Rational formulation design through retrospective machine learning methodology: Case study ibuprofen

Abstract

The presented work investigates critical aspects of rational formulation design through machine learning (ML) methodology to identify essential patterns in immediate release ibuprofen oral dosage products influencing its pharmacokinetic profile. Registry data were extracted and standardized into a consistent format using pandas (v1.3.5) in Python 3.9, with special attention to variant nomenclature for identical excipients. Patterns regarding the usage of dissolution-modifying excipients as well as ibuprofen variants were used to investigate their influence on clinical pharmacokinetic profile. Film coated tablets emerged as the most common immediate release dosage form of ibuprofen utilizing ibuprofen acid as the active ingredient and sodium lauryl sulfate as surfactant/wetting agent. Ibuprofen special variants, such as ibuprofen sodium dihydrate, ibuprofen lysine and ibuprofen arginine, offer more rapid drug release and onset with significantly reduced tmax and increased Cmax as well as generally lower bioavailability variance compared to standard immediate release ibuprofen oral dosage forms. The approaches presented in this article will be helpful in better understanding of rational formulation strategies and support regulatory scientific decisions ensuring predictable bioavailability and reproducible clinical responses.

Introduction

While traditional pharmaceutical development relies on trial-and-error and empirical classification, Machine Learning (ML) offers a transformative advantage by identifying complex, non-linear relationships within high-dimensional datasets that escape conventional statistical analysis. (Murray et al., 2023, Wang et al., 2025) The presented work investigates critical aspects of rational formulation design through (ML) methodology to identify essential patterns in immediate release ibuprofen oral dosage products influencing its pharmacokinetic profile. Ibuprofen is available in diverse oral dosage forms (tablets, soft gel capsules, oral suspensions), and the abundance of clinical data available for ibuprofen formulations with distinct pharmacokinetic profiles presents an ideal opportunity to enhance understanding of rational drug formulation design via implementation of pattern-recognition algorithms. Such an approach could be a helpful tool from a regulatory scientific perspective for ensuring predictable bioavailability and reproducible clinical responses, while enhancing understanding of excipient effects not only for ibuprofen but also for other drugs that pose even greater formulation challenges.

Ibuprofen is a poorly soluble weak acid that exhibits characteristics of a BCS Class II compound. The compound demonstrates minimal solubility in gastric conditions but experiences a marked increase in solubility within the small intestinal environment, facilitating rapid absorption in the duodenal and jejunal regions. (Bermejo et al., 2018) While formulation composition can influence ibuprofen’s pharmacokinetic profile, comprehensive investigations exploring specific formulation design patterns affecting these parameters remain limited. In this regard, understanding the role of excipients on pharmacokinetic profiles plays a crucial role and warrants careful consideration in bioequivalence studies.

In general, excipients may affect drug biological action through various mechanisms, including alterations in gastrointestinal transit time, transcellular or paracellular permeability, active transport processes, or pre-systemic drug metabolism. (Martinez et al., 2022) Excipient effects on pharmacokinetics of drugs are well-known for BCS Class III substances, exhibiting high solubility and low permeability. In particular, excipients affecting intestinal transit (e.g., sorbitol, mannitol) or dissolution rate (e.g., lubricants such as magnesium stearate) may feature absorption-modifying properties. (Metry and Polli, 2022) Surfactants represent another category of critical absorption-modifying excipients, demonstrating P-glycoprotein (P-gp) inhibition in cell culture models. (Metry and Polli, 2022) Surfactants can further be used as permeability enhancers for poorly soluble peptide drugs as exemplified by salcaprozate sodium (SNAC) in the FDA approval of oral semaglutide (Rybelsus).

Excipient effects on ibuprofen dissolution have also been examined previously in vitro. Research focusing on lubricant effects in ibuprofen tablets demonstrated that lubricant-ibuprofen combinations can generate eutectic mixtures with reduced melting points. (Roberts et al., 2004) Concentration of superdisintegrant is another factor that can significantly influence dissolution rate of ibuprofen tablets due to percolation effects. (Dvořák et al., 2020) Surfactants generally enhance ibuprofen dissolution rate, but the ultimate impact varies significantly based on surfactant classification, i.e., anionic, cationic, or non-ionic. (Rangel-Yagui et al., 2005) Studies have further shown that solid dispersions of solid macrogol (such as PEG 8000) and ibuprofen increase the latter’s dissolution rate and bioavailability in rat models. (Newa et al., 2008) Furthermore, while pH-adjusting (alkalizing) excipients theoretically should enhance ibuprofen’s intrinsic dissolution rate, the anticipated improvements, unlike those observed with paracetamol, were not confirmed. (Shaw et al., 2005).

