Recent advances in nano-photonic techniques for pharmaceutical drug monitoring with emphasis on Raman spectroscopy

2.1 Surface-enhanced Raman spectroscopy for drug monitoring

Simply increasing the laser power to reach high sensitivities is not a valid option as bulky high-power lasers are not practical for on-site or bed-side application. Instead, micro- and nano-structured materials are employed to make the best use of the available laser power. One way is to locally enhance the optical field via the surface plasmon resonance effect. This localized electromagnetic field enhancement is achieved when the wavelength of the illuminating light is resonant with the plasmon frequency of a nano-scaled metal structure [48], [49], [50]. The Raman signal of sample molecules in the close vicinity of the metal structure is consequently strongly enhanced. The method based on this local but high Raman signal enhancement is called surface-enhanced Raman spectroscopy (SERS) [51]. An additional enhancement of the Raman signals occurs when molecules are chemisorbed on the SERS-active nano-metals (chemical enhancement), creating new molecular states [52]. Also, colloidal metal solutions can be used as SERS substrates [53].

Nowadays, SERS accounts for approximately 20% of the published articles on Raman drug sensing, according to the Web of Science database (from 298 publications between 2010 and 2018, containing the keywords “Raman” and “drug”, 60 feature “SERS”). Several reviews report on the diverse applications of SERS, e.g. in pharmacy [54], [55], [56], for TDM [52], [57], for biological and biomedical applications [58], [59], [60], [61], and in forensics [40], [62]. In this review, we focus on advances in SERS with potential towards drug sensing in body fluids which were published after 2017 (see Table 1).

2.1.1 Advances in SERS substrate development for drug sensing applications

For sensing applications, SERS substrates need to provide reproducibly high Raman signal enhancements. Large variations between different substrates of the same type would introduce unacceptable quantification errors. Currently, standard deviations of about 20% are still considered reasonable [72]. However, clinical and forensic applications must fulfill much higher standards and require significant improvements. Yan et al. [63] developed a method that ensures high reproducibility of gold (Au) nanoparticle (NP)-based SERS substrates. The technique is based on NPs transforming from the wet state to the dry state, forming optimal hot spots, arranged in a 3D long-range-ordered structure, which is stable for dozens of seconds [63]. The presented substrate provides good reproducibility; quantitative measurements showed a standard deviation of only 8.1% [63]. They tested the detection of the party drug 3,4-methylenedioxy-methamphetamine and the synthetic psychoactive drug α-methyltryptamine hydrochloride in water for concentrations down to 10 μm and 1 μm, respectively [63].

Besides gold, silver (Ag) is a frequently used metal for SERS substrates. Gao et al. presented Ag-Au-hybrid NPs with unusually long shelf life (12 months) [69] and strong binding of thiol groups, which is very useful for biomedical applications [69]. Ag NPs, in this case seeded on graphene nanosheets, have also proven to be sensitive for the anticancer drug 2-thiouracil down to 10 nm in water [67]. Haddad et al. [70] proposed an approach for monitoring the synthetic opioid fentanyl (also in mixture with heroin) using a simple paper-based substrate impregnated with Ag NPs [70]. Moreover, Hong et al. [91] developed Ag-nanogratings to monitor the antibiotics enofloxacin and ciprofloxacin, using a portable Raman spectrometer and chemometrics for quantification [91].

In addition to the noble metals, copper (Cu) was demonstrated to enable SERS. The Cu-NPs were temporally protected against oxidation by porous calcium carbonate microspheres and enabled the sensitive detection of two antibacterial drugs (ceftriaxone and sulfadimethoxine) in injection solutions and tablets [72].

Also, colloidal Ag [68] and Au [66] NPs were useful for the detection of antibiotics [66], [68] and antimalarial active ingredients [68] in aqueous solutions and allowed quantification in pharmaceutical dosage forms such as (substandard) tablets [68] and eye drops [66]. However, in these cases either the preparation of the colloidal substrate [66] or the sample preparation [68] is quite time-consuming and extensive. An innovative strategy for TDM of the anti-cancer drug erlotinib using an Au-colloid-SERS assay was introduced by Litti et al. [83]. They used an internal reference – propynyl fluorescent red (PFR), a good Raman scatterer – which competes with the drug for the binding spots on the surface of functionalized gold NPs (see Figure 3). Usually, erlotinib levels are low and PFR dominantly binds to the surface of the Au NP clusters. The resulting intense PFR SERS signal is negatively proportional to the target drug concentration.

