required to achieve expected sensitivities necessitates the use of more detection components to achieve the desired throughput. As the analytical procedures increase in complexity, so does the need for a rigorous data validation plan.

Both Canberra's VMS-based Alpha Management Software (AMS) and the PC-based Alpha Analyst include important functions which allow the tracking and reporting of key quality parameters that relate to both instrument operation and the laboratory processes. These functions will be discussed in this technical brief.

Figure 1
Preparation of laboratory sample from unknown matrices, including creation of appropriate control samples.

The Facets of Data Validation

Data validation is focused upon two key areas:

• The method of sample handling
• The proper operation and use of the instrumentation

Laboratories typically analyze samples according to a pre-defined set of protocols. These protocols must include a comprehensive definition of how to prepare a sample and how to process it. The protocols typically differ depending upon the sample matrix and the requirements specified by those who originally provided the samples. Regardless, it is necessary to document that the protocols are sufficient to yield accurate analysis and that the technicians processing the samples are following the correct procedures. This documentation is usually achieved by the parallel processing of known control samples.

On the other hand, instrument operation is independent of sample handling. An instrument must be properly calibrated to yield accurate results. It must also be free of mechanical or electronic failure, and the software parameters which guide analysis must be properly defined.

Finally, it is desirable to extract as much "known" information out of an individual "unknown" sample as possible. Common preparation protocols for low level alpha emitting samples inherently allow some knowledge to be extracted.

Therefore, passing judgement on the accuracy of any given sample usually involves the assimilation of the following data from the appropriate source:

• Is the instrument properly calibrated?
• Is the instrument operational at the time of counting?
• Are the control sample analyses within an acceptable range?
• Are the chemical yields within an acceptable range?

If all of the above are true, then it is possible to infer that the unknown sample was indeed properly handled and analyzed correctly. This is the basis of data validation for isotope specific alpha analysis.

General Concepts

Validation protocols require the management of a large amount of data. It is important that data validation does not become an unusually large burden upon a laboratory. Therefore, an automated method of tracking this data is absolutely necessary.

Figure 2
Control chart representing spectrometer performance over a 3 month period, and documenting a specific failure.

Both AMS and the Alpha Analyst maintain a series of Quality Control files which make up a database of key validation parameters. These standard functions include the automatic distribution of key parameters to the validation database to minimize technician interaction and therefore improve accuracy and efficiency.

Furthermore, two classes of data are maintained: results from the current batch of samples being processed, and historical results. The type of data maintained includes:

• Instrument		Backgrounds
		Instrument Check
		Calibration
• Samples		Reagent Blanks
		Control Samples
		Chemical Yields

The application of this data to the validation process will be described in the following sections.

Instrument Verification

Is the instrument properly calibrated?

The process of calibrating a single spectrometer is usually straight
forward. AMS includes several commands that automate both initial and update calibrations, while the Alpha Analyst offers the same features, but in a less automated fashion. These commands are performed periodically (weekly or at longer intervals) and cover energy, peak width and efficiency calibrations.

Once calibrated, an important question must be asked: Is the calibration acceptable?

The calibration data itself contains the answer. Key parameters such as centroid position, peak width, percent efficiency, and so on, can be examined. Furthermore, these parameters can be examined in reference to the same parameters on a different spectrometer (batch mode) using AMS, or can be referenced to parameters collected previously on an individual spectrometer on either platform.

Through use of statistical tests, calibration parameters can be automatically compared against expected results. Abnormal changes are flagged so the operator is immediately aware of a calibration problem. Control charts and reports can be generated documenting that the instrument was properly calibrated. The reports can become part of the packet that is provided to the owner of any unknown samples that are processed using these new calibration parameters.

Finally, the data is stored away in QA files, providing a repository which can be quickly accessed in the event data is challenged.

The resultant documentation is proof that the instrument was properly calibrated.

Is the instrument operational at the time of counting?

