inConversation with Dr Deep Upadhyay
Our guest, Dr. Deep Upadhyay, Chief Quality Officer at Vasudha Pharma Chem Limited, brings over 25 years of experience in global regulated pharmaceutical organisations. He leads quality operations, regulatory compliance, cultural transformation, and key digital initiatives, with expertise spanning development, regulatory submissions, and global product launches. He has built strong quality systems, strengthened risk-based compliance, and driven operational excellence, while also leading enterprise-wide ERP modernisation to enhance data integrity and efficiency.
Quality Control (QC) has long been one of the most trusted and essential functions in the pharmaceutical industry. From its early-20th-century origins in product inspection, QC evolved through Walter A. Shewhart’s statistical quality control into a structured system of testing, verification, and certification. In pharmaceuticals, tragedies like the Sulfanilamide and Thalidomide incidents made QC a regulatory imperative, establishing it as the custodian of product reliability, consistency, and compliance under GMP and modern quality systems.
Today, the industry operates in a far more data- and process-driven environment. With Quality by Design (QbD), digital systems, and advanced analytics, quality must be built into the process rather than tested into the finished product. Yet in many organisations, QC still functions mainly as a checkpoint, confirming quality after the fact through OOS or OOT signals that surface only after manufacturing is complete.
The future of QC lies in transforming from a gatekeeper into a decision engine, detecting failures, predicting trends, and driving proactive decisions across manufacturing and quality systems. Historically a testing function that became a gatekeeper, QC must now emerge as a strategic decision-support system. Better decisions lead to better processes, better medicines, and ultimately safer outcomes for patients.
Question 1: In your experience leading quality functions, where do you see the most critical disconnect between “data generated by QC” and “decisions taken by Quality”? What structural or governance changes are required to ensure QC outputs directly influence batch disposition, rather than being treated as post-facto confirmation?
QC remains one of the most valuable and trusted functions in pharma, providing the scientific foundation for critical quality decisions. The real opportunity today is to unlock the full value of the enormous amount of data QC already generates.
The most critical disconnect lies in how QC data is interpreted and used. In many organisations, it is treated primarily as a compliance checkpoint, pass or fail against predefined specifications. While this gatekeeping role is essential, the deeper value of QC data lies in using it as a source of process insight. For example, a batch may pass impurity specifications, yet trend analysis can reveal gradual impurity build-up, increased variability, or uncommon unknown peaks in chromatographic profiles. Because the batch technically “meets specification,” these signals are often overlooked, and opportunities for early risk detection, proactive intervention, and process improvement are missed. QC then becomes a post-facto confirmation system rather than a forward-looking decision-support function.
Bridging this gap requires both structural and cultural transformation. Batch disposition must move from a pure pass/fail mindset toward a risk-based evaluation model incorporating trend analysis, process capability, stability signals, control charts, and predictive insights. Even in-specification results should trigger assessment if drift or unusual variability is observed.
An integrated review mechanism involving QC, QA, and Manufacturing is equally important. Cross-functional assessment of critical or borderline cases enables holistic decision-making by combining analytical insight, manufacturing context, and quality risk evaluation. Finally, digital analytics platforms must be leveraged not merely for data archival, but for trend monitoring, early warning signals, and actionable insights.
QC must evolve from “data generator” to “insight provider,” while QA evolves from “gatekeeper” to “risk-based decision maker.”
Question 2: Batch release decisions often rely on discrete specifications, even when underlying process trends suggest emerging risk. How can QC evolve to incorporate trend-based or probabilistic decision frameworks without introducing subjectivity or regulatory ambiguity into release decisions?
Batch release decisions must continue to rely on explicitly defined, unambiguous specifications, these provide the regulatory basis for pass/fail decisions and ensure consistency, objectivity, and compliance.
That said, pharmaceutical quality systems have evolved well beyond simple specification-based release. Today, release assessments consider analytical results, process performance, material variability, manufacturing events, regulatory commitments, electronic data integrity, and overall product quality assurance. The industry has already moved from “quality control” toward broader “quality assurance.”
Real manufacturing rarely behaves in black-and-white terms. “Grey-zone” signals, gradual impurity drift, unusual chromatographic patterns, colour variation, flowability changes, or rising process variability, may not breach specifications but still indicate emerging risk. QC results also gain greater meaning when linked with manufacturing parameters; for example, higher tablet compression force may impact dissolution behaviour. Such cause-and-effect relationships should be scientifically established, monitored, and trended.
The defined specification answers: “Did the batch pass?” Process intelligence asks: “Is the process stable, predictable, and under control?”
To avoid subjectivity, such signals must not rely on individual judgement. Mature quality systems should convert them into predefined decision tables, trend alerts, statistical limits, control charts, and rule-based triggers, scientifically justified and governed through formal Quality Risk Management aligned with ICH Q9.
