Bioburden Lesson Number 1: Microbes Never Take a Break
An in-process product contamination example emphasizes the importance of getting it right early--in this case cleaning the drainpipes between runs. Third in our Cleanroom series.
Recap: A series of clean room investigations dragged on for years because a manufacturing site minimized the importance of individual EM and water bioburden events. The site also failed to connect the recovery of closely related species from In-process bioburden samples. As the root cause went untreated, in-process bioburden counts increased in frequency and severity. This led to a stop in production and misapplied corrective actions.
We can learn from this mistake. For individual events, the site blamed common, generic root causes to close individual event records. A comprehensive analysis of the microorganisms observed could have solidified scope of the problem and returned the environment to a state of control faster.
In part one and part two of this series, we addressed the common understanding that humans are the primary source of cleanroom contamination. Part 1 looking into skin flora organisms found in a cleanroom. Part 2 discussed the blame personnel received for water organisms showing up in water samples.
This third part of the series discusses an in-process product contamination example. The root cause was also misapplied to people. The hits occurred on some of the first lots manufactured at this site, within a month of the previous water investigation conclusion. Microbes never take a break!
PRODUCT BIOBURDEN TESTING
The first couple lots at this site had action level counts of Bacillus cereus. Production couldn’t get past the first manufacturing step without the product being overrun with contamination.
Below is diagram of the manufacturing process flow from raw material receipt to sample collection. Raw material is received frozen in bottles. Operators transfer the bottles from large carts onto an enclosed assembly line in a controlled / non classified room. The assembly line carries the bottles through a wall into an ISO 8 clean room. A robot cuts open the bottles and transfers them into small thawing vessels. As the material thaws, it drains into 1 of 2 large manufacturing tanks. The first bioburden sample is collected from that tank.
The raw material is expensive - between US$800,000 and $1 million per lot. This bioburden issue impacted the last 2 practice runs prior to going live. The root cause wasn’t correctly identified, so the first two live lots were also destroyed.
Based on the raw material’s source, it was reasonable to believe it was sterile. At the very least, it was expected to have an extremely low bioburden count. Bioburden was introduced somewhere between room 1 and room 4, in this diagram.
During a routine run, there is no human interaction with the manufacturing process from rooms 2-4. However, maintenance was often needed on the robots mid-run, so personnel entered that room for these lots. As you can guess, the investigation quickly focused into room 1, as point of entry, and the maintenance activities of room 2.
Environmental swab samples were collected all over rooms 1 and 2. Surprisingly few swabs had significant recovery. Even fewer had Bacillus. Despite this evidence, the investigation progressed as if contamination was introduced by personnel in these rooms. Three separate root cause investigations – one after each contaminated lot – came to an incorrect human error conclusion. How were humans blamed? These were some of the theories:
- Somehow significant counts of Bacillus from an operator’s shoes in room 1 got in contact with the raw material bottles. This Bacillus persisted on the bottles through the assembly line belt rinsing and blow-drying process. The robot transferred these bottles to the thawing vessel, and Bacillus fell in.
- Gaskets around the windows to the conveyor belt were being insufficiently cleaned. Even though no Bacillus was found behind the gaskets, procedures were changed so they had to be removed and sanitized after each run. Constant manipulation of these gaskets made them fit incorrectly leading to leaks, operators were blamed for the leaks, as if they installed them wrong.
- R&D personnel used new, but non-sterile, tubing to collect a small sample from the top of the tank during the second practice run.
Important people on the investigation team believed Bacillus, introduced from these sources, could proliferate so much that we’d get a Too Numerous To Count (TNTC) result from a 1 mL sample. That’s knowing the batch is held at 2*C for less than a day prior to sample collection and testing.
The Pool Skimmer
To be clear on the defense of human operators – they’re not perfect. They make mistakes. An interesting mistake came up during this investigation. Due to issues optimizing the thawing process, chunks of raw material were impacting the tank mixers. One of the operators brought a pool skimmer into the cleanroom to supplement the mixing process. They dipped the skimmer directly into the GMP product tank and mixed product with it for over half an hour!
This mixing occurred during the first practice run with this bioburden issue. The full blame of the event was placed on this red-herring pool skimmer. The skimmer ultimately had no impact. However, because this action stood out so much from the norm, it delayed finding the actual root cause.
Cleaning validation of these tanks was ongoing during the practice runs. In some ways the validation results supported the contamination-by-personnel theory. The bioburden tests for the rinse samples were below all response limits. Most of the samples had 0 recoveries.
However, the few organisms identified from these samples were B. cereus. The contaminating organism was present in the tank even when it was supposed to be in its cleanest state. At the time, analysts couldn’t explain how the contamination grew to such high levels between cleaning and sample collection, so the investigation team kept their focus on personnel practices.
During this investigation some found themselves hiding in Room 3 to really think about the problem. It was a hidden CNC maintenance area. There was a variety of product, water, CIP supply, and drainpipes connecting in all different ways. How was a cleaning rinse sample able to represent all product contact surfaces?
For example, there was a product inlet to the tank in room 4, but the rinse sample for that tank came in through spray balls. That water never would have touched the product inlet!
The engineering team was able to concoct a special rinse test. Cleanings for the thaw and pooling vessels were normally done individually. However, they arranged the rinse water supply and processing valves to mimic the product pathway for a bioburden sample. The tanks were cleaned, and routine rinse samples were collected. They had no bioburden counts. Immediately after, the special water sample was collected B. cereus covered the plate!
Working off this data, the investigation identified a small section of piping that was not cleaned between runs. The tank rinse samples had good results because the rinse water never contacted this pipe section. Once the programming logic was fixed for CIP supply and valve openings, this issue was never seen again.
The investigation team came into this event with the mindset that personnel were major contamination contributors. Despite the months spent analyzing and “correcting” operator practices, the contamination event was not directly from humans.
If you have enjoyed this article check out Jon's entire series on cleanroom investigations.