Biosensors and Nanotheranostics

1. Introduction

Over the past several decades, nanomedicine has reshaped therapeutic design by enabling targeted drug delivery, limiting systemic toxicity, and improving the pharmacokinetic behaviour of complex pharmaceutical agents (Lu et al., 2016; Ballacchino et al., 2021). Among the many organic and inorganic nanocarriers that have been explored, liposomes, which are spherical vesicles built from synthetic or natural phospholipid bilayers surrounding an aqueous core, remain the most successful and most widely translated platform in the clinic (Bozzuto & Molinari, 2015; Basak et al., 2025). Their layered architecture is the source of a rare versatility: a single vesicle can carry hydrophilic drugs within its inner lumen, hydrophobic compounds within the lipophilic membrane, and amphiphilic moieties at the lipid-water interface (Lu et al., 2016; Kiuchi et al., 2025).

Building on this versatility, the convergence of diagnostic imaging and therapy within one nanoplatform has given rise to the field of theranostics (Lammers et al., 2011; Skupin-Mrugalska et al., 2021). Theranostic liposomes are a sophisticated subset of multifunctional nanomedicines capable of real-time, non-invasive bioimaging, target tracking, and stimuli-responsive therapy (Lu et al., 2016; Ballacchino et al., 2021; Pharmaceutics Review, 2025). By co-encapsulating therapeutic payloads, such as hydrophobic photosensitizers (e.g., zinc phthalocyanine, ZnPc), chemotherapeutics (e.g., doxorubicin, paclitaxel, or curcumin), or nucleic acids, alongside diagnostic probes such as gadolinium-based MRI contrast agents, gold nanoparticles, or fluorescent dyes, these vesicles make image-guided and personalized treatment feasible within a single construct (Skupin-Mrugalska et al., 2021; Al-Ahmady et al., 2018; Ballacchino et al., 2021).

Despite this promise, the path from laboratory bench to commercial product is repeatedly obstructed at the manufacturing stage (Amrani & Tabrizian, 2018; Basak et al., 2025). Traditional macroscale batch techniques, including thin-film lipid hydration, reverse-phase evaporation, ethanol injection, and detergent dialysis, depend on chaotic bulk mixing across macroscopic dimensions (Amrani & Tabrizian, 2018; Lamprou et al., 2021). Because concentration and thermal gradients are unavoidable in such vessels, the resulting vesicle populations tend to be heterogeneous, highly polydisperse, and often multilamellar (Amrani & Tabrizian, 2018; Basak et al., 2025). As a consequence, these workflows require secondary downsizing, typically high-pressure homogenization or repeated membrane extrusion, simply to reach an acceptable particle size (Bozzuto & Molinari, 2015; Kiuchi et al., 2025).

These post-processing steps carry their own penalties. They expose labile active pharmaceutical ingredients to shear-induced degradation, deliver low loading efficiencies for hydrophobic payloads, clog extrusion membranes, risk metal contamination from sonicating probes, and introduce substantial batch-to-batch variability (Khadke et al., 2019; Amrani & Tabrizian, 2018). The difficulty compounds when the goal is a complex theranostic construct carrying several therapeutic and diagnostic payloads at once, because bulk methods are prone to competitive displacement, poor co-encapsulation yields, and inconsistent payload ratios, which together make translation under GMP conditions especially hard (Skupin-Mrugalska et al., 2021).

Microfluidic manufacturing has emerged as a response to precisely these physical and operational limits (Jahn et al., 2004; Skupin-Mrugalska et al., 2021; Amrani & Tabrizian, 2018). Within sub-millimetre channels, typically 5 to 500 micrometres, processing is governed by low Reynolds numbers and laminar flow, so that mixing proceeds through rapid, highly predictable convective and diffusive transport rather than turbulence (Amrani & Tabrizian, 2018; Lamprou et al., 2021). In the canonical hydrodynamic flow focusing configuration, lipids dissolved in a water-miscible organic solvent are focused by flanking aqueous streams (Jahn et al., 2004; Ran et al., 2016). As the solvent diffuses outward and local polarity rises, lipids self-assemble into planar bilayer discs that close spontaneously into uniform small unilamellar vesicles (Al-Ahmady et al., 2018; Ran et al., 2016). This single-step route gives exquisite control over critical quality attributes (CQAs), including mean hydrodynamic diameter (tunable from 20 to 200 nm), a narrow polydispersity index (PDI < 0.2), surface charge, and lamellarity (Kiuchi et al., 2025; Han et al., 2022), and it has been reported to raise encapsulation efficiency for hydrophobic photosensitizers and gold nanoparticles two- to three-fold relative to conventional methods (Skupin-Mrugalska et al., 2021; Al-Ahmady et al., 2018).

