As contract manufacturers (and researchers) along with generic companies become more important to the Pharma landscape, there are several topics, rules, and technologies worth noting. With more interest in PAT, QbD, and continuous manufacturing (CM), these companies will need to become familiar with some guidances and guidelines.

Foremost is 21CFR11, which was originally written to assist in submitting NDAs and ANDAs.  But, as Robert Burns wrote, “The best laid plans of mice and men…” Or, when Ronald Reagan joked, “I’m from the government. I’m here to help you.”

Basically, 21 CFR Part 11 guidelines define what constitutes an electronic record and set the requirements for their creation, maintenance, and archiving. They emphasize the need to ensure that electronic records are accurate, legible, and retrievable throughout their intended retention period. Sounds simple enough and meant to keep NDAs and ANDAs together and simplify submission acceptances. However, no “little” rule is simple. As added provisions state:

1. Controlled, Audit-Ready Records. Digitization of records helps quality and manufacturing departments ensure that documentation is complete, compliant, and readily available. Compliance software protects all electronic records within a secure document repository accessible to authorized users.
2. Reliable Electronic Signatures. Secure documents will provide fields for all the information required for FDA 21 CFR Part 11-compliant signatures, including name, date/time of signing, and meaning of signature.
3. Automated Audit Trail Capture. The FDA expects electronic record users to have a system in place to capture secure, computer-generated, and time stamped audit trails to independently record the date and time of entries and actions involved in creating, modifying, or deleting.
4. User Authentication/Authorization Enforcement. 21 CFR Part 11 compliance software should offer multiple levels of security to ensure the authenticity of each user, document, and electronic signature in the system.
5. Comprehensive Training and Training Records. There should be the creation and deployment of simple or extensive training courses when trainees are required to learn tasks and demonstrate competency.
6. Accelerated Validation. To lighten validation burdens for any company seeking 21 CFR Part 11 compliance, software has been developed that allow companies to dramatically accelerate validation time. The software packages should also ensure that all audit trails go through revalidation with each update.
7. Sustained Data Integrity. A reliable digital system reduces the risk of data loss, tampering, or unauthorized modification. Using software specifically designed for 21 CFR Part 11 compliance, assures the integrity of the data in electronic records.

Some of the unique “side-effects” of Part 11 can be time-consuming, as well and not clearly advantageous over the “old-fashioned” way of storing information. Indeed, some parts ensure that IT staff need to be added for compliance. The one that affects lab/QC/QA staff the most is the need to store digital (as in “raw”) data for up to 20 years. In recent years, the storage of the terabytes of raw data has been alleviated by the fantastic decrease in price of digital storage devices concurrent with the increase in storage capacity (and shrinkage in size).

What has changed is what is required to be stored. In the “good old days,” we stored the hardcopies of both the master manufacturing formula (MMF) and analytical data sheets in a file cabinet, clearly marked for retrieval and examination by an Agency inspector. However, under Part 11, the raw digital data must be stored: spectra, LC/GC chromatograms, etc. It is incumbent, then, for the IT team to assure that every time there is a company-wide change in computer operating systems (e.g., Windows 10 to Windows 11) that all these raw data can be recalled and, using the analytical software, generate the same analytical values as the initial analyses.

This simple requirement could take many, many man hours. If a company has been in business for more than a year, it is likely that they have generated thousands of bundles of data from analyzing their products. Every time a piece of software (both the overall operating system and specialized software from the instrument manufacturers) is updated, the IT department needs to show that older results can be duplicated with the new operating systems, every time! If the company is truly into QbD and continuous manufacturing, the amount of data points expands exponentially. Instead of 10-20 tablet assays, there could be as many as 1 to 5,000,000 units analyzed for every lot made.

Fortunately, the time and waste saved by CM and QbD more than makes up for the added burden of updating and validating the stored raw data. One “loophole” in 21 CFR 11 is that any measurement not used to release the batch does not need to be stored digitally. So, electronic outputs from every monitor, from the weighing mechanisms (CM) to monitors on the blending or drying portions could plausibly be considered “not used for release” and simply have any outputs stored as printouts. The same is true with secondary or auxiliary monitors.