Figure 3:

Schematic working principle of SERS mediated monitoring of the anti-cancer drug erlotinib.

Erlotinib and PFR compete for the azide-groups on the surface of the functionalized Au NP. In common applications, erlotinib levels are low and the strong Raman scatterer PFR dominates on the surface of the Au NP clusters. The resulting intense SERS signal is negatively proportional to the target drug concentration. Adapted from Litti et al. [83], reproduced with the kind permission of Elsevier, Copyright (2019) Elsevier.

2.1.2 Towards SERS measurements in body fluids

Body fluids are challenging media for Raman sensing in general. The components of these liquids give rise to signals which can overlap with the signal from the substance of interest. Especially proteins complicate the measurement procedure, as they are fluorescent and thus cover the Raman signals of the analyte (e.g. drug). For SERS measurements, it should also be considered that proteins can bind the low-concentrated drugs or cover the NP surface of the SERS substrates, therefore inhibiting the activity of the “hot spots” [52].

Recently, special substrates were introduced to address the protein problem. Yue et al. [82] developed semipermeable alginate microparticles containing gold NPs [82]. The alginate membrane acts like a filter and only allows small molecules to enter and interact with the gold NPs. Large proteins are thus hindered from forming a protein corona around the Au NPs (Figure 4), resulting in a good SERS signal of the analyte [82]. In serum, this method could reach a reproducibility of 10% [82]. Panikar et al. introduced a brush layer to prevent SERS-hot spot blockages and fouling by blood serum proteins [84]. The monolayer is built by hierarchically modified, self-assembled zwitterionic L-cysteine [84]. The complete substrate is called graphene oxide (GO)-supported L-cysteine-functionalized star-like gold NPs (SAuNPs@GO) (Figure 5) [84]. It has additional advantages, such as a very high Raman signal enhancement factor as well as high stability and reproducibility [84]. Both approaches aim to find application in point-of-care TDM in cancer therapy [82], [84].

Figure 4:

Depiction of the protein problem in SERS measurements and possible solutions.

(A, a) The nonspecific adsorption of proteins on Au NPs leads to “protein coronas” which block the analyte (e.g. drug) from entering the hot spots close to the NPs surface. (A, b) SERS measurements with colloidal gold: while the spectrum of 1.0×10⁻⁶ 4-aminothiopheno (4-ATP) in aqueous solution (red) shows pronounced peaks, the spectrum for 1.0×10⁻⁶m 4-ATP in serum (blue) has no characteristic features due to the protein problem. (B, a) Semipermeable microparticles prevent the diffusion of proteins but allow the diffusion of small molecules. The hot spots remain accessible, and the protein problem is avoided. (B, b) SERS measurements with colloidal gold within semipermeable microparticles: The spectra of 1.0×10⁻⁶m 4-ATP in water and in serum both show strong characteristic features. Adapted from Yue et al. [82], reproduced with the kind permission of Elsevier, Copyright (2018) Elsevier.

Figure 5:

Schematic illustration of the production and application of a SERS substrate modified to avoid the protein problem.

(A) Fabrication of the graphene oxide (GO)-supported functionalized starlike gold nanoparticles. L-cysteine is used to functionalize the NPs. (B) The L-cystein monolayer acts as brush layer and prevents blood serum proteins from blocking the SERS hot spots. Adapted from Panikar et al. [84], reproduced with the kind permission of the American Chemical Society, Copyright (2019) American Chemical Society.

Another approach to overcome matrix-related challenges is to combine sample pretreatment methods with the sensing platform [73], [74], [76], [79], [81], [86], [88], [90]. For drug sensing in body fluids, several studies reported the combination of SERS and high-performance liquid chromatography (HPLC) [56]. Unfortunately, this combination compromises the advantage of Raman spectroscopy: quick and straightforward point-of-care applicability. Other reported separation techniques like liquid extraction [73], [74], [77], (ultra-)filtration [59], [79], [80], [88], and passing the sample through a purification column with silica gel [76] are more compatible with point-of-care use.