What microfluidics gains in control at the bench it has historically struggled to keep at scale. Single-channel chips operate at modest throughputs of microlitres to millilitres per minute, which fall short of clinical and commercial demand (Amrani & Tabrizian, 2018; Han et al., 2022; Liu et al., 2022). Simply widening the channels is self-defeating, since it destroys the laminar regime and degrades monodispersity. Instead, the field has turned to process intensification: parallelization, or numbering-up, of identical channels, 3D-printed multilayer chips, chaotic micromixers such as staggered herringbone structures, and microfluidic vortex focusing (Han et al., 2022; Shan et al., 2021; Kabay et al., 2022; Liu et al., 2022). High-resolution 3D printing has enabled multilayer architectures running at total flow rates up to 474 mL/min (Shan et al., 2021), while vortex-focusing devices can exceed 20 g/h while holding size and PDI within tight bounds (Han et al., 2022). To make such throughput GMP-ready, continuous manufacturing is increasingly coupled with Quality by Design (QbD) and Process Analytical Technology (PAT), so that in-line dynamic light scattering, nanoparticle tracking analysis, and UV-Vis monitoring feed automated control loops that hold batch consistency within FDA and EMA frameworks (Basak et al., 2025; Kabay et al., 2022; Patel et al., 2025).

Against this backdrop, the present review critically examines recent advances, technological paradigms, and regulatory frameworks governing the scalable microfluidic manufacture of theranostic liposomes. Its specific aims are as follows: To evaluate microfluidic mixing dynamics and platform architectures (2D/3D hydrodynamic flow focusing, staggered herringbone mixers, static mixers, and vortex focusing) as determinants of the CQAs of theranostic liposomes, namely size, PDI, surface charge, and lamellarity. To compare payload loading mechanisms and entrapment efficiencies of multi-modal agents, including hydrophobic chemotherapeutics, photosensitizers, nucleic acids, and inorganic imaging nanoparticles, prepared by microfluidics versus conventional bulk methods. To synthesize process-engineering strategies for scalable manufacturing, focusing on parallelization, multilayer 3D-printed devices, high-throughput vortex focusing, and continuous-flow operational design for industrially viable output. To outline GMP-ready process-design frameworks that integrate QbD, PAT-based real-time monitoring, automated feedback, downstream purification, and the regulatory expectations of global agencies (FDA and EMA).

2. Microfluidic Self-Assembly, Scale-Up, and Quality by Design for Theranostic Liposome Manufacturing

This section organizes the current evidence base along the logical arc of a manufacturing problem: first the microscale physics that gives microfluidics its control, then a head-to-head comparison with conventional bulk methods, the specific mechanics of theranostic co-encapsulation, the engineering of scale-up, and finally the quality and process-control architecture required for regulated production.

2.1 The Physics of Microfluidic Self-Assembly and CQA Tuning

Liposomes are self-assembled phospholipid vesicles comprising an aqueous lumen enclosed by one or more bilayers, and their amphiphilic organization permits the simultaneous entrapment of hydrophilic compounds in the core, hydrophobic drugs within the acyl-chain region, and surface-bound ligands at the lipid-water interface (Bozzuto & Molinari, 2015). In a microfluidic device, the event that governs vesicle formation is controlled mixing at the microscale, where low Reynolds numbers (Re < 2300) enforce a laminar regime dominated by viscous rather than inertial forces (Jahn et al., 2004). Within a two- or three-dimensional hydrodynamic flow focusing (HFF) geometry, phospholipids dissolved in a water-miscible organic solvent are introduced through a central channel and squeezed into a thin sheet by flanking aqueous streams (Jahn et al., 2004; Ran et al., 2016; Amrani & Tabrizian, 2018). The full sequence, from focused stream to closed vesicle, is summarized in (Figure 1).

As the miscible solvent diffuses laterally into the surrounding buffer, local polarity climbs sharply, lipid solubility collapses, and amphiphilic molecules aggregate into intermediate planar disc-like bilayer fragments (Al-Ahmady et al., 2018; Ran et al., 2016). These discs grow in a reaction-rate-limited manner until the line energy of their exposed hydrophobic edges exceeds the bending energy needed to curve the membrane, at which point they close into spherical small unilamellar vesicles (SUVs) (Han et al., 2022). The practical consequence is that vesicle diameter is not set by chance but by transport kinetics the operator can dial.

Microfluidic processing therefore offers direct command of the CQAs that matter clinically: mean hydrodynamic diameter, PDI, zeta potential, and lamellarity (Kiuchi et al., 2025; Patel et al., 2025). Diameter can be tuned across roughly 20 to 200 nm with narrow, monodisperse distributions (PDI < 0.2) by adjusting the flow rate ratio (FRR) between aqueous and organic channels and the total flow rate (TFR) (Skupin-Mrugalska et al., 2021; Ran et al., 2016; Amrani & Tabrizian, 2018). Increasing the FRR narrows the focused lipid core, shortens the diffusive path for solvent exchange, and accelerates the solubility transition (Ran et al., 2016; Han et al., 2022); because this limits the growth window of the intermediate discs before closure, the result is smaller vesicles with lower size variance (Han et al., 2022; Kastner et al., 2014). Beyond size, static mixers and controlled flow velocities let researchers regulate lamellarity itself, tuning the number of concentric bilayers through lipid concentration and fluid velocity, a degree of structural control that bulk mixing, which tends to default to multilamellar or large unilamellar vesicles, cannot offer (Kiuchi et al., 2025; Ota et al., 2023).

