Pills have been used to deliver a measured medicinal dose for over three millennia. When the British artist, inventor, and watchmaker William Brockedon produced a compression machine for graphite to produce a better pencil lead, his invention was noticed by a savvy pharmaceutical company who hired him to retool the instrument to produce tablets. Brockedon went on to receive the first-ever patent for such a device, in 1843, for “Shaping Pills, Lozenges and Black Lead by Pressure in Dies.” His invention permitted the efficient mass production of commercially viable dosage forms, and kickstarted the modern-day pharmaceutical industry. It is truly remarkable that for more than 150 years of pill making, Brockedon’s basic methodology has barely changed.
Currently, pharmaceuticals in tablet form are by far the most common formulation of prescription and over-the-counter drugs worldwide. According to the FDA, 46 percent of approved drugs are manufactured in tablet form, and billions of doses are prescribed and consumed in the U.S. each year. Yet, manufacturing a formulation in a compressed tablet form with consistency and reliability is not simply a matter of taking the powdered drug and applying pressure.
The complexity of pills
Many formulations have multiple active pharmaceutical ingredients (APIs), and each of these has their own granulation, wetting, and solubility profiles. But even with a single API, formulations include excipients (non-active ingredients), such as diluents (an inactive filler to achieve a reasonable final pill size), and disintegrating agents to regulate the tablet’s dissolution time after administration.
Often there are binders, which give adhesiveness to the powder during the granulation phase of the ingredient compounding; lubricants, which reduce friction during the compression and ejection of the freshly minted tablet; and glidants that assist in powder fluidity. Colorants for identification and branding may be added. Critical to many tablet formulations are coatings to ease swallowing; to mask unpleasant tastes; to protect the ingredients from light, moisture, and air; to control the drug’s release site within the gastrointestinal tract; or for aesthetic reasons.
Establish your company as a technology leader. For 50 years, the R&D 100 Awards, widely recognized as the “Oscars of Invention,” have showcased products of technological significance. Learn more.
With such a cornucopia of ingredients, ensuring tablet consistency both within and between batches is a complex task. Manufacturers want, and the U.S. Food and Drug Administration (FDA) insists, that there is minimal variation in quantity and physical properties of the APIs from pill to pill. This is achieved in part by carefully mixing and granulating the compound to achieve a homogeneous dispersion of the various drugs and excipients throughout the batch before the pills are compressed.
Many diagnostic techniques are applied to samples taken from each batch to ensure that the tablet weight, amount of drug, dissolution properties, hardness, and coating properties are within the manufacturer’s tolerance for that product. Tablets are checked for impurities which may have found their way into a batch.
These quality control (QC) checks are usually done manually or semi-manually by human operator visually inspecting samples and performing an array of measurements upon the sample. If the batch passes the QC tests, the pills are then packaged.
Machine vision techniques and tools are becoming commonplace during the packaging phase. There are good algorithms to detect and reject individual cracked or partial tablets; to look for container defects; to ensure that containers are filled properly and with the right quantity of tablets; and to detect label defects, loose contaminants, and a host of other macroscopic issues that need to be flagged.
In contrast to this, automated machine vision techniques are not yet widely deployed during the within batch and between batch pill QC testing. One reason is that some of the tests require humans to prepare the sample. For instance, if a tablet needs to be imaged with a scanning electron microscope, then it will certainly be handled by a technician.
However, due to remarkable recent advances in imaging and machine learning, automating some of these procedures is now feasible. For instance, much QC information can come from optically observing the distribution pattern of the various APIs and excipients within a pill.
Typically, the process begins with a system designed to take tablet samples from process line and then split them open using either pressure or a knife, after which the half-tablet is robotically placed on a microscope stage. Using techniques such as photometric stereo, which calculates shape from multiple images taken from the same viewpoint but under different lighting conditions, the newly exposed surface topography can be calculated (see Figure 1). The hills and valleys of the pill’s surface can be mapped instantly and automatically using this technique.
By calculating statistics of the exposed surface such as roughness, average heights and depths of the hills and valleys and their variances and quantity, and their distribution, correlations can be made to more traditional assays. Anomalies in API quantity, dissolution, tablet friability and hardness, and contamination can likely be detected and flagged. Humans currently do much of this work, but there are natural limitations in consistency and accuracy, especially taking into account the effects of fatigue and human error.
Recently, modern machine learning techniques, such as deep neural net learning, have been applied to such image analysis problems with great success. They are currently mature enough to be integrated into the pharmaceutical process line to increase efficiencies and reduce costs for such quality control protocols. The inspections can be done in real-time and automatically red- or green-light a batch—or they can provide automated feedback to the process tools’ instructions to increase the mix time, change the amount of binder, or alter the compression parameters. Like Brockedon’s 1843 invention, the efficiencies gained by deploying such deeply integrated, intelligent automated testing and process control instrumentation could usher in another pharmaceutical manufacturing revolution.
Filed Under: Drug Discovery