Berger et al. reported about a point-of-care applicable platform for SERS, based on a paper membrane for vertical flow separation [79]. The filter membrane is combined with inkjet-printed Ag NPs (Figure 6). This method has high potential as TDM tool in the clinic. The proposed procedure is simple; one measurement only takes 15 min overall, and the single components are cheap.

Figure 6:

Working principle of a point-of-care applicable platform for SERS based on a paper-filtering membrane.

The micro-structure is visible in the SEM picture of the Ag NPs on cellulose paper. After application of the sample liquid, proteins are separated from the active ingredients by vertical flow filtering. The SERS signal of the analytes can then be read from the paper SERS sensor. Adapted from Berger et al. [79], reproduced with the kind permission of Elsevier, Copyright (2017) Elsevier.

Besides blood, urine is another body fluid of interest. It is easily available without any invasive measures such as a needlestick. As a sample matrix, it is easier to handle because it does not contain as many proteins as serum or plasma. Bindesri et al. [75] combined electrochemistry and SERS, using a fabric-based electrochemical plasmonic sensor [75]. All electrodes were impregnated with conducting silver ink, and the fabric was dried. An additional carbon ink line was deposited onto the working electrode and topped with Ag NP paste. The counter electrode was also coated with carbon ink, while the reference electrode remained unchanged (silver ink). When the sensor surface charges are manipulated through the application of voltage, neutral analyte species interact with the silver metal surface. Uncharged analytes can thus be measured quickly and easily. In this study, the authors succesfully tested their sensor for TDM of the antibiotic levofloxacin in synthetic urine down to a concentration of 1 mm [75].

Markina et al. [76] developed a fast detection method for another antibiotic (ceftriaxone) in urine, enabling SERS quantification in 10 min overall. The following simple pretreatment steps were used: pH adjustment, passing the sample through a purification column with silica gel, and filtering through a cellulose filter [76].

Urine is interesting not only for TDM applications, but also for forensic sensing, especially to determine drug abuse easily. Liquid-liquid microextraction is a common approach [73], [74], [77] for this purpose. This method can be quick as well; the overall process requires just 5–6 min [77].

2.2 Fiber-enhanced Raman spectroscopy for drug monitoring

Another very promising approach for high signal amplification with potential for on-site and point-of-care sensing was recently introduced with fiber-enhanced Raman spectroscopy (FERS) [47], [92], [93], [94], [95], [96]. In FERS, the laser light is coupled into a hollow-core optical fiber, which is also filled with the sample solution. The interaction between the guided laser light and the analyte molecules is thus increased over an extended length. As a result, the number of molecules interacting with the light is highly increased (Figure 7). For the fiber enhancement, several factors, such as laser intensity, fiber guiding properties and length, and the coupling efficiency play a major role [47]. These parameters differ for different fiber types. For FERS sensing with hollow-core photonic crystal fibers (HC-PCFs), a tremendous increase in signal intensity is possible. Ten to fifty times higher sensitivities and improvements in limits of detection (LODs) of more than one order of magnitude have been achieved for pharmaceuticals dissolved in aqueous environment [97], [98]. FERS is a very flexible technique which can be miniaturized and shows high potential for point-of-care monitoring of drugs [97], [98] or biomarkers [99] (see Table 2).

Figure 7:

Comparison between the limited collection volume in conventional Raman spectroscopy (A) and the extended interaction volume in FERS (B).

Enhancement effect in FERS: The volume of interaction and light collection in conventional measurements (A) is limited to the focal volume. In FERS (B), the illuminating laser light is guided along the sample filled core of a fiber. The light-analyte interaction is thus increased and the Raman signal is enhanced. Adapted from Frosch et al. [47], reproduced with the kind permission of the American Chemical Society, Copyright (2013) American Chemical Society.