2.2 Microfluidic versus Conventional Bulk Manufacturing

Figure 1. Microfluidic hydrodynamic flow focusing and the mechanism of vesicle self-assembly. Lipids dissolved in a water-miscible organic solvent enter through a central channel and are hydrodynamically focused into a thin sheet by flanking aqueous buffer streams. As solvent diffuses outward and local polarity rises, lipids aggregate into planar bilayer discs that close into small unilamellar vesicles once edge energy exceeds membrane bending energy. The process runs under laminar flow (Re < 2300) with diffusion-limited mixing, yielding size-tunable (20-200 nm), low-PDI vesicles in a single continuous step.

Figure 2. Architecture of a co-encapsulated theranostic liposome. Schematic of a multifunctional vesicle showing how distinct cargoes partition by physicochemistry: hydrophilic agents (doxorubicin, Gd-DTPA chelates, gold nanorods, iron oxide nanoparticles) reside in the aqueous lumen, hydrophobic agents (zinc phthalocyanine, gold nanoparticles) embed within the lipid bilayer, and PEGylated phospholipids decorate the surface. Single-step microfluidic assembly loads these components simultaneously at defined stoichiometry, enabling image-guided, stimuli-responsive therapy from one construct.

The historical barrier to liposomal translation has been the physics of macroscale batch processing itself (Amrani & Tabrizian, 2018; Lamprou et al., 2021; Ballacchino et al., 2021). Thin-film hydration, reverse-phase evaporation, ethanol injection, and detergent dialysis all rely on chaotic bulk mixing in which concentration, shear, and thermal gradients coexist throughout the vessel (Basak et al., 2025; Lamprou et al., 2021). Bulk hydration consequently yields large, polydisperse, multilamellar populations that must be downsized by repeated homogenization or extrusion (Bozzuto & Molinari, 2015; Skupin-Mrugalska et al., 2021). A point-by-point contrast of the two paradigms, spanning fluid dynamics, size control, lamellarity, payload entrapment, throughput, and GMP readiness, is given in (Table 1).

Those secondary operations are not free. High-pressure extrusion and probe sonication subject delicate ingredients, among them nucleic acids, peptides, and proteins, to shear and localized heating, with attendant degradation or leakage (Khadke et al., 2019; Amrani & Tabrizian, 2018). Ultrasonic tip erosion can seed metallic particulate contaminants, while extrusion membranes clog and cap yield (Basak et al., 2025; Ballacchino et al., 2021). Layered on top is the inter-batch inconsistency that follows from small variations in manual mixing speed, vessel geometry, and cooling rate (Basak et al., 2025).

Continuous-flow microfluidics addresses these problems at their root by consolidating hydration, particle formation, and drug loading into one automated step (Skupin-Mrugalska et al., 2021; Al-Ahmady et al., 2018; Ran et al., 2016). Because self-assembly repeats within identical channel geometries under steady-state conditions, batch-to-batch variation is largely designed out rather than corrected after the fact (Basak et al., 2025; Lamprou et al., 2021). Comparative work reports that microfluidic theranostic liposomes reach median diameters up to 2.5 times smaller (under 50 nm) with markedly lower polydispersity (PDI < 0.15) than thin-film hydration followed by extrusion (Skupin-Mrugalska et al., 2021).

2.3 Co-Encapsulation Mechanics of Theranostic Cargoes

The defining feature of a theranostic liposome is that it carries therapy and diagnosis together, enabling real-time imaging, target tracking, and non-invasive monitoring of treatment from within one vesicle (Lu et al., 2016; Skupin-Mrugalska et al., 2021). The generic architecture, hydrophilic agents in the lumen, hydrophobic agents in the bilayer, and PEGylated or targeting lipids at the surface, is depicted in (Figure 2). Assembling such constructs by bulk methods is notoriously unreliable, since multi-component payloads tend to displace one another or phase-separate, producing low yields, broad size distributions, and erratic drug-to-probe stoichiometry (Skupin-Mrugalska et al., 2021; Al-Ahmady et al., 2018).

Microfluidic continuous assembly improves incorporation of both hydrophilic and hydrophobic agents (Al-Ahmady et al., 2018; Skupin-Mrugalska et al., 2021). For hydrophobic cargoes seated in the bilayer, such as ZnPc photosensitizers for photodynamic therapy or hydrophobic gold nanoparticles, microfluidic mixing delivers a two- to three-fold increase in loading over reverse-phase evaporation or thin-film hydration (Skupin-Mrugalska et al., 2021; Al-Ahmady et al., 2018). In one illustrative case, 2-4 nm gold nanoparticles co-assembled into liposomal membranes achieved a gold-to-lipid weight ratio above 60%, roughly three-fold higher than reverse-phase evaporation, with cryo-TEM and small-angle neutron scattering confirming clean embedding within the hydrophobic domain and only a slight stiffening of the bilayer rather than aggregation or disruption (Al-Ahmady et al., 2018).

