The 2008 version of this document did imply, but did not clearly specify, that endotoxin testing should be performed on implanted products. The draft guidance from 2008 stated:
The reviewer should document the testing performed to demonstrate that all blood contacting devices, permanent implants, devices that contact cerebrospinal fluid, and devices labeled pyrogen-free or non-pyrogenic are, in fact, non-pyrogenic.
The 2016 version is unmistakable regarding the expectations. It states:
Pyrogenicity testing is used to help protect patients from the risk of febrile reaction due to either gram-negative bacterial endotoxins or other sources of pyrogens (e.g., material-mediated pyrogens). To address the presence of bacterial endotoxins, devices that fall under the following categories should meet pyrogen limit specifications:Shortly afterward the following statement occurs:
In addition, for devices that should meet pyrogen limit specifications, we recommend the labeling state that the device is non-pyrogenic.
The new version clearly indicates that implants require pyrogenicity testing to reduce the risk of reactions from both bacterial endotoxin and other material-mediated pyrogens. In addition, this is the first time FDA has recommended that implants be labeled as non-pyrogenic based on the fact they should be meeting the limits established for general medical devices in contact with the cardiovascular system.
Although this document sets new regulatory expectations for a new review of 510(k) submissions, it does not change the fact that all implants that claim to be non-pyrogenic should be tested for bacterial endotoxins. In addition, the risk to the end user for all implants should be addressed based on exposure regardless of the label claim. This means devices that do not have a non-pyrogenic label but are intended to have contact with the cardiovascular system, lymphatic system, or cerebrospinal fluid should be evaluated for the presence of endotoxin according to an established sampling plan.
The New BET Requirements
With the release of this document, FDA is requiring sponsors to provide information for new and revised 510(k) submissions. In reviewing these requirements, it is important to understand the difference between pyrogen testing (an animal-based test described in USP <151>) and BET [an in-vitro test using the Limulus amebocyte lysate (LAL) test described in USP <85> and AAMI ST72]. Following are the requirements as outlined in the document:
The sponsor should provide the information outlined below:
- A description of the method used to make the determination that the device meets pyrogen limit specifications [e.g., bacterial endotoxins test, also known as the Limulus amebocyte lysate test]
- A statement confirming that endotoxin testing will be conducted on every batch or if not, information regarding the sampling plan used for in-process testing and/or finished product release, as recommended in the FDA guidance, “Pyrogen and Endotoxins Testing: Questions and Answers”
- Identification of the chosen testing limit
- An explanation supporting the selected endotoxin limit
The following text highlights some of these requirements:
- A description of the method used. According to FDA, material-mediated pyrogens are chemicals that can leach from a medical device and should be evaluated as part of the biocompatibility assessment using the USP <151> Pyrogen Test. Fortunately, pyrogen testing is not being required as a lot release test and does not need to be included as part of a sampling plan unless there is the risk of non-endotoxin pyrogens. Bacterial endotoxin should be evaluated using one of the three techniques outlined in USP <85> Bacterial Endotoxins Test. The three currently recognized test methods are: gel-clot technique, turbidimetric technique, and chromogenic technique.
- A statement that endotoxin testing will be conducted on every batch or a sampling plan. The typical approach for endotoxin testing is to test every lot or batch after sterilization as this will represent the entire manufacturing process. However, testing every lot after sterilization can be impractical when devices are expensive, lot sizes are small, or high demand requires rapid production. Alternatives to batch testing are acceptable and can be justified based on a rational explanation and demonstration that the manufacturing process is in control. A discussion of reduced sampling plan in the context of implant products is included in the Options for Compliance section of this article.
- Identification of the chosen testing limit. According to the FDA document, the endotoxin limit for general medical devices is 20 EU/device unless the device is in contact with cerebrospinal fluid, where the limit is 2.15 EU/device. These limits are consistent with those from USP <161> Medical Devices—Bacterial Endotoxin and Pyrogen Tests.
- Explanation supporting the selected endotoxin limits. The endotoxin limits for devices is based on patient contact, because different exposure routes to a patient can cause different effects depending on the endotoxin quantity. A device in contact with cerebrospinal fluid is required to meet the lower endotoxin limit as it takes less endotoxin to elicit a pyrogenic response when exposed to the intrathecal space. Additionally, although not mentioned in the FDA document, it should be noted that industry requirements1 have added that ophthalmic implants are required to meet an endotoxin limit of 0.2 EU/device due to the risk of toxic anterior segment syndrome.