QC can evolve through a layered decision framework:
- Layer 1: Maintain final release decisions strictly against approved specifications to preserve regulatory clarity.
- Layer 2: Build a “process health” assessment engine using trend analysis, variability monitoring, historical behaviour, SPC, and out-of-trend signals as early warnings before OOS events occur.
- Layer 3: Integrate QC data with process understanding and manufacturing parameters so decisions are based on scientific cause-and-effect rather than isolated numbers.
- Layer 4: Separate “batch compliance” from “process confidence.” A batch may meet specifications while still indicating drift requiring proactive investigation.
This aligns with modern frameworks like ICH Q10 and Continued Process Verification (CPV). APQRs and CPV programmes further help identify long-term drifts and recurring patterns not evident during routine release. While such evaluations are resource-intensive in manual systems, digitalisation, advanced analytics, IoT-enabled systems, and eventually digital twins can significantly strengthen trend detection and predictive capability.
The future of QC is not replacing specifications, but augmenting them with scientifically driven process intelligence.
Question 3: In your experience, what builds real confidence in QC data when making critical quality decisions? Beyond meeting specifications, what practical signals, such as consistency, repeatability, or alignment with process behaviour, indicate that the data can truly be trusted?
QC specifications safeguard the quality, safety, and efficacy of pharmaceutical products. Meeting them confirms compliance with predefined acceptance criteria established during development and approved by regulators, providing the formal and legally binding basis for batch release.
Specifications are built on process understanding, quality risk assessments, clinical considerations, and toxicological evaluations. Parameters such as impurities, microbial limits, dissolution, Delivered Dose Uniformity (DDU), Aerodynamic Particle Size Distribution (APSD), and assay directly support patient safety and product efficacy.
However, meeting specifications alone does not automatically build complete confidence. Specifications represent the minimum acceptable standard for release; critical quality decisions require confidence that the data genuinely reflects the true condition of the product and process. Real confidence comes from consistency, repeatability, predictability, and alignment with process behaviour.
Key practical signals include:
- Consistency of results across batches, analysts, instruments, and laboratories.
- Repeatability and low variability in analytical performance over time.
- Alignment between QC data and manufacturing behaviour, e.g., expected relationships between compression force and dissolution, or process conditions and impurity trends.
- Stable trend patterns without unexplained shifts, drifts, or abnormal variability.
- Chromatographic and analytical profiles consistent with historical knowledge, free of unusual unknown peaks or unexpected noise.
- Data integrity assurance through ALCOA++ principles, attributable, contemporaneous, original, accurate, complete, and reliable.
- Correlation between laboratory observations and real manufacturing conditions rather than isolated numerical compliance.
Equally important is the absence of contradictory signals. A batch meeting specification while showing unusual variability, inconsistent trends, unexpected process behaviour, or weak data-integrity controls should reduce confidence, even if the result is technically “passing.” Contamination from foreign particles, process mix-ups, or data-integrity concerns may not always be detected through routine specification testing alone.
Trusted QC data, therefore, is not merely data that passes specifications, it is data that is scientifically reliable, reproducible, process-aligned, and supported by strong quality systems. True confidence comes when the organisation believes the analytical result genuinely represents the actual quality state of the product, process, and patient risk.
Question 4: In established manufacturing environments, QC improvements are often incremental. Can you share examples of specific changes, in workflows, systems, or team practices, that have tangibly improved decision-making or efficiency in QC without adding significant compliance burden?
In mature manufacturing environments, Quality Control improvements often appear incremental because validated processes and optimised systems already provide a stable baseline. QC naturally becomes more of a diagnostic and certifying function. Even so, the overall value generated by QC is transformational, contributing to risk reduction, prevention of quality failures, patient safety, and process reliability. Small improvements in data interpretation, workflow efficiency, or trend detection can create substantial organisational impact.
For example, integrating QC systems such as LIMS with ERP and automated manufacturing platforms has improved visibility of process variability and enabled earlier detection of abnormal trends. Historical analytical data, chromatographic trends, analyst variability, and OOT patterns can now be evaluated through advanced analytics to support predictive OOS prevention, early warning systems, and identification of hidden variability.
A major industry example is the nitrosamine risk assessment evolution. Initially detected through QC testing, it fundamentally shifted the industry from finished-product testing toward deeper understanding of synthesis pathways, impurity formation mechanisms, and molecular risk assessment, demonstrating how QC insights can drive transformational change beyond the laboratory itself.
At the operational level, many impactful improvements come not from expensive technologies but from eliminating non-value-adding activities and improving decision architecture. Practical examples:
- Reducing sample waiting time caused by poor planning and hidden workflow delays.
- Eliminating unnecessary review layers beyond essential data verification and four-eye review principles.
- Removing tests that generate information but do not support any meaningful quality or business decision.
