Automation, Digitization and use of new technologies in Pharmaceutical operations.

 

- Ganadhish Kamat

In the current times Pharmaceutical industry is facing lot of challenges. These include –

• Pricing pressure due to various initiatives by governments to control healthcare cost in the respective countries & consolidation of buyers in US.

• Large number of generic approvals even for complex generics on day 1 in US thereby not giving huge dividends for being first to market the generic version.

• High cost of development, high investments & long gestation period for complex generics.

• Increased regulatory expectations & scrutiny.

• Increasing costs due to rise cost of materials, transportation, energy, wages etc.

Only way the industry can overcome these issues and come out winner is by adopting philosophy of continuous improvement. Continuous improvement will help in improving efficiency, productivity, yields, resource utilization etc and help driving down the cost. It will also help in making the processes robust there by reducing the failure rates, improving compliance, and reducing COPQ (Cost of poor quality). One of the important levers to drive continuous improvement is adopting new age technologies including digital.

We have seen how companies like Google, Amazon, Uber have become successful by adopting  digital technologies in last decade. During same time we have also seen how erstwhile successful companies like Kodak, Xerox or Blockbuster etc declined because of failure to adopt new technologies.

Pharma companies today are at different levels of digitization and automation. At the base there are business process automations like ERP, LIMS, MES etc. Then there are linkages of these base systems with manufacturing and testing equipment for recipe based process controls and data acquisitions. Mechanizations and robotics are used for eliminating human efforts and improving reliability and consistency of the process. Once all the base automations and data acquisitions are in place, then comes the data analytics for continuous improvement.

Most pharmaceutical manufacturing equipment today are capable to run with recipes and have data acquisition systems. They capture lot more data than those controlled through batch records as Critical Process Parameters (CPPs). Testing instruments in the QC labs also captures all the test data including the data of the starting materials, in-process materials and finished products apart from the data of test parameters. Having so much data and doing nothing about it is like sitting on gold mine without starting to explore. Analyzing all the data and establishing meaningful information for the purpose of continuous improvement is humanly impossible without new technologies like Data science, Artificial Intelligence (AI) and Machine learning (ML). Thankfully today these tools are available which can help us in analysis of huge data, including textual data (with the help of LLM). These tools can provide meaningful insight from analysis of huge data and help us in identifying improvement opportunities, improve speed and accuracy of decisions and also help in automating certain complex processes so far were only possible to be done by humans such as speaking, writing, visual comparison etc. 

Driving continuous improvement using data -

One simple way of using data is creating Analytical dashboards which can give useful information without having to go through piles of data scattered over various systems. Analytical dashboards can be designed to provide meaningful information about Key Performance indicators, health of the processes and quality systems using all the available data across various systems mentioned earlier. These dashboards can help managers in daily management, to take necessary course corrections when required. The dashboards can also help in identifying improvement opportunities.

Data analytics tools can be used to perform multivariate analysis and establish relationships between, Material Attributes, Process parameters and Critical Quality Attributes (CQAs) which were not established as critical during product and process development. These tools can be designed to identify the golden tunnel to run the processes to give best performance with respect to Quality attributes as well as yields. The AI tools can also be deployed to facilitate the operator in machine setting for each batch, based on the analysis of material attributes of the input materials and quality attributes of the intermediates thereby significantly reducing setup time and reducing wastages by setting the machines first time right.  Best part of these tools is they become more intelligent as more data becomes available.

Lot of information gathered from manufacturing operations like that mentioned earlier can be leveraged by R&D for optimal process development for new products. Other information available in R&D related to past products including drug- excipient interactions, degradation pathways, structure activity relationships, stability data, bioavailability data, dissolution profiles etc can be mined using AI tools to get first time right development there by reducing the development cycle time. This can also help in reducing the resources required for development by minimizing the experimental work.