The bandgap of specific HC-PCFs can be shifted into the visible range, when the fiber is filled with aqueous analyte solution (see Figure 8A). This effect was exploited for the monitoring of the antibiotic levofloxacin in the visible range (excitation wavelength 532 nm) down to a concentration of 20 μm [60]. The required sample volume for the filling of the sensor fiber is very low (8 nl). By selectively filling the central hollow core of the fiber [98], broadband step-index guiding is achieved. This enables the use of excitation wavelengths throughout the whole visible range (Figure 8B). Thus, the excitation wavelength can be chosen depending on the analyte requirements, taking spectral characteristics such as absorption maxima or fluorescence bands into account. In this way, additional enhancement can be achieved through the resonance Raman effect [99], and the spectral range where fluorescence is likely to occur might be avoided. Using a selectively filled fiber, a LOD of 1.7 μm was achieved for the antibiotic moxifloxacin in water [51].

Figure 8:

Transmission spectra of two different hollow core fibers for FERS of liquids.

(A) Compared with the air-filled HC-PCF, the spectrum of the water filled fiber is blue-shifted into the visible spectral range. The sensor fiber is thus suited for FERS with a laser wavelength of 532 nm and Raman spectra up to 4000 cm⁻¹. (B) In this case, not the whole microstructure of the fiber is water filled. Instead, just the central core of the fiber is filled with water, leading to a change in guiding properties (from band-gap to step-index guidance). The measured spectrum roughly follows the prediction and shows that the fiber offers broadband transmission in the visible range. Adapted from Yan et al. [59], [97], reproduced with the kind permission of the Royal Society of Chemistry (A), Copyright (2017) Royal Society of Chemistry and the American Chemical Society (B), Copyright (2017) American Chemical Society.

FERS based on HC-PCFs was further exploited in combination with SERS [107]. Yet, it remains to be proven whether these approaches will be practicable for drug monitoring. So far, fiber-SERS techniques were only tested for the detection of model substances, biomarkers, or cells for biomedical applications but not yet for approved pharmaceuticals [107].

There is a major drawback using HC-PCFs. The small central hollow core is difficult to fill with analyte solutions and also hampers stable light coupling. To efficiently integrate the fiber into a spectroscopic system, the coupling needs to be very precise. To achieve 90% performance, a lateral offset of just ±1.5 μm is allowed [97].

This challenge can be solved by using customized air-clad fibers (Figure 9A) with bigger central core, which were specifically tailored for highly sensitive drug monitoring [101]. FERS with these fibers enabled a signal enhancement factor of about 60 times for the antibiotic cefuroxime [101]. The broadband guiding mechanism of this selectively filled fiber type allows the use of different excitation wavelengths in the visible and also up to the near infrared (NIR) range. By choosing the excitation wavelength of 835 nm, fluorescence background of human urine could be avoided [101]. Drug monitoring in human urine was possible down to concentrations of 100 μm [101]. Moreover, the method was validated on real urine samples from volunteers and cross referenced with HPLC measurements. Thus, this approach proves to be a very promising alternative to the time-consuming, lab-based TDM methods used up to date.

Figure 9:

Cross sections of a novel silica air-clad hollow-core fiber (A) and polymer optical hollow-core fiber (B) for FERS sensing.

(A) SEM image of the cross-section of an air-clad fiber which was developed specifically for highly sensitive drug monitoring. The diameter of the central core is approximately 30 μm. (B) Experimental setup for FERS for glucose sensing. The microstructure of the polymer hollow-core optical fiber is modified on one end: all openings but the central core are closed. This end is dipped into the sample solution which is sucked into the central core as a result of the capillary effect. Laser coupling and signal collection through an objective lens takes place at the other end of the fiber. The inset shows a microscope image of the upper end face of the micro-structured polymer hollow-core optical fiber. Adapted from Yan et al. [107] (A) and from Azkune et al. [102] (B), reproduced with the kind permission of the American Chemical Society, Copyright (2018) American Chemical Society (A) and IEEE (B).

Recently, Azkune et al. reported about a micro-structured polymer hollow-core optical fiber (POF) system (Figure 9B) that was used for glucose sensing [102]. This FERS platform could perspectively be used for TDM of the sodium-glucose co-transporter 2 inhibitors in the therapy of diabetes mellitus. Through careful preparation of the fiber and additional steps in data processing and analysis, glucose sensing in urine could be demonstrated via POF-FERS [102].

