From Development to Commercial Products: Mind the Gaps

Article by Philippe Tschopp | Glatt Pharmaceutical Services
Based on the presentation delivered at IPTS 2026 in Ankara

The route from a promising formulation to a reliable commercial product is often drawn as a straight line. In practice, it is a chain of decisions, translations and learning loops. Costly problems tend to emerge when one discipline hands its assumptions to the next.

Most crises that emerge late in development do not result from a lack of scientific intelligence. They occur because a formulation, material, process, analytical method or supply assumption fails to travel across an interface. A product that performed well at laboratory scale may behave differently in larger equipment. An excipient that meets its monograph can still show different functional behaviour. A process parameter that appeared critical during development may have been only a local indicator of a deeper physical mechanism.

Pharmaceutical development should therefore not be treated as a sequence of isolated specialist tasks. A drug product is a coupled system. The API, excipients, process sequence, equipment geometry, analytical methods and supply chain all shape the final performance. Optimising one element can destabilise another.

Listen to a summary of this topic:

Start with a clear direction

Before selecting a technology, the development team needs a clear direction. The Quality Target Product Profile is part of it, but the direction should also include the patient need, required product performance and commercial boundary conditions. Dose, release profile, stability, patient use, manufacturing capacity, regional supply and cost belong in the same discussion. Together, they define what a viable product must become.

Without that direction, development can produce elegant answers to the wrong problem. A technically sophisticated formulation has limited value if it cannot reach the required dose, tolerate realistic material variability or be supplied reliably in the markets where the product is needed.

Select the technology from the failure mode

Technology selection should start with what needs to change in the product, rather than with the equipment already available. For a solubility or exposure problem, an amorphous solid dispersion, a lipid formulation or nanomilling may be appropriate. Each option addresses the problem through a different mechanism and introduces its own questions about stability, precipitation, dose and downstream processing.

If flow or compactability is the main limitation, direct compression, wet granulation or dry granulation are more relevant options. The choice must account for segregation, lubricant sensitivity, tablet strength and dissolution. For modified release, a matrix, functional coating or multiparticulate system may fit, but release uniformity, coat integrity and food effects still need to be understood.

Every technology solves some problems and creates others. The best choice is the platform whose mechanism addresses the dominant failure mode and leaves a risk profile that the team can control under commercial conditions.

Scale changes the operating environment

Results generated at small scale do not automatically translate into commercial performance because the laboratory offers conditions that production cannot reproduce. Material paths are short. Sampling is frequent. Equipment is accessible. Operators can intervene immediately. At production scale, residence and hold times increase, heat and mass transfer behave differently, segregation becomes a real risk during transfer, and control loops respond on a different timescale. Yield, throughput and cleaning also become decisive.

Scaling up changes the physical operating environment. Energy input, flow patterns, dead zones, the ratio of surface area to volume, spray distribution, evaporation, diffusion, exposure time and sensor location can all shift at once.

Copying setpoints does not reproduce that environment. The impeller speed, spray rate or feed rate used in a laboratory unit may not produce the same material history in commercial equipment. Equivalent shear, temperature, moisture history, residence time or coating conditions often require different settings. The goal is to transfer the mechanism and confirm the product response. Identical numbers on a control panel do not demonstrate equivalence.

Compendial sameness is not functional sameness

An excipient can meet its monograph and still behave differently in a formulation or process. Particle size distribution, surface area, porosity, moisture, sorption, bulk density, flow and deformation behaviour may vary within a compliant material population. These characteristics can influence feeding, mixing, granulation, compression, coating, dissolution and stability.

The excipient name alone does not define the formulation. Function, grade, supplier and acceptable material range are also part of it. Qualification belongs in development and should begin before purchasing decisions are finalised.

Formulation scientists also face a commercial constraint: the technically ideal excipient may not be reliably available. Approved precedent, regional availability, supplier support, change control practices, cost and continuity of supply all influence the choice. These factors are part of the real optimisation problem. A robust commercial choice must perform technically and remain qualifiable and sourceable across the product lifecycle.

The development target should therefore be an acceptable material space rather than a single excipient batch that happened to work once. The material characteristics that determine functionality, their ranges and their interactions with the process must be understood and documented.

Build causal understanding

Lists of critical material attributes and critical process parameters become useful when the links between them are understood. A parameter is critical because it changes a material state. That state change then drives the product response.

Take spray rate during granulation. It influences the balance between wetting and evaporation. This affects granule density and porosity, which can change flow and compactability and ultimately tablet strength or dissolution. The same causal chain can be traced for lubrication time, roller pressure, drying temperature or coating conditions.

A long parameter list without that causal logic remains a collection of correlations. It may describe the equipment used during development, but it becomes less useful when the process moves to a different scale, site or equipment design. Mechanistic thinking makes the knowledge transferable.

The same principle applies to Design of Experiments. Good statistics cannot rescue a weak hypothesis or an irrelevant factor range. A useful design begins with a plausible mechanism, realistic material and scale variation, a method sensitive enough to detect meaningful change, and a clear decision that will follow from the result. Fewer, well targeted experiments often create more useful knowledge than a broad design built around the wrong question.

Integrate predictive modelling into development

Mechanistic understanding provides the basis for prediction. Once the relationships between material attributes, process conditions and product performance are sufficiently understood, they can be used to anticipate what may happen outside the experiments already completed. This is where predictive modelling becomes part of development.

Predictive modelling should not sit beside development as a separate digital activity. Its purpose is to use existing knowledge for the next decision. A model can help identify material risks before laboratory work, test heat transfer, mass transfer or residence time scenarios before scaling up, and compare process signatures before technology transfer. During commercial manufacturing, the same logic can support the detection and investigation of drift.

The type of model should match the decision. An empirical model may be adequate within a clearly defined range. A hybrid model combines physical understanding with data and may remain useful across equipment after calibration. A mechanistic model can explore causal behaviour, but it depends on the quality of its assumptions and parameters. Greater complexity only helps when it improves the decision.

The model must remain inside the learning loop. Development data provide the starting point. Experiments test the prediction. The results update the model and determine the next experiment, scale condition or control strategy. In this way, modelling reduces uncertainty earlier and makes development knowledge more transferable.

This connection becomes stronger when modelling is combined with process analytical technology, soft sensors and material genealogy. Soft sensors estimate quality attributes or process states from signals available in real time. Genealogy connects the finished batch with raw material lots, intermediate states, equipment routes, process conditions and hold times. Together, these elements connect material variability, process history and product performance across the lifecycle.

Once this evidence is connected, technology transfer becomes less dependent on copying equipment settings and more focused on preserving the relevant mechanism and material history.

Make every decision transferable

Technology transfer should not be the first moment when the team asks whether its knowledge can travel. Transferability needs to be a development deliverable at every gate:

• Feasibility: What failure mode is being addressed, and can the proposed material and process route work?
• Selection: Which mechanism supports the chosen platform, and why is it preferable to the alternatives?
• Optimisation: Which ranges and interactions define a robust operating window?
• Scaling up: Which process signature or material history must remain equivalent?
• Validation and lifecycle: How will the process be controlled, monitored and updated when materials, equipment or supply conditions change?

At each gate, the record should capture the assumption, evidence, applicable range, residual risk and next decision. This creates a connected knowledge chain instead of a document handover.

Author: Philippe Tschopp writes about pharmaceutical excipients, formulation and oral drug delivery, with a focus on excipient functionality, practical formulation considerations and lesser-known material properties.

See our next webinar: