Case Study of Contaminated WFI in China

What happens when a deviation investigation into microbiological contamination in a water system determines the wrong root cause?  Then, what happens if a corrective action is implemented that results in the contamination immediately coming back worse than it was and now also showing up as well in product samples?  And what does a pharma company do if this takes place at a CMO in China?

[Related: Check out a FREE comparison report for two of the leading CMOs, featuring GMP inspection observation data.]

At the inaugural PDA Europe Aseptic Animal Health conference held virtually in October 2020, Boehringer Ingelheim (BI) Quality Auditor Olivier Chancel presented a compelling case study focused on a contaminated water for injection (WFI) system at a contract manufacturing organization (CMO) producing vaccines in China and the work that was needed to get to the correct root cause and resolve the issue.

Chancel was sent by BI to a CMO in China that was having recurring problems with contamination of its WFI system. His mission was to find the root cause of the issue and put measures in place to prevent the contamination from recurring.

“What I would like to start with are the consequences of treating a direct cause as the root cause in a deviation investigation,” Chancel said.  “While it is beneficial and appropriate to correct the contributing causes to a problem, addressing only these will not solve the root cause—the underlying issue that led to the problem in the first place.  The root cause of the problem remains untouched and will likely cause further problems in the future.”

The incident he described at the PDA conference took place July 2017 at a CMO that was and remains a subsidiary of a German company but is not part of Boehringer Ingelheim.

WFI System Contamination

Chancel began his account of the issue with an overview of the water system.  The WFI distribution loop at the CMO vaccine plant was contaminated with microorganisms.  He discovered no issue with the design of the system, characterizing it as “state-of-the-art.”  The WFI was produced by double reverse osmosis with a distribution loop that recirculates at ambient temperature.

He commented that the microbiological trends were “very, very bad.”  Frequent contaminations with Pseudomonas were identified by biochemical tests.  There were, however, no endotoxin excursions.  The company opened a deviation investigation and concluded that the root cause was poor WFI sampling practice. 

“That kind of root cause is quite common when the company does not want to investigate too deeply,” Chancel commented.

The preventive action was to train QC technicians again on sampling. “I submit that retraining is not considered as an appropriate preventive action,” Chancel commented.  The corrective action was sanitization of the distribution loop with super-heated water.

While it is beneficial and appropriate to correct the contributing causes to a problem, addressing only these will not solve the root cause

The initial investigation by the company concluded that there was no impact to product from the contamination in the WFI system.  It maintained that the WFI is always sterilized before use, either because it is the raw material used to produce clean steam, or by sterilizing filtration.  In addition, it considered that all limulus amebocyte lysate (LAL) test results were less than the detection limit during excursions for gram-negative rods in WFI, and therefore were within appropriate limits according to the U.S., EU, and the Japanese pharmacopeias. 

LAL is an aqueous extract of blood cells from the Atlantic horseshoe crab Limulus polyphemus.  LAL reacts with bacterial endotoxin lipopolysaccharide, which is a membrane component of gram-negative bacteria.  It is widely used for the detection and quantification of bacterial endotoxins.

Chancel posed the question, “Why was no trace of endotoxin detected for excursions with gram-negative rods?”  He explored the answer later in his presentation.

Corrective Action Unsuccessful

After sanitization of the loop—the corrective action—“against all expectations, bioburden excursions came back quicker and bigger than even before the sanitization,” Chancel revealed.

Regarding the identification of the contaminant, a couple of weeks after the beginning of the investigation the contaminant proved to be Sphingomonas paucimobilis, not Pseudomonas. 

Why is a different contaminant showing up now?  At the very beginning of the investigation, the company used a biochemical identification method to identify Pseudomonas.  Two weeks later, it got the results from a contracted lab, which used a PCR (polymerase chain reaction) method to identify the contaminant.

I submit that retraining is not considered as an appropriate preventive action

Sphingomonas and Pseudomonas are remarkably similar, but they are not the same microorganisms. The small differences between Sphingomonas and Pseudomonas cannot be detected by the LAL test but can by the PCR test.  This could explain why endotoxins were not detected.

Subsequently, the firm discovered new bioburden excursions, this time in product samples.

The company had different buildings, and each building had a dedicated WFI distribution loop and a dedicated production system.  After a few weeks, the firm detected Sphingomonas contamination “absolutely everywhere—in all the tanks,” Chancel reported.

“How do you explain these observations, which may appear as a paradox at first look?” he queried.  “At this step, we can conclude that we have not just one root cause of contamination, but maybe several.”

Root Cause Investigated

The direct cause of the contaminated WFI samples was determined by Chancel to be poor cleaning of the WFI distribution loop.  The root cause of contaminated WFI samples was found to be poor preventive maintenance, related to the frequency of change of a charcoal filter.