Calibrations are not performed everyday but samples are counted daily. It is therefore necessary to implement some means of verifying basic instrument operation on any given day. Most alpha spectrometers contain a pulser which is very effective in performing a rudimentary instrument check.

In order to effectively utilize the pulser information, an automated means of analyzing the resultant pulser spectra must be implemented. By comparing the pulser peak centroid, FWHM and counts/second parameters against expected values, an operator can gain a high degree of confidence that individual spectrometers are operational. Control charts and reports can be generated providing documentation of this process.

Once instrument operation has been established, sample counting for the day can proceed.

Figure 3
Quality control report documenting the measurement and limit checks for validating tracer peaks in a sample batch.

Managing Control Samples

Are the control sample analyses within an acceptable range?

The processes described above verify and document performance of the instrument, which is important but incomplete when viewed from a broad perspective. Instrument tests can not determine the accuracy of the chemistry or mounting processes which are extremely critical when working with low level alpha emitting samples.

Control samples are used as a comprehensive method of verifying each stage of sample analysis: sample reduction, separation, mounting and analysis. Control samples are created in parallel with the preparation of unknowns, and typically involve spiking an uncontaminated blank with known amounts of the isotopes of interest. The spikes are in addition to the normal introduction of a tracer spike. A laboratory may, for example, generate one control sample for every twenty unknowns.

A special type of a control is a reagent blank. For a blank, only the tracer spike is introduced and no other activity is expected. The processed results should reach the same conclusion. A laboratory may, for example, generate one blank for every twenty unknowns.  

The resultant samples should yield known results. These known results can be entered into QA files, and processed control sample results can be stored accordingly. During the analysis phase, the pre-determined region of interest which bounds known peaks in the spectrum can be viewed to see if there is abnormal peak shift or broadening. If this occurs for the controls, it must also be a problem with the unknowns.

Reports and control charts can then be viewed as part of the quality procedures. The technicians can infer that if the controls are acceptable, then so too are the unknowns associated with those controls. The appropriate documentation can be bundled with results from unknown analyses to produce a complete quality reporting package.

Individual Sample Validation

Are the chemical yields within an acceptable range?

Each individual sample has a small amount of information which can be used as a limited "sanity check". First, most if not all samples require the introduction of a known amount of a tracer isotope. A technician viewing data can use the width and position of the tracer peak to judge validity.

Furthermore, a chemical yield is computed for the sample based upon the ratio of known vs. measured amounts. If the chemical yield is not within an acceptable range, then the operator can be flagged that something in the preparation process went wrong.

As with the other tests, all of this data can be entered into QA files allowing immediate feedback to the operator as to the acceptability of the data. Documentation regarding the range of yields can be included in the report packet for each appropriate tracer isotope.

Reporting

Reports and control charts from instrument and sample validation data are typically used as documentation to corroborate unknown sample analyses. Calibration and instrument check reports/charts are generated at the time of the calibration. Control sample reports can be automatically generated for the batch or batches to which they apply.

The Sample Batch Processing modes allow processing unknown samples at the same time the associated controls and reagent blanks are processed. Therefore, an automated, efficient method of collating all the reports necessary to form a complete unknown analysis report packet, including the control samples that validate the data, is available.

Historical Record Keeping

All of the data mentioned above is maintained by AMS in QA files. The archived data serves two purposes: to act as a baseline or point of reference and to be recalled in case analyses from a certain day or period are questioned.

All data stored in QA files can be easily accessed. Reports, control charts or interactive queries are all available. Statistical review of the data is possible to identify trends or other anomalies.

Conclusion

The process of data validation is a critical element for an overall laboratory quality assurance plan. The ability to provide detailed documentation regarding instrument and laboratory performance at the time of unknown sample analyses increases the credibility of the results.

Data validation must be a comprehensive program which encompasses each step of sample handling and processing. The procedures should be automated to ensure accuracy and allow efficient implementation. Both the VMS-based Alpha Management Software (AMS) and the PC-based Alpha Analyst include functions that augment these procedures by generating the supporting documentation in addition to rigorous unknown analyses.

---

Top of Page