The same single-step mixing supports genuine multi-component co-encapsulation (Ran et al., 2016; Fernandes et al., 2023). Investigators have built thermosensitive liposomes that hold doxorubicin in the lumen alongside superparamagnetic iron oxide nanoparticles and gold nanorods (Lv et al., 2021; Ran et al., 2016). Operating near an FRR of 4:1 and a TFR of 25 mL/min, microfluidics produced uniform folate-targeted magnetoliposomes (around 230 nm, PDI < 0.15) with high doxorubicin entrapment and superparamagnetic behaviour (Lv et al., 2021). Near-infrared irradiation of the embedded nanorods drove local heating above the lipid phase-transition temperature and triggered controlled release, with the paired chemo-photothermal action producing synergistic cytotoxicity in tumour models (Lv et al., 2021; Ran et al., 2016). Co-loading paramagnetic MRI agents such as PE-DTPA-Gd with ZnPc likewise improved longitudinal and transverse relaxivity relative to thin-film-hydrated vesicles, an effect attributed to the smaller diameter, higher surface-area-to-volume ratio, and altered water permeability of uniformly loaded membranes (Skupin-Mrugalska et al., 2021; Skupin-Mrugalska et al., 2018). The broader family of multi-modal payload combinations and their triggered-

Table 1. Mechanistic and performance comparison of conventional bulk versus microfluidic liposome manufacturing. This table contrasts the two manufacturing paradigms across eight dimensions, spanning fluid-mixing dynamics, particle-size control, polydispersity, lamellarity, secondary processing, hydrophobic co-encapsulation, scalability, and process-analytical integration. For each dimension it states the limitation of bulk methods, the corresponding microfluidic advantage, and the downstream consequence for critical quality attributes and GMP readiness. It is intended as the high-level map against which the study-level detail in Tables 2 to 4 should be read.

2. Representative microfluidic liposome studies with device, formulation, process, and outcome parameters. This matrix compiles seven representative studies, from 2D flow focusing and static mixers to 3D-printed and vortex-focusing platforms, detailing device architecture, lipid composition and molar ratio, operating flow parameters (FRR, TFR, or total fluid velocity), resulting hydrodynamic size and PDI, entrapment or loading metrics, and the intended theranostic application. Values are reported as stated in the source studies; because geometries and reporting conventions differ, entries are indicative for cross-study comparison rather than strictly commensurable.

release mechanisms is catalogued in (Table 3).

2.4 Advanced Micromixing Architectures and High-Throughput Scale-Up

For all its analytical appeal, single-channel benchtop microfluidics produces only milligram quantities at microlitre-to-millilitre flow rates, far short of what trials or markets require (Amrani & Tabrizian, 2018; Han et al., 2022; Liu et al., 2022). Scaling by enlarging channels fails, because larger bores break the laminar regime, broaden diffusive length scales, and forfeit size control (Shan et al., 2021; Han et al., 2022). The field's answer has been process intensification that preserves the microscale physics while multiplying it, as mapped in (Figure 3) and quantified across representative devices in (Table 2) and (Table 4).

Four strategies dominate the literature:

• Parallelization (numbering-up). Replicating dozens to hundreds of identical channels on one substrate; branched-manifold arrays of 8 to 128 HFF units have reached production rates above 0.7 kg/day while holding local mixing parameters constant across channels (Liu et al., 2022; Kabay et al., 2022).

• Chaotic advection and static mixers. Internal microstructures such as staggered herringbone mixers, Tesla elements, or static-mixer inserts generate transverse chaotic advection that repeatedly splits, folds, and recombines the streams, shortening diffusion distances and enabling rapid mixing at higher flow rates (Kiuchi et al., 2025; Kabay et al., 2022; Ota et al., 2023).

• Additive manufacturing of multilayer devices. High-resolution 3D printing, notably projection micro-stereolithography at roughly 10 micrometre resolution, yields stacked-channel chips that withstand high pressure without delamination and sustain total flow rates up to 474 mL/min; at such velocities the regime shifts from laminar to controlled turbulence (Re > 2300), where intense shear still produces sub-100 nm, low-PDI liposomes (Shan et al., 2021; Ballacchino et al., 2021).

• Microfluidic vortex focusing (MVF). Miniature hydrocyclonic chambers establish a rotational field around an axial lipid core; tangential buffer injection drives rapid convective stretching near the laminar limit, supporting mass production above 20 g/h in a single device while tuning diameter from 127 nm down to 27 nm at an exceptionally low PDI (Han et al., 2022).

2.5 Batch Consistency, Quality by Design, and Process Analytical Technology

Moving a theranostic liposome from a validated formulation to a regulated product requires more than a capable mixer; it requires a control architecture (Basak et al., 2025; Patel et al., 2025; Pharmaceutics Review, 2025). Under FDA, EMA, and ICH Q8 to Q10 guidance, nanomedicines are treated as complex injectables demanding extensive physicochemical characterization and validated process stability (Basak et al., 2025; Patel et al., 2025). The integrated QbD-and-PAT control loop that satisfies these expectations is diagrammed in (Figure 4), and the device-level PAT and compliance features of leading scale-up platforms are compiled in (Table 4).