If you are using, for example, a Raman or NIR monitor in a bioprocess to prompt the operator to withdraw a sample for traditional analyses. Since the compendial analyses are used for product release, the spectra form the monitors need not be stored in digital form, excusing it from the requirement to recalculate it with newer software. So, part 11 need not be used as an excuse not to switch to CM.

Profitability and supply chain

The congressional bills passed recently that are intended to lower the costs of drugs to Medicare patients through negotiated and mandatory ceilings on cost have several intended and non-intended consequences.

1.  The cost of insulin has dropped, allowing many diabetics to obtain the needed doses. Also, several proprietary drugs have had their prices reduced to the benefit of patients. Unfortunately, many generic pharma companies are having trouble making a profit using traditional GMP manufacturing methods.

This is causing several multinational companies to shrink their footprint in America, depending on supplies (APIs and excipients) as well as finished products from places such as China and India. After the supply chain disasters during the COVID-19 pandemic, this is a step backwards from recent “on-shoring” efforts in the U.S. The profits from antibiotics have shrunk recently—only one company in the U.S. still makes amoxicillin (and is almost on the verge of closing) and the diminishing number of COVID vaccinations is causing Pfizer to begin layoffs.

2.  One potential consequence of this legislation would be to encourage a shift from GMP to QbD and CM. Yes, there is the initial investment in instrumentation and training, but the reductions in space, HVAC, waste, and time-to-market are worth the investments.

There is a symbiotic relationship between the pharma manufacturers and the instrument manufacturers: as the need for instrumentation grows, the quality of instrumentation improves (more profits available for R&D and sales increase). On the flip side, as the quality of instruments increases (and, due to volume sales, the price decreases), the demand for and use of process monitoring units will increase. This may be what is called “a win-win situation.”

Miscellaneous

The Near-Infrared chapter in the USP has had a re-write and number change. First, its new address. Formerly, it had been given the number or “for information only.” (If we meet, I’ll explain the political reasons for this.) That meant that any NDA or ANDA submitted with NIR as its mode of analysis could be ignored by an Agency reviewer. It is now in USP-NF and must be considered by a reviewer as the primary release method—only 40 years after being introduced to the Pharma industry, lightning speed for the USP!). This may be considered the “good news.” However, there is some bad news, as well.

When was written, it was crafted by industrial and academic practitioners of NIRS (with the input of former FDA scientists). It was based on the best practices in the agriculture, chemical, pharmaceutical, and textile industries. Suggestions were made as to measures of “goodness” since NIR is a predictive method, not an actual assay (e.g., there are no reference standard tablets or beefsteaks).

The fact that the correlation between results of an NIR analysis and a reference chemical analysis was the result of two different methods, each with its own inherent errors. The correlation coefficient was used to show a relationship between the NIR spectrum and measured chemical or physical property of the sample. It was never assumed to be a measure of linearity, until the FDA Guidance and the new USP chapter were released. The USP chapter was rewritten by a UV/Vis expert and the FDA Guidance by a computer programmer.

The two documents contain a few minor “boo-boos” such as substituting transmittance for transmission or reflectance for reflection. Reflection and transmission are methods of measurement; reflectance and transmittance are the calculated values from the spectra generated by those methods. A more egregious error is the calling for a correlation coefficient value of 0.995 or better and a measure of linearity. While this works quite well in UV/Vis (and as a consequence HPLC, since it uses a UV detector), it is not suited for NIR.

For UV and Visible work, one accurately weighs an API and dissolves it in a known volume, producing a known solution. Serial dilutions are made, and the resultant spectra graphed versus the concentration, giving a Beer’s law graph. Even a moderately trained technician should be able to produce a graph with a slope of 0.995 or better. In NIRS work, the values are unknown and must be determined by a compendial method. These resultant values are used with their spectra to generate an equation for predicting new samples. When you consider that 1) NIR spectra seldom have a zero Absorbance when there is no API, 2) the NIR method has its own measuring errors, and 3) the HPLC method, or whichever method is used, has its error, using a correlation coefficient built on them is not a true measure of linearity—there are numerous others, such as a Durbin-Watson equation, that will determine if a method is non-linear.

So, in short, welcome to 2024. Use your best judgement when doing analyses and good luck.