Investigation of non-conformities is another very important area in the Pharmaceuticals which often take long time and huge resources due to difficulty in gathering relevant data from the batch records and its analysis. Also sometimes analysis of the data available in the batch records doe not lead to identification of the root cause because many time the failures are due to multivariate effect and sometimes even those parameters have impact which have not been earlier identified as CMAs or CPPs. Having huge amount of data of material attributes, process parameters, machine parameters, environmental parameters etc and ability to do meaningful analysis of such data in short time can help in establishing correlations which might have not been understood in the past. Of course such analysis is not possible without data analytics tools or AI tools. AI tools can also help in trending of incidences or failures and the root causes identified in earlier investigations to provide right directions for new investigations. They can also help in giving early warnings for likely failures based on monitoring of trends of various parameters and correlating them with past failures.

Out of Specification results especially those due to laboratory failures is another pain point in the industry. Data analysis of system suitability parameters and other parameters of chromatography analysis on continual basis can be used to predict the systems or column failures in advance, so that OOS or OOT results which can be caused by these can be avoided by taking proactive measures.

Use of other new technologies in operations -

Process analytical Technology has been talked about for long time but is still not being widely deployed in the industry. There are lot of sensor available today which can give us important information about the process. This information can be used as feed back or feed forward for continuously correcting the process to yield right output. Some commonly used examples are –

- Using compaction force monitoring for adjusting the fill weight in the tablet compression machine

- Use of Outlet temperature of FBD for termination of drying process

- Use of NIR for endpoint determination of blending & eliminating blend testing. 

- Use of current drawn by motor to determine end point of granulation

- Use of tablet bed temperature to adjust the spray rate of coating solution.

- Monitoring of dissolved oxygen and other gases for understanding and controlling fermentation process

Many IoT sensors are now available for continuously monitoring parameters like temperature, pressures vibrations etc. This data can be analyzed using AI tools for predictive maintenance of the equipment there by minimizing breakdowns and improving OEE. Controls of transportation conditions is important to ensure the quality of the medicines. IoT can be used to continuously track the shipments of medicines not only to monitor the transportation conditions but also whereabouts of the shipment to facilitate on time and safe delivery.

Rapid analytical techniques like NIR, Raman spectrophotometry, rapid microbiology etc can bring in significant change in the way we control the quality of products. Establishing appropriate on line or at line controls in the process using these technologies can help us to move towards parametric releases there by eliminating need for end product testing and significantly bringing down process cycle times.

Conventional batch manufacturing processes used in Pharmaceutical manufacturing has some disadvantages like fixed batch sizes, large foot print of the facility due to separate unit processes, high material handling, in-process material storage, stagewise analysis before proceeding to next stage etc. All these result in high resource requirements and also high process cycle times. Continuous manufacturing processes have been in use in chemical industry for long time but have not yet entered Pharmaceutical industry on large scale. Adopting continuous manufacturing can significantly improve the efficiency in Pharmaceutical operations both in API manufacturing and drug product manufacturing. Apart from reducing the process cycle times and the foot print of manufacturing facility, it can allow flexibility of batch sizes so that manufacturing can be done to match the demand there by reducing inventories and COPE due to write-offs of unsold stocks. Smaller foot print can help in reducing energy consumption thereby also having positive impact on the environment.

Robotic Process automation can be used for performing repetitive tasks which require following pre-defined rules such analytical data review, batch record review, review of audit trails, transfer of data from one source to other etc. 

New areas for use of RPA, AI, Gen AI, LLM -

The use cases mentioned earlier in this article are based on my practical experience of implementing these in the organizations I was associated with. I feel there are many more areas in pharmaceutical operations where these tools can be deployed to drive significant benefits in terms of improved efficiency, quality and compliance. Some of such ideas I have described below. Some of these may not be currently feasible due to limitation of technology at the moment but the speed with which the technology is developing, companies can start thinking about these solutions to be ahead of the competition.  

1. Facial recognition -

Correctly capturing the identity of the person carrying out activity is crucial in pharmaceuticals. This is achieved by signature in paper documents and using login name and password in digital records. This poses multiple problems such as need for memorizing multiple passwords for different systems, non-compliances due to sharing of passwords, unauthorized access due to long auto logout time, or need for repeated logins due to short auto logout time resulting in loss of productivity. If facial recognition is used for automatic login like in smartphones all these issues can be resolved. Facial recognition can also be used for access controls to critical areas like aseptic areas eliminating potential non-compliances.