A QbD programme begins by defining the Quality Target Product Profile and then identifying CQAs, Critical Material Attributes (CMAs), and Critical Process Parameters (CPPs) (Patel et al., 2025; Pharmaceutics Review, 2025). For theranostic liposomes the CQAs typically include particle size, PDI, zeta potential, lamellarity, encapsulation efficiency, MRI relaxivity or photothermal efficiency, and residual solvent (Patel et al., 2025). Risk tools such as Failure Mode and Effects Analysis and Design-of-Experiments response-surface methods then map the multidimensional design spaces within which FRR, lipid concentration, and mixing temperature may vary without compromising product performance (Basak et al., 2025; Patel et al., 2025).

To displace destructive end-product testing, microfluidic lines increasingly carry in-line and at-line PAT (Basak et al., 2025; Kabay et al., 2022). Flow-through dynamic light scattering records hydrodynamic diameter and PDI in the outflow, nanoparticle tracking analysis reports real-time particle concentration, and in-line UV-Vis or HPLC quantifies drug concentration and unencapsulated payload (Amrani & Tabrizian, 2018; Patel et al., 2025; Basak et al., 2025). Linking these modules to automated closed-loop controllers lets the system self-correct: should an in-line sensor detect upward drift in diameter, the controller adjusts total flow rate or pump pressure to bring the product back within specification (Basak et al., 2025; Kabay et al., 2022). Coupling synthesis with continuous tangential-flow filtration or dialysis extends this into an all-in-one closed platform for purification, solvent removal, and sterile filtration, minimizing human handling and meeting global GMP mandates (Kabay et al., 2022; Basak et al., 2025; Patel et al., 2025).

3. Methods

Because this is a narrative review with a structured, reproducible evidence base, the methodology describes how the literature was identified, screened, and synthesized rather than any wet-laboratory procedure. The protocol was designed so that an independent reader could reconstruct the same corpus and arrive at a comparable synthesis.

3.1 Information Sources and Search Strategy

Four bibliographic databases were queried: PubMed/MEDLINE, Scopus, Web of Science Core Collection, and Google Scholar, the last used only to capture grey literature and forward citations. The search window spanned January 2004, the year the hydrodynamic flow focusing of liposomes was first reported, through June 2026. Searches were restricted to peer-reviewed, English-language records. The final search was executed in June 2025. A Boolean string combined three concept blocks with the AND operator, with synonyms within each block joined by OR: (i) carrier terms (liposome* OR "lipid nanoparticle*" OR nanovesicle* OR "lipid vesicle*"); (ii) process terms (microfluidic* OR "flow focusing" OR "staggered herringbone" OR "vortex focusing" OR "static mixer" OR "continuous manufacturing" OR "numbering-up" OR "3D-printed"); and (iii) application or control terms (theranostic* OR "co-encapsulation" OR "Quality by Design" OR "Process Analytical Technology" OR GMP OR scale-up). Where supported, MeSH descriptors ("Liposomes", "Microfluidics", "Theranostic Nanomedicine", "Drug Compounding") were combined with the free-text terms to improve recall. Reference lists of included articles were hand-searched to recover sources the database queries missed.

3.2 Eligibility Criteria

Records were eligible if they met all of the following: (i) primary experimental studies, authoritative reviews, or regulatory-guidance analyses; (ii) a microfluidic or continuous-flow method used to prepare, load, or scale up liposomes or lipid vesicles, or a direct benchmarking of such a method against a conventional bulk technique; and (iii) at least one quantitatively reported outcome among particle size, PDI, zeta potential, lamellarity, encapsulation or loading efficiency, volumetric throughput, or a defined process-control or regulatory parameter. Studies were excluded when they addressed non-lipid nanocarriers exclusively, reported no extractable quantitative outcome, or were available only as conference abstracts, editorials, or non-English texts.

3.3 Study Selection

Retrieved records were de-duplicated, then screened in two stages, first on title and abstract and then on full text, against the criteria above.. The final corpus that informed the synthesis comprised the primary and secondary sources listed in the References, spanning foundational flow-focusing physics, comparative bench studies, multi-modal theranostic formulations, high-throughput scale-up engineering, and QbD/PAT and regulatory analyses.

3.4 Data Extraction and Synthesis

From each included study a standardized set of fields was extracted into a structured matrix: device architecture and mixer type, lipid composition and molar ratios, operating parameters (FRR, TFR, and total fluid velocity), resulting hydrodynamic size and PDI, encapsulation or loading metrics, throughput, and the theranostic application or process-control feature reported. Because device geometries, lipid systems, and reporting conventions varied too widely for meaningful meta-analytic pooling, the evidence was synthesized narratively and organized into four comparison matrices (Table 1 through Table 4). Quantitative values are reported as stated in the source studies, with units harmonized for cross-study comparison; no numerical re-estimation or imputation was performed. Where primary studies disagreed, the discrepancy is noted rather than reconciled.

4. Flow Parameters as Master Levers in Microfluidic Theranostic Liposome Manufacturing

Read together, the included studies tell a reasonably coherent story, though not an entirely tidy one. The through-line is that a small number of flow variables exert outsized control over the product, and that this control carries forward from vesicle geometry to payload performance to, more tentatively, manufacturing scale. The four matrices referenced throughout, (Table 1) through (Table 4), collect the underlying numbers.

