Tools, Technologies and Training for Healthcare Laboratories

Milan 2014 Consensus Draft on Quality Specifications

On November 24th and 25th of 2014, the first EFLM Strategic Conference on 'Defining analytical performance goals 15 years after the Stockholm Conference on Quality Specifications in Laboratory Medicine' was held. On November 24th, at the end of the first day, a DRAFT consensus was handed out (the DRAFT had in fact been written before the meeting was held). However, the conference organizers later requested that this DRAFT not be shared with the public. Here is only a summary of the contents of the DRAFT. The official final Milan Mandate is available, so this document is only useful to comparative purposes now.

 [Editorial note: Originally an actual copy of the 2014 MILAN DRAFT CONSENSUS: WRITTEN BEFORE MEETING appeared on this page. At the request of the conference organizers, however, it was directed that the document be taken down. What follows below is NOT a verbatim copy of the DRAFT consensus. Instead, it is a brief summary written in our own words. To see the Final Milan Mandate, you must visit the EFLM website:]

First EFLM Strategic Conference on 'Defining analytical performance goals 15 years after the Stockholm Conference on Quality Specifications in Laboratory Medicine'

The first paragraph was to congratulate the meeting on its success in achieving a new set of quality specifications and a new arrangement of those specifications.

Consensus statement:

Analytical performance goals

The DRAFT then notes that there will now be THREE different models for measurands.

Model 1. Based on the effect of analytical performance on clinical outcomes

In order to develop quality specifications using outcomes, you must complete one of the following:

a. an Outcome study investigating the impact of analytical performance on clinical outcomes

b. a Simulation study investigating the impact of analytical performance on the probability of clinical outcomes

c. a Survey of clinicians' and/or experts' opinion investigating the impact of analytical performance on medical decisions

These goals will be most realistic since they are based on actual medical decision-making. However, only a few tests have a direct link to medical decisions, so this model is not universally applicable to all tests.

Model 2. Based on components of biological variation of the measurand

This is familiar territory. The goal here is to make sure the "analytical noise" does not drown out the biological signal. This works for tests where the biologic variation is not so small that the analytical specifications for noise are too stringent to be practical. However, in the new DRAFT it was pointed out that there are in fact  significant limitations to this approach including the relevance and validity of the biological variation data.

Model 3. Based on state of the art

This is least desirable model, but it based on the realistic performance "as is" of the marketplace. If the best laboratories can only achieve a certain quality, but cannot achieve the quality demanded by models 1 and 2, then the world will have to accept the current performance (for now). While technology is improved and (presumably) manufacturers develop better assays, the laboratory should set its specifications using this third model. 

Explanatory notes

Obviously, the models are different and have distinct principles. It is preferred to use models 1 and 2 where possible. 

Also note that under this scheme, some tests will have multiple goals. For instance, if glucose is used differently in critical care than in self-monitoring and diagnosis, there could be multiple quality specifications that come from outcomes studies (model 1), and those specifications may not agree with a quality specification that comes from biologic variation (model 2). Labs will need to accept this fact: multiple specifications may be developed for the same test. Laboratories will need to choose which model and goal to use. Indeed, manufacturers and EQA groups will need to choose which criteria to use as they build new assays and assess laboratory performance.

Performance goals for the pre- and post-analytical phases

Finally, it was noted that it is high time to include specifications for preanalytical and post-analytical processes.

Conference co-chairs: Mauro Panteghini, Sverre Sandberg

Members of the Scientific Committee: Callum Fraser, Andrea Rita Horvath, Rob Jansen, Graham Jones, Wytze Ooserhuis, Per Hyltoft Petersen, Heinz Schimmel, Ken Sikaris

[Editorial note: Remember, this is only a description of a preliminary DRAFT consensus document, written by the organizers. The DRAFT that was distibuted to participants was not supposed to be shared with the public. The writing of the DRAFT was not yet influenced by any of the actual meeting participants.]