Consideration of the molecular properties of the drug also plays an important role in formulation strategy. Based on drug chemical structure, different decision support systems were developed to predict most suitable formulation strategy for a wide range of poorly soluble drug substances. (Zane et al., 2019) The key discriminatory parameters in these decision trees encompass e.g. the logarithm of dose number (log D0), logarithm of distribution coefficient (log D) for octanol-buffer in small intestines range, number of hydrogen bond donors (HBD), hydrogen bond acceptors (HBA), and molecular surface area. (Branchu et al., 2007) In this context, conventional formulation strategies would be defined as materials and processes that typically do not alter drug bioavailability (only the dissolution rate), including milling, micronisation, and incorporation of disintegrating and wetting agents. In contrast, unconventional formulation approaches would aim at increasing the bioavailability and include such strategies as self-emulsifying drug delivery systems (SEDDS), nanoparticles, and amorphous solid dispersions. According to this decision-support model ibuprofen is deemed suitable for conventional formulation approaches.

Recent developments have introduced another classification framework, the Developability Classification System (DCS), covering not only poorly soluble compounds but all BCS classes. (Butler and Dressman, 2010, Rosenberger et al., 2018) The latter system primarily focuses on drug exposure (AUC) rather than Cmax and tmax, parameters typically crucial for bioequivalence studies. For instance, BCS Class II substances with low solubility and high permeability may be categorized as either DCS Class IIa (dissolution-limited) or IIb (solubility-limited). Thus, it is postulated that DCS Class IIa substances would only require dissolution rate enhancement through traditional formulation approaches (e.g., surfactants, salts), while BCS IIb substances would demand alternative strategies (e.g., amorphous solid dispersions, nanoparticles). Ibuprofen, despite its BCS Class II designation, is classified as a DCS Class I substance and considered relatively straightforward to formulate. While systems like the DCS provide a useful baseline, they are often too reductive to account for the synergistic or antagonistic effects of specific excipient combinations on tmax and Cmax. This is where ML provides distinct value: it can integrate complex excipient choices and dosage form design beyond API molecular properties to predict bioequivalence outcomes with higher granularity than qualitative decision trees, relying exclusively on drug molecular descriptors.

Various nominally conventional and unconventional strategies have been reported for ibuprofen immediate release formulations. According to DCS methodology, ibuprofen solubility is not dependent on particles size (critical particle size above 295 μm) (Rosenberger et al., 2018) and, therefore, micronization and nanoparticle formation would not be considered a suitable formulation strategy for this substance. However, it was demonstrated that crystalline ibuprofen (acid) nanoparticles (270 nm) show much faster dissolution than conventional marketed product, releasing 85% of total dose within 2.5 min at pH = 2. (Plakkot et al., 2011) Similarly, amorphous ibuprofen (acid) formulations were shown to release the drug much faster than commercial products and were capable of maintaining substantial supersaturation level in simulated gastric fluid, where ibuprofen solubility is the lowest. (Mantas et al., 2019, Mantas et al., 2020).

Different classification and decision support systems predict that ibuprofen formulation should be straightforward and based on conventional formulation strategies. However, the direct link between using specific excipients, formulation principles and clinical outcomes of ibuprofen has not been clearly outlined, especially regarding bioequivalence. Unlike many pharmaceutical compounds, where bioequivalence assessment primarily relies on statistical analysis of AUC and Cmax, ibuprofen requires additional considerations due to the clinical relevance of absorption rate and onset. The EMA guidelines specify immediate release ibuprofen bioequivalence criteria to be based on AUC0-t, Cmax and tmax in the fasted state, specifically requiring 90% confidence intervals of 80–125% for AUC0-t and Cmax, along with comparable median (less than 20% difference) and range for tmax.

In the modern pharmaceutical landscape, accelerating the transition from drug discovery to clinical entry is a commercial and therapeutic imperative. As the number of new drug candidates is increasing exponentially with implementation of high-throughput screening and computational chemistry approaches, rational formulation design is critical to minimize the risk of failure in clinical trials, improve high attrition rates from drug candidate to approved product, expedite development timelines, and mitigate the risk of advancing suboptimal drug products into costly clinical programs. As there is no clinical data available for new drug candidates, let alone comparison between different formulation approaches, it is critical to comprehensively evaluate the existing clinical data for established drug product to gain better understanding of the influence of excipient choices and dosage form design on pharmacokinetic profile. By utilizing ML to evaluate the extensive clinical data available for established products like ibuprofen, this research provides a potential framework to bypass iterative trial-and-error cycles. The aim of this study is to identify patterns in immediate release ibuprofen oral dosage formulations that critically influence clinical pharmacokinetic profiles using ML principles as a model system for rational drug formulation.

Download the full article as PDF here Rational formulation design through retrospective machine learning methodology

or continue reading here

Dylan Garamani, Erik Sjögren, Albert Mihranyan, Rational formulation design through retrospective machine learning methodology: Case study ibuprofen, International Journal of Pharmaceutics, Volume 693, 2026,
126667, ISSN 0378-5173, https://doi.org/10.1016/j.ijpharm.2026.126667.

Enjoy our new free webinar, registration & information here:

Continuous manufacturing: Excipient characterization and formulation development