4.1 Flow Parameters as the Master Levers over Vesicle CQAs

If one finding recurs more than any other, it is this: the flow rate ratio and the total flow rate, more than lipid chemistry, decide what comes out of the channel (Skupin-Mrugalska et al., 2021; Han et al., 2022). The mechanism is the one sketched in (Figure 1), raising the FRR narrows the lipid sheet and speeds solvent exchange, so vesicles close earlier and smaller. The magnitudes are striking. In 2D flow focusing with DPPC mixtures, pushing the FRR from 5:1 to 50:1 dragged mean diameters from roughly 800 nm down to 60 to 100 nm while the PDI held below 0.35 (Amrani & Tabrizian, 2018). Vortex focusing did something similar over a full logarithmic FRR sweep at a fixed 60 mL/min, taking DMPC:cholesterol liposomes from 127 nm to 61 nm, and down to 27 nm once PEG2k-PE was added, all at a PDI near 0.04 that, frankly, outperforms most chaotic-advection micromixers reported alongside it (Han et al., 2022). The representative formulations and their resulting CQAs are laid out in (Table 2).

Size is not the only handle. Using an in-line static mixer with 0.6 to 1.0 mm channels, lamellarity could be walked from a single bilayer to as many as sixteen concentric rings by varying total fluid velocity between about 3 and 71 cm/s and initial lipid concentration between 30 and 120 mg/mL (Kiuchi et al., 2025). That higher lamellarity stiffened the membrane and raised payload density for biosensing, apparently without tipping the particles into aggregation, a trade-off worth keeping in mind, since more layers are not automatically better (Kiuchi et al., 2025; Ota et al., 2023). And when microfluidic micromixing was benchmarked directly against thin-film hydration plus extrusion for a POPG/POPC/PE-DTPA-Gd/ZnPc system, the microfluidic vesicles came out as much as 2.5 times smaller (18.5 to 35.2 nm versus 51.6 nm) while staying narrowly distributed (PDI 0.153 to 0.175) (Skupin-Mrugalska et al., 2021).

4.2 Co-Encapsulation Efficiency and Multi-Modal Performance

The second consistent signal concerns how much cargo actually makes it in, and here the advantage for hydrophobic and multi-modal payloads is hard to dismiss (Skupin-Mrugalska et al., 2021; Lv et al., 2021). Across the hydrophobic cases, microfluidic assembly returned a two- to three-fold gain in encapsulation efficiency over thin-film hydration and extrusion (Skupin-Mrugalska et al., 2021). For ZnPc co-loaded with PE-DTPA-Gd, loading rose to 60.6 to 76.2% (about 17 to 22 micrograms per millilitre) against only 23.8 to 28.5% for the conventional route (Skupin-Mrugalska et al., 2021). The downstream effects followed: enhanced longitudinal and transverse relaxivity, which the authors tie to the larger surface-to-volume ratio of sub-35 nm vesicles, and, in SCC-25 and FaDu head-and-neck carcinoma lines, half-maximal inhibitory concentrations of 0.22 to 0.27 micromolar after modest light irradiation (Skupin-Mrugalska et al., 2021; Skupin-Mrugalska et al., 2018). The schematic in (Figure 2) shows where each of these agents sits in the vesicle.

Perhaps the more persuasive demonstrations are the triple-payload systems. One-step chip mixing co-encapsulated 10 nm iron oxide nanoparticles, gold nanorods, and doxorubicin inside folate-targeted thermosensitive liposomes (Lv et al., 2021). At an FRR of 4:1 and 25 mL/min, the process gave uniform vesicles near 230 nm (PDI 0.11), a doxorubicin capacity of 0.57 mg/mL at up to 79% efficiency, and a saturation magnetization of 20 emu/g (Lv et al., 2021). Under 980 nm irradiation the embedded nanorods heated the membrane through its transition, releasing 95% of the drug within about three hours against under 16% at ambient temperature, and in folate-positive bladder cancer cells viability fell from 97.6% to 7.4% (Lv et al., 2021). A comparable logic held for 2 to 4 nm gold nanoparticles packed into DPPC bilayers, where gold-to-lipid ratios topped 60 wt%, a roughly three-fold improvement over reverse-phase evaporation, with cryo-TEM confirming the particles sat cleanly in the membrane core (Al-Ahmady et al., 2018). The fuller catalogue of payload pairings, loading metrics, and triggering mechanisms is given in (Table 3).

4.3 Throughput: Closing the Gap to Industrial Scale

Whether any of this scales is the question that has nagged the field, and the evidence here is encouraging if still early (Shan et al., 2021; Han et al., 2022). The two routes that stand out, summarized in (Figure 3) and (Table 4), attack the problem differently. Three-layer PµSL-printed chips with 400 by 400 micrometre channels tolerated extreme injection velocities without delaminating and reached a total flow rate of 474 mL/min, the highest reported for liposome preparation (Shan et al., 2021). At that velocity the flow is no longer laminar at all; the regime crosses into controlled turbulence (Re near 6583), yet rapid shear still returned 64 to 84 nm vesicles at PDI below 0.20, which is not the outcome one might naively predict once turbulence enters the picture (Shan et al., 2021).

