12.11.2019 | LOGFILE Feature 42/2019

The alternative to the classic FMEA: FMEA according to Spengler and Juhnke

In 8 steps to self-learning FMEA

In 8 steps to self-learning FMEA

6 minutes reading time | by Felix Tobias Kern


The legislation's demand on the pharmaceutical industry for an integrated risk approach (e.g. according to ICH Q9) implies major challenges in terms of effort and costs.

The FMEA (Failure Mode and Effects Analysis) approach is often used to identify potential weak points and to reduce possible errors, but this entails numerous disadvantages:

  • Complex root cause analysis of possible errors
  • Mostly large team of experts necessary
  • Costly brainstorming sessions in specialist groups to identify risks
  • Incorrect or distorted results as result of over- or under-presence of an area
  • Overestimation of rare special cases (e.g. radioactive contamination by nuclear accidents (Fukushima))
  • Short shelf life due to constantly optimising and changing processes

The FMEA according to Spengler and Juhnke offers an alternative to classic FMEAs:

  • Focus on error analysis without having to know the causes
  • Only the causes of unacceptable, real risks are analysed.
  • Only small team necessary: Savings and fewer resources
  • Self-learning system through continuous updating from the QA system and analysis of actual data
  • Use of common spreadsheet systems possible

The modular approach consists of 8 clearly defined steps:

  1. In the first step, the basic requirements are checked, such as the existence of at least one prototype of the product and at least the definition of the process.
  2. As a first document, a list of process hazards, i.e. documented product defects that can be caused by the process and endanger the patient, is compiled in a second step. Sources for this can be e.g. error pattern catalogs and deviations. For each hazard, a severity rank of 1 - 10 is defined by risk management, i.e. how the manufacturing process endangers the patient (assessment of the extent of damage). In addition, the probability of occurrence (1 - 10) is determined for each hazard. The assessment of both influencing variables can be based on DIN EN ISO 13485 and, if possible, should be determined from actual data, e.g. from analogy to similar products and plants, as well as from the results of internal product controls and the number of deviations that have occurred.
  3. As a second document, a process step list is required, i.e. a list of the entire process in the corresponding substeps from the point of view of the product.
  4. A systematic cross matrix is created from both documents in a 4th step by combining each process step with each process hazard.
  5. Any nonsensical combinations can be filtered out in a pre-evaluation in step 5 with appropriate justifications. Within the framework of this pre-evaluation, the probability of detection of the process hazard is also assessed at the corresponding process step. This can also be based on DIN EN ISO 13485, as well as on automatic and manual control steps for error detection. At the end of the pre-evaluation, the risk priority number (RPN) is calculated from the product of severity, probability of occurrence and detectability for each hazard process step combination.
  6. In the final main evaluation, 3 areas for the RPNs will be identified within a small team of experts: A green area for which no measures are necessary, a yellow area for which possible measures have to be discussed in the expert team in order to reduce the risk and a red area for which risk minimising measures are necessary.
  7. If risk minimising measures are necessary after the main evaluation, these must be defined and implemented in a 7th step. This can be for example a change of the process design or a check by further validations. The risks are then reassessed.
  8. Since this FMEA variant is a living, self-learning system, regular checks are necessary in step 8. For example, the hazard list is checked for up-to-dateness. The assumptions, in particular the assumed event frequency, are to be verified by real data (e.g. annual reports).

In this way, you get a self-learning FMEA that focuses on error-risk analysis and bypasses the errors of classic FMEAs. An example with a combination of 3 hazard process steps for a capsule filling process can be seen in Figure 1. You can read comprehensive details about the alternative method in the publication by Spengler and Juhnke [1].

Figure 1   Example FMEA according to Spengler and Juhnke


[1] Jan-Peter Spengler, Hanno Juhnke: Process FMEA, Pharm. Ind. 77, No. 6, 839 - 843 (2015)



Our recommendations
GMP Compliance Adviser