- Simplifying excessive documentation that adds compliance burden without improving control.
Such changes significantly improve analyst productivity, cycle time, instrument utilisation, and decision speed without weakening compliance. They look incremental but the cumulative impact is transformational.
Another important shift is moving from transactional thinking to trend-based quality management. Individual investigations remain essential, but greater value comes from analysing recurring patterns across batches, methods, analysts, or instruments. A single OOT result may trigger an isolated investigation, but long-term trends across batches can reveal gradual process drift, recurring variability, or systemic weaknesses. Similarly, trending lab errors and repeat analyses can identify opportunities to improve training, instrument reliability, and workflow design, reducing investigation workload and duplicate testing.
Many hidden inefficiencies in QC labs arise not from lack of resources, but from reactive systems, excessive caution-driven activities, fragmented planning, and insufficient use of data-driven review mechanisms. The most sustainable improvements come from simplifying workflows, improving scientific interpretation, focusing on trends rather than isolated events, and removing activities that do not contribute meaningfully to product quality or decision-making.
Question 5: As QC begins to play a more active role in decision-making, how should organisations rethink investigation frameworks (e.g., OOS, OOT)? Specifically, how can QC data be leveraged to move from reactive deviation handling to earlier, risk-based intervention without overloading the quality system?
Traditional OOS and OOT systems are triggered only after something crosses a limit. By that stage, the batch is already manufactured, time is lost, and the organisation is reacting rather than preventing.
As QC takes a more active role, the focus must shift from investigating failures to understanding trends and early signals. Not every small variation needs a formal deviation, but unusual patterns, gradual drifts, recurring minor events, or rising variability should trigger scientific review before they become major issues.
The system needs to be risk-based, not event-based. QC data should be trended across batches, methods, materials, analysts, and process parameters to catch hidden changes early. Simple tools, trend charts, predefined alert limits, and periodic cross-functional reviews, provide strong early warning without adding compliance burden.
Organisations should also separate “monitoring” from “formal investigation.” Every signal does not require a deviation, but important trends should lead to timely discussion, process review, or preventive action. The goal is not more investigations but earlier understanding, a mature quality system should identify weak signals before they become deviations, OOS results, or market complaints.
The investigation focus should be to learn and implement, not to close.
Question 6: Looking forward, as capabilities like advanced analytics, real-time monitoring, and digital integration mature, do you see the traditional boundaries between QC, QA, and manufacturing shifting? What operating model changes will be necessary for QC to function as a true decision engine rather than a downstream control function?
Yes, the traditional boundaries are melting. Digitalisation and IoT will reshape how quality decisions are made.
Historically, QC, QA, and manufacturing operated in sequence, manufacturing produced the batch, QC tested it, QA reviewed and released it. This model worked well in a compliance-driven environment where quality was largely verified at the end of the process. With real-time monitoring, advanced analytics, MES, LIMS, ERP connectivity, process automation, and golden-batch analytics, the industry is moving toward a far more interconnected model. Process data, analytical data, equipment performance, environmental monitoring, and quality events can now be viewed together in near real time.
In this environment, QC cannot remain only a downstream testing or certification function. QC data becomes a live diagnostic input into manufacturing and quality decisions, and the role of QC evolves from “testing and reporting” to “process intelligence and decision support.”
Operating model changes required:
- Collaborative and integrated batch disposition processes rather than sequential ones.
- A shift from event-based QMS to continuous monitoring and predictive quality systems.
- Stronger use of data analytics, process understanding, and statistical interpretation.
- Digital systems playing a central role in decision-making.
- Integrated platforms combining LIMS, QMS, MES, and ERP for trend visibility, process correlation, automated alerts, and predictive insights.
- Adoption of digital twins and AI-assisted analytics to strengthen proactive decision-making.
Importantly, this shift must not dilute GMP responsibilities or accountability. It should create a more connected and scientifically informed quality ecosystem where decisions are based on specification compliance supported by data analytics indicating process stability, future risk, and patient impact.
The future operating model will not eliminate human intelligence or the QC, QA, and manufacturing functions, but it will make them far more integrated, data-driven, and collaborative. QC will act as an active decision engine, contributing continuously to process understanding, risk management, and product quality assurance.
Disclaimer: The views and opinions expressed in this editorial are those of the interviewees and are based on their professional experience in pharmaceutical engineering and sterile manufacturing. They do not necessarily reflect the official views, policies, or positions of Hello Pharma, its management, or its affiliates. Hello Pharma does not endorse or take responsibility for any specific technical, commercial, or regulatory interpretations presented in this article. Readers are encouraged to independently evaluate the information shared, review applicable regulatory guidance, and rely on their own experience, expertise, and professional judgment before making decisions related to equipment selection, system design, validation strategy, or regulatory compliance.