Vortex focusing took the other path, trading peak flow rate for sustained mass output (Han et al., 2022). At 80 mL/min and lipid feeds up to 60 mM, single hydrocyclone devices held continuous production above 20 g/h for PEGylated

Figure 3. Process-intensification routes from benchtop microfluidics to industrial throughput. Four engineering strategies scale production while preserving microscale mixing physics. Parallelization (numbering-up) replicates 8 to 128 identical channels for outputs above 0.7 kg/day; chaotic advection and static mixers shorten diffusion distances at higher flow; 3D-printed multilayer chips reach total flow rates up to 474 mL/min; and microfluidic vortex focusing sustains more than 20 g/h at a PDI near 0.04. Each route avoids the loss of monodispersity that simply widening channels would cause.

Figure 4. GMP-ready control architecture integrating Quality by Design and Process Analytical Technology. A closed-loop manufacturing scheme in which critical material attributes and critical process parameters feed a continuous microfluidic synthesis stage, whose output is monitored in real time by in-line dynamic light scattering, flow-through UV-Vis/HPLC, and nanoparticle tracking analysis. Automated feedback dynamically adjusts flow rates to correct size drift and couples with tangential-flow filtration for continuous purification, delivering batch consistency compliant with FDA, EMA, and ICH Q8 to Q10 guidance.

lipids and about 7.2 g/h for dicetyl phosphate formulations, with light scattering indicating essentially complete conversion into unilamellar vesicles and no detectable micellar contamination (Han et al., 2022). For rapid screening rather than mass production, 2D flow focusing generated monodisperse liposomes at outlet concentrations up to 1740 by 10^8 particles per millilitre within two minutes, a throughput well suited to pilot-scale work (Amrani & Tabrizian, 2018). None of these devices has, to be clear, been shown end-to-end under full GMP conditions in the sources reviewed; the throughput case is strong, the manufacturing-qualification case is still being built.

4.4 Process Control and the Road to QbD Compliance

The last body of evidence is less about numbers than about architecture, how a continuous line is kept within specification (Patel et al., 2025; Basak et al., 2025). The regulatory starting point is consistent across sources: FDA, EMA, and ICH Q8 to Q10 designate size, PDI, zeta potential, lamellarity, encapsulation efficiency, and residual solvent as CQAs that must be controlled, not merely measured (Patel et al., 2025; Basak et al., 2025). The control loop that follows, inputs, synthesis, in-line measurement, and automated correction, is the one drawn in (Figure 4), with platform-level features tabulated in (Table 4).

In practice this means FMEA and DoE response-surface work to define a design space within which TFR, FRR, lipid concentration, and temperature may move without harming the product (Patel et al., 2025; Basak et al., 2025). It means replacing off-line release testing with flow-through DLS for size and PDI, flow-through UV-Vis or HPLC for drug concentration and leakage, and NTA for particle count (Patel et al., 2025; Basak et al., 2025). And it means wiring those sensors to a controller that can, for instance, nudge the total flow rate upward the moment vesicle size begins to drift, holding batch consistency in a way manual processes never quite managed (Basak et al., 2025; Kabay et al., 2022). It is this closed loop, more than any single device record, that makes the GMP case credible.

5. Microfluidics Reframes Theranostic Liposome Manufacturing as a Process Control Problem

What the synthesis above adds up to is a shift in where the hard problem lives. For years the difficulty in theranostic liposome development was framed as a formulation question, which lipids, which payloads, which ratios. The microfluidic literature suggests the more stubborn constraint was always the process, and that once the process becomes controllable, several long-standing problems soften at once.

5.1 Why Microscale Control Reframes the Manufacturing Problem

The central advantage is almost deceptively simple: microfluidics makes vesicle formation deterministic rather than statistical (Jahn et al., 2004; Han et al., 2022). Because assembly is governed by diffusive transport in a laminar sheet (Figure 1), the same inputs give the same vesicles, batch after batch, which is exactly the property conventional bulk mixing cannot guarantee (Basak et al., 2025; Lamprou et al., 2021). The contrasts gathered in (Table 1) are therefore not just a list of incremental improvements; they describe a move from a process defined by its uncontrolled gradients to one defined by its controlled parameters. That reframing matters because reproducibility, not peak performance, is what regulators ultimately certify (Patel et al., 2025). A method that reliably produces a slightly suboptimal vesicle is, for manufacturing purposes, worth more than one that occasionally produces a perfect one.

5.2 Co-Encapsulation as the Real Theranostic Bottleneck

If there is a place where microfluidics earns its keep most clearly, it is multi-modal co-encapsulation (Skupin-Mrugalska et al., 2021; Lv et al., 2021). Theranostic constructs ask a great deal of a carrier, a drug, a contrast agent, sometimes a photothermal element, all held at a defined ratio in one vesicle (Figure 2), and this is precisely the scenario bulk methods handle worst, through competitive displacement and erratic stoichiometry (Al-Ahmady et al., 2018). The evidence in (Table 3) indicates that single-step continuous assembly largely sidesteps that failure mode, delivering not only higher hydrophobic loading but functional gains that follow from it, sharper MRI relaxivity, cleaner triggered release, and in several models genuinely synergistic chemo-photothermal cytotoxicity (Skupin-Mrugalska et al., 2021; Lv et al., 2021; Ran et al., 2016; Fernandes et al., 2023). One should be careful not to overgeneralize from a handful of well-characterized systems, but the direction of the evidence is consistent.