2. Text to voice and voice to text conversion -

Operating personnel in manufacturing and QC have to constantly refer to the instructions in batch records, test methods and SOPs to perform their tasks correctly, this often slows down the activity and creates distraction. Sometimes people may rely on their memory to avoid repeatedly reading the instructions which may result in errors. If the verbal instructions are constantly provided to the operators based on the written procedure by using text to voice conversion, such errors can be avoided while improving the speed of execution. 

Operators & analysts are also required to contemporaneously document the activities carried out by them in the form of start & end time, signature & date and capture certain observations. If this can be done by voice command instead of manual documentation, it will significantly improve the efficiency and also truly comply with requirement of contemporaneous documentation. This will be very helpful especially in aseptic areas for documenting interventions. 

3. Visual verification of activities -

GMP requires all major activities to be verified by second individual. This second verification often gets flouted due to absence of second individual at the time of the activity but gets documented as done without actual verification unless there is electronic time stamping. Ensuring availability of second person for verification, results in delays due to waiting for the person to arrive or requires deployment of more resources. If a visual device such as google glass or camera mounted on the operator does this verification and documents it contemporaneously in the batch records, compliance can be achieved with better reliability without deploying additional resources. Such tools can be very useful in activities such as checking and documenting microbiological test results, documenting aseptic area intervention etc. Some of the common errors in the lab which result in OOS results are use of wrong glassware, incorrect volume makeup, incorrect meniscus adjustment during pipetting etc. Visual verification device can detect such errors and warn the analyst to do necessary correction there by avoiding such errors. 

4. Gen AI for writing procedures, reports, creating training modules, training evaluation etc

Pharmaceutical operations involve creation of large number of procedures, protocols, reports etc. I have often noticed that technical people in India lack language skills to prepare error free documents. Significant resources are deployed in creating these documents, reviewing and correcting them multiple times. Gen AI could be solution for this. The technical people can provide all the technical details required for the procedure or protocol and use Gen AI for generating the procedure in write format. All new procedures and revision in procedures requires training of operating personnel. Gen AI could be used for creating training modules based on the procedures, creating assessment tests for evaluating effectiveness of the training and also perform the evaluation of responses from participants. 

Gen AI could be good tool for generating reports such as APQRs, Validation reports, periodic safety update reports of Pharmacovigilance etc. Gen AI could not only fetch the required information from different data sources and compile it in right format but also perform analysis of data to arrive at appropriate conclusions and write narratives for various sections. 

5. Using AI for processing of chromatographic data -

Majority of the testing in the pharmaceuticals is now performed using chromatography. After performing the analysis, the raw data needs to be processed to get the final results. Chromatography software provide tools for processing of data. Using these, processing of chromatograms for tests such as assay, dissolution etc can be easily performed. However in case of tests such as related substances, where there are many closely eluting peaks, processing often requires manual intervention to suppress small peaks due to noise, to ensure proper separation and integration of closely eluting peaks and correctly labeling peaks of known impurities. Since this is done manually, it is time consuming, prone to errors, and prone to intentional manipulations to obtain favorable results. The last one is often major non-compliance issue cited in FDA 483s and warning letters. AI could be very good tool to automate this activity thereby saving lot of time, improving reliability of test results, improving compliance and improving OEE of instruments.   

6. Robotic process automation & AI for Pharmacovigilance -

Pharmacovigilance involve processing of ADRs, data base searches, creating expedited reports,  creation of periodic safety update reports and signal detection. Processing of ADRs involves determining seriousness and expectedness of the event. Since this is done based on the pre-defined rules, it can be easily automated using RPA. AI & Gen AI can be used for rest of the activities. 

Conclusion -

Automation, digitization and new technologies such as AI, ML, RPA, IoT etc can certainly help in improving efficiency as well as quality and compliance in pharmaceutical industry. It must be kept in mind that these technologies are not replacement of basic scientific knowledge but addon to the scientific knowledge. Implementation of these technologies in pharmaceuticals comes with unique challenge. Reliability of any such tool needs to be demonstrated by comprehensive validation. Since many of the ideas discussed above are new, implementation will involve significant amount of development and testing till they can be rolled out for commercial use. This will require significant amount of resources and time. So it is important to select few which will bring maximum benefits to the organization.