Regardless of the sampling plan, the focus of the regulations associated with bacterial endotoxin testing is centered on the demonstration that the manufacturing process is well controlled from a microbiological perspective.
The established endotoxin limits for devices are based on pooling a maximum of 10 devices. This does not imply that the selection of 10 devices, regardless of the lot size, represents a statistically significant sample.
Process monitoring of water, raw materials, and critical manufacturing procedures can be used to demonstrate control and support implementation of alternative sampling plans. The most challenging aspect of the new FDA expectation is that alternatives to batch testing are acceptable, but no guidance on acceptable practices is provided. Instead, the FDA guidance document “Pyrogen and Endotoxin Testing: Questions and Answers” is referenced. This document does not shed much light on the subject, however, stating, “Specifically, firms should take into account aspects of the manufacturing design, including consistency of a manufacturing process, impact of in-process hold times, endotoxins removal steps, and finished product endotoxins specifications.” AAMI ST72, which is an FDA-consensus standard, provides more insight into the selection of an alternative sampling plan.
Following are options for compliance with the requirements.
Test Every Batch (Lot Release or Batch Testing)
This approach is generally considered the most appropriate for medical devices where endotoxin testing is indicated, and there is essentially no justification necessary. In testing of every batch, there is minimal risk to the manufacturer, as any endotoxin excursion is noted prior to release of product. This gives the manufacturer the ability to decide whether re-working, discarding, or any other approach is appropriate for the batch.
The disadvantages are that there is a test cost associated with every batch, although the cost is small, and there might be a time delay while awaiting the BET results. In the case of implant manufacturers, perhaps the greatest disadvantage is the high cost of the devices that are tested. Often the cost of the devices, which are lost to testing, is significantly higher than the cost of the test itself.
Batch testing is likely the best approach for devices in contact with cerebrospinal fluid due to the increased sensitivity of that area of the human body to the presence of endotoxin. Batch testing for these products is also beneficial because the safety factor built into the endotoxin limit of 2.15 EU/device is currently unknown. Batch testing is better able to detect a smaller increase in endotoxin for these devices, which might illicit an endotoxin response. These smaller levels of endotoxin may not elicit the same response in devices with other types of bodily contact.
If the product contains water, or can support microbial growth for any other reason, batch testing is the only option. At the time of manufacture, the endotoxin might be under control, but by the time the product is sterilized, the microbes may have grown and the new endotoxin result might exceed the limit. For these types of products, there is usually too much possibility for batch-to-batch variability to allow for anything other than BET of every batch.
Related to products that can support microbial growth, products that are manufactured with inconsistent manufacturing processes will also indicate the need for batch testing. An inconsistent manufacturing process, or one that is not under strict controls, can see large swings in the amount of deposited endotoxin, especially when water is involved in the process. An inconsistent manufacturing process where no water is involved can show more consistent results and might allow for an alternative test approach. It must first be established, however, that there is not water used in the processes of any of the component suppliers. Although the device manufacturer may not use water in its process, it is common to find that one of the component suppliers does use water, and equally common to learn that few controls are in place with that supplier’s processes.
In summary, this approach is best used for low-cost devices or in situations where it is not worth the risk to justify a less-frequent test scheme.
Alternatives to Batch Testing (Skip-Lot Testing or Alternative Sampling Plan)
There is always additional potential risk when adopting a sampling plan, which is different from testing every batch. There are circumstances, however, where the additional risk is acceptable and appropriate. The primary risk is that product batches might be released, which, due to subsequent test data, are called into question from an endotoxin perspective.
Whenever an alternative sampling plan is being considered, an acceptable level of risk must be established by the company and a risk assessment performed to determine an appropriate sampling frequency. The amount of risk brought on by an alternative sampling plan should be reduced as much as possible. The reduction of risk is usually accomplished by an understanding of the entire manufacturing process, process controls, process monitoring, and historical data.
Understanding the Manufacturing Process and Controls
As previously mentioned, it is common to assume that since there are no water-based processes on-site at the manufacturer, there are also no water-based processes anywhere else in the manufacturing process. Anytime tubing or metals are part of a process (to name some of the more common component types), however, there is a likelihood that the products have had contact with water somewhere upstream from the manufacturer. Manufacturers can no longer blindly accept product components from suppliers without any understanding of the microbiological and chemical aspects of the processes. Though not the focus of this article, chemical residuals from raw materials as well as from molding, manufacturing, and cleaning processes can also be problematic to device manufacturers.