5.3 Scale-Up: Competing Philosophies, Unresolved Trade-offs

Table 3. Multi-modal theranostic payload combinations and their microfluidic co-encapsulation and triggering mechanisms. This table categorizes five representative theranostic cargo combinations by diagnostic modality, therapeutic mechanism, microfluidic loading strategy, encapsulation and physicochemical metrics, and the stimulus that triggers drug release. It spans single-therapy bimodal systems (ZnPc + Gd) through triple- and quadruple-payload constructs for combination chemo-photothermal-photodynamic therapy and brain-targeted delivery, illustrating how single-step continuous assembly supports high-stoichiometry co-loading that bulk methods struggle to achieve.

4. Industrial scale-up routes with throughput, mixing physics, PAT, and regulatory-compliance features. This table presents four process-intensification routes, numbering-up arrays, 3D-printed multilayer chips, vortex focusing, and integrated all-in-one platforms, with their representative devices, volumetric output, governing mixing mechanism, in-line Process Analytical Technology, and the regulatory or GMP considerations each addresses. Read alongside Figure 3, it maps how each strategy preserves microscale mixing fidelity while raising throughput, and where closed-loop control enables a benchtop-to-commercial transition compliant with FDA, EMA, and ICH Q8 to Q10 expectations.

On scale-up the literature is genuinely interesting precisely because it does not converge on one answer (Shan et al., 2021; Han et al., 2022; Liu et al., 2022). The strategies in (Figure 3) and (Table 4) embody different bets. Parallelization preserves the exact single-channel physics and multiplies it, which is conceptually clean but shifts the engineering burden onto uniform flow distribution and channel fouling across an array (Liu et al., 2022; Kabay et al., 2022). Multilayer 3D-printed chips chase raw flow rate and, interestingly, abandon strict laminarity altogether once they cross into controlled turbulence, yet still hold size and PDI, which complicates the tidy narrative that laminar flow is the thing that matters (Shan et al., 2021). Vortex focusing optimizes for sustained mass output instead of peak rate (Han et al., 2022). It would be premature to crown a winner; the more honest reading is that the optimal route probably depends on the product, batch sizes for a rare-disease theranostic and for a mass-market agent are not the same problem, and the field has not yet had to choose.

5.4 The Regulatory Frontier and What Remains Missing

The QbD-and-PAT architecture in (Figure 4) is, on paper, a near-complete answer to the GMP question (Basak et al., 2025; Patel et al., 2025). In-line DLS, NTA, and UV-Vis feeding a closed-loop controller gives continuous verification and self-correction, the two things end-of-batch testing cannot (Kabay et al., 2022). Yet a gap persists between demonstration and qualification. Most PAT integrations in the reviewed sources are shown as feasible rather than validated across extended GMP campaigns, and the regulatory classification of continuously manufactured nanomedicines is still maturing (Patel et al., 2025; Pharmaceutics Review, 2025). The technology, in other words, appears ready before the regulatory precedent does, a familiar pattern, and one that tends to resolve only once a first product clears the path.

5.5 Limitations of the Evidence Base and Future Directions

Several caveats temper the optimism. Reporting is heterogeneous; device geometries, lipid systems, and the way CQAs are measured differ enough that cross-study comparison, including the matrices in (Table 2) and (Table 3), should be read as indicative rather than strictly commensurable (Kastner et al., 2014; Ota et al., 2023). Much of the theranostic performance data comes from a small set of model systems and cell lines, so in vivo durability, long-term stability, and immunogenicity remain comparatively under-reported (Fernandes et al., 2023; Sonkar et al., 2021). And the throughput records, impressive as they are, have mostly been established for relatively simple formulations; whether a 20 g/h vortex device holds its monodispersity for a three-payload theranostic is an open question (Han et al., 2022; Lv et al., 2021). Useful next steps would include standardized reporting conventions, head-to-head scale-up comparisons on identical theranostic formulations, and, above all, a fully documented GMP campaign that takes one such construct from chip to clinic (Basak et al., 2025; Patel et al., 2025).

6. Conclusion

Microfluidic continuous-flow manufacturing has moved theranostic liposomes from a formulation curiosity toward a manufacturable medicine. By replacing chaotic bulk mixing with deterministic, diffusion-governed self-assembly, it delivers size-tunable, narrowly dispersed vesicles in a single step and, with them, the two- to three-fold gains in hydrophobic and multi-modal payload loading that theranostic constructs demand. Process-intensification routes, parallelized arrays, multilayer 3D-printed chips, and vortex focusing, now reach throughputs compatible with clinical and commercial supply without sacrificing monodispersity, while integrated Quality by Design and Process Analytical Technology supply the real-time, self-correcting control that regulated production requires. What remains is less a matter of capability than of consolidation: standardized reporting, head-to-head scale-up studies on identical formulations, and a fully documented GMP campaign that carries one construct from chip to clinic. For a field long constrained at the bench, the remaining barriers now appear more organizational than technical. On present evidence, microfluidics is the most credible route yet to reproducible, scalable theranostic nanomedicine.

References