“I saw the charcoal filter. The capsule, the outer packing of this filter, was clear. I should have been able to see black charcoal inside,” Chancel maintained.  “But the charcoal was not black, was not dark, but was a sort of greenish color because of some contaminants from the water.”

Contaminated product samples were a result of frequent retro-contamination of the flexible drain on the QC lab equipment during the bioburden testing by contaminated water.  Since products from multiple buildings and water for injection were analyzed on the same equipment in the QC lab, the contaminated product results were false positives.  The firm, however, did not know that and believed that all their products were contaminated.

Biofilm is the Culprit

Chancel posed the following question: Why did the bioburden detected after the sanitization come back so quickly and in even greater concentration than before the sanitization?  To explain this, he introduced the concept of a biofilm.

Biofilms occur when bacteria adhere to each other and to surfaces in an aqueous medium.  Many bacterial species secrete a polysaccharide or “slime” that promotes the attachment of additional bacteria and provides some degree of protection to the biofilm colony.  Gram-negative bacteria can readily attach to surfaces due to their morphology and are known to promote biofilm formation.  Biofilms are not common in well-maintained pharmaceutical water systems.  Once formed, however, they are difficult to remove.

He pointed to a phenomenon called “the yo-yo effect” that can occur with a biofilm in a fluid handling system.  Frequently, if a physical or chemical treatment is not appropriate, the biofilm can come back after the treatment quickly and more prevalent.

A biofilm matrix is composed of multiple chemistries.  Heat applied with the appropriate frequency is the best prevention of a biofilm in a water loop.  But once the biofilm has formed, a heat treatment alone will generally only kill the contaminants—it will have an extremely limited impact on the removal of the biofilm matrix.  If the biofilm is not completely removed, residual organic and cellular debris can promote the growth of new ubiquitous bacteria and biofilm colonization.

Why did the bioburden detected after the sanitization come back so quickly?

Pharmaceutical waters, including water for injection, are clean, but not sterile, and can seed the system with new water-borne bacteria.  A vicious cycle can be triggered: the more often the heat treatment, the quicker and the bigger the biofilm.

The removal of the biofilm matrix is a necessary step in any remediation process.  Along with a heat treatment and thorough cleaning of the pipes, it is a necessary part of contamination control.

To remove the matrix requires an alkaline treatment—such as NaOH or KOH—to hydrolyze the biofilm and help clean it from the surfaces.  “The higher the temperature the better, and the higher the concentration of alkali the better,” Chancel advised.  “Following that, a heat treatment, using super-heat if possible, can be effective after biofilm removal because of the heat penetration it provides into the crevices where biofilm could reside.”

Sphingomonas Detection by LAL

Chancel next explored why the CMO was unable to detect Sphingomonas with the LAL test.

Water for injection is considered an essential bulk raw material in sterile preparations, and monitoring its bioburden is mandatory in the pharmaceutical industry.  A bioburden of greater than 10 CFU per 100ml should trigger an investigation with an impact assessment.

“But we need to be careful when Sphingomonas are identified,” Chancel cautioned.  “The LAL test cannot be considered, as it is not relevant for these bacteria. It will result in a false negative.”

Some species of bacteria have been able to adapt to the stringent conditions of WFI, including low nutrient conditions and osmotic pressure. 

“The presence and adaptation of Sphingomonas in purified waters, including water for injection, has already been widely documented,” Chancel said. 

The LAL test is used to detect or quantify endotoxins from gram-negative bacteria based on a reaction with lipopolysaccharides, or LPS, a membrane component of gram-negative bacteria.  Sphingomonas, unlike other gram-negative rods, however, does not contain any LPS in its outer capsule.  Instead, it contains GSL, Glyco sphingo lipids, which are not detected by the LAL test.

We need to be careful when Sphingomonas are identified

Consequently, an endotoxin result of less than 0.25 IU per ml by LAL does not mean that WFI conforms to specifications. It may just mean that the endotoxins are not detected.

Cross-contamination of Test Equipment

Regarding cross-contamination of bioburden test equipment, Chancel explained that QC labs usually consider that filtration funnels are a convenient and classical solution for bioburden testing.  Using them reduces the risk of contamination when technicians transfer membranes on a medium. 

“The risk of a false positive is reduced, but it is not reduced to zero,” he emphasized.

The BI auditor reviewed common practices in QC labs that perform bioburden testing on both water and product (Figure 1).

In addition to the “usual practices,” Chancel suggested that if the filtration system is not used for more than seven days or if it is moved, to sanitize the equipment again.

He explained that the risk of false positives is even higher if the same filtration system is used to assess the bioburden of both products and pharmaceutical waters. 

“It is probably for this reason that some QC labs have made the decision to have different systems—one for pharmaceutical products, another one for clean pharmaceutical waters, and a third one for less clean pharmaceutical waters, for instance, drinkable water or purified water.”

[Related: Download a FREE report comparing FDA GMP enforcement actions for two of the leading CMOs here.]