If water is involved in a supplier’s process, the process controls must be well understood by the manufacturer, and the supplier’s controls must be robust enough to adopt an alternative sampling plan. Suppliers must also begin to demonstrate more awareness and vigilance over the potential for endotoxin in their processes. It is the experience of the authors that very few suppliers of metallic implants can provide any history of endotoxin test data from their water systems or finished products. Those that can provide this information should be congratulated, and the others need to implement the controls and process monitoring to properly support their customers. This change to the FDA guidance document will increase the need for manufacturers to rely on their suppliers to have these controls and information in place.
Process monitoring should not consist solely of an occasional bioburden test of the water or a finished product. It should be a well thought out approach based on an understanding of microbiology and endotoxin contamination. Endotoxin can be present even though the bioburden test (i.e., the viable microorganisms) shows results of zero colonies. Also, water in the distribution system can be void of endotoxin, but testing from tanks, bins, baths, or sonicators can demonstrate high counts.
A complete endotoxin process monitoring structure must start with an understanding of the water system and an assessment of its sufficiency to maintain low levels of bioburden and endotoxin. It should include an initial assessment of all points where water comes into direct or indirect contact with the component or product. The outcome of the initial assessment should include a sufficient quantity of bioburden and endotoxin data to demonstrate control over the process and allow a determination of which locations in the system or process provide an accurate view of the bioburden and endotoxin condition overall. The initial assessment should also assist in determining the necessary frequency of testing at the selected locations.
Historical data provides a solid understanding of the true control over a process. For example, it is one thing to show control at any given point in time, but it is another to show control over a year of testing. Although historical data cannot be used to supplant process control, it is an effective means of demonstrating that the existing process controls are functional. As more data are accumulated, and as the data continue to demonstrate acceptable results, there is potential for further reduction of the sampling frequency.
Although helpful, it is not necessary that the historical data consist of 100 percent ideal results. If high or failing results are obtained, it is critical to document that an investigation was performed, corrective action was implemented, and subsequent test results demonstrate the effectiveness of the action.
Implementation of an Alternative Sampling Plan
With these aspects of endotoxin control in place, it is usually an easier decision to implement an alternative sampling plan. It is also easier to justify, and significantly lessen, the risk assumed by the manufacturer. AAMI ST67 provides additional guidance on this topic and is a valuable resource.
Note that under an alternative sampling plan, there will usually be an increase in the amount of process control testing that is undertaken, but this can often be traded for a reduction in the amount of finished product testing.
Another item to consider is whether finished product, which is of sufficient quality to be sold, must be tested. Testing coupons, surrogate product, or product that is rejected due to dimensional or cosmetic defects can be used if a rationale is written and if they still represent the entire manufacturing process.
BET Assessment and Rationale for No Routine Testing
There might be circumstances where testing on a routine basis is not performed. These circumstances are usually limited to a few situations. A good example is when endotoxin removal or inactivation is part of the process and when it occurs late enough in the process that there is no reasonable, subsequent possibility for contamination with endotoxin. In this example, it could be justified that due to process controls and demonstration of successful implementation of the removal or inactivation step, there is no need to perform endotoxin testing on a routine basis. However, endotoxin testing should still be performed initially to confirm the implementation of the elimination or reduction step, and then as part of change control and/or on a periodic basis (e.g., yearly). Process monitoring might still be indicated, depending on the results of the initial testing.
The changes in the FDA document represent a shift in concern with respect to endotoxin on implanted products. This shift can often be handled in a reasonable manner and can result in reduced risk to both the manufacturer and the patient. The authors have noticed a consistent request from FDA requesting additional information regarding endotoxin for implanted products. We have also noticed that it is typically not required or expected, for implanted products, that testing of every batch be performed—as long as a well-written rationale is in place that addresses the process controls, process monitors, and rationalizes an appropriate sampling plan. This is another attempt from regulatory bodies to ensure that risk-based approaches are being used to address patient safety. Thus, it is appropriate that a risk-based approach be used to respond to these changes.
1 FDA, Endotoxin Testing Recommendations for Single-Use Intraocular Ophthalmic Devices, August 2015
Martell Winters is a senior scientist and consulting manager at Nelson Laboratories, a microbiology testing lab that provides medical device testing solutions for medtech companies. Quinton Inglet is the section leader for bacterial endotoxin testing at the company.