2.3 Fiber-array based Raman chemical imaging for drug monitoring

In pharmaceutical applications, it would be ideal to gain complete qualitative and quantitative information about the chemical composition of a formulation preferably in a short time. The distribution and the particle size of the APIs strongly influence the biopharmaceutical profile of the formulations. Yet, these properties are neither directly analyzed during the production process nor in the formulation. For the quality assuring analysis of the active ingredients’ concentrations and their pharmacokinetic profile, the samples have to be dissolved. This is time consuming, and the analysis requires expensive instrumentation. This aspect is especially critical in the case of counterfeit medicines, which are spreading all over the world, and in a greater extent in poor countries.

To address this problem, a fiber-array-based chemical imaging method [103], [108], [109] was adapted and successfully applied for counterfeit and substandard tests of the antimalarial tablet Riamet^® [103]. This technique exploits the flexibility and efficiency of light guidance in optical fibers and enables the readout of numerous spectra from different regions of the sample [103], [108], [109] (see Figure 10). Therefore, chemical imaging of a large area is possible with high spatial resolution. In the antimalarial tablets, the two active ingredients artemether and lumefantrine were quantified simultaneously in situ, without dissolution. With this setup, the distribution of active ingredients can be visualized in a time efficient way, as in one shot 64 regions can be imaged with a spatial resolution of about 1 μm [103]. Consequently, fiber-array based Raman imaging has a high potential to become an on-line or at-line quality control method for pharmaceuticals.

Figure 10:

Sketch of the experimental setup for fiber array-based Raman hyperspectral imaging.

The rectangular area of illumination in the focal plane is achieved with a long propagation length of the laser light within a step index multimode fiber (MF). The backscattered signal is imaged onto an 8×8 fiber array. The fiber arrangement within the fiber bundle (FB) changes from 8×8 square to 64×1 line. The linear array is positioned in the slit plane of the spectrometer enabling hyperspectral imaging with reduced effort for acquisition. Adapted from Frosch et al. [103], reproduced with the kind permission of MDPI.

Furthermore, fiber-array based Raman imaging could help to speed up SERS measurements. In SERS, a lot of spectra are needed for robust quantitative measurements (e.g. in the study of Souza et al. [89] at least 100 spectra are required). Frequently, the sample surface is mapped for this reason. However, the time-consuming mapping is limiting the measurement speed [110] which could be mitigated using an array-based imaging approach [103].

2.4 Spatially offset Raman spectroscopy for drug monitoring

Spatially offset Raman spectroscopy (SORS) allows the detection of analytes below a non-absorbing cover material, such as in the case of packaged pharmaceuticals [111]. This is realized by the acquisition of two Raman spectra: first from the container, in conventional geometry, and second from a deeper layer by lateral offset from the illumination spot. The principle is the following (Figure 11): as the incident laser light reaches the sample, part of the light will reach deeper layers of the sample (with exponentially decaying intensity). Equally, some of the diffuse Raman-scattered photons will migrate upwards with a lateral offset. [112] The second Raman spectrum recorded with the lateral offset is then scaled and subtracted from the first, yielding the subsurface Raman spectrum [111]. The deeper the layer of interest, the larger is the necessary spatial offset of the detection.

Figure 11:

Principle of spatially offset Raman spectroscopy.

Raman scattered light from deeper layers can be collected in a distance Δx from the spot of illumination. The larger the distance, the higher the ratio of signal from deeper layers, but also the smaller the signal intensity. Δx=0 is equal to the signal collection geometry of conventional Raman spectroscopy.

Using this method in combination with multivariate analysis, several pharmaceuticals were detected [111] and quantified [104]. Olds et al. [104] reported about the quantification of the frequently used analgesic acetaminophen (also called paracetamol) and the decongestant agent phenylephrine in a powder mixture with glucose and caffeine down to 5% [104] through the polypropylene container. Zhang et al. [113] reported the detection of acetaminophen and the antibiotic drug metronidazole through polyethylene and polytetrafluorethylene bottles [113].

The SORS technique was advanced and developed into a handheld device called “Resolve” by the company Cobalt Light Systems Ltd (part of Agilent Technologies). The method can thus be applied in the field, e.g. for the detection of toxic industrial chemicals, flammable compounds, explosives, narcotics, and chemical warfare agents [114].

The sensitivity and specificity of this technique can be further increased by combining SORS with SERS: the so called SESORS [115], [116]. Stone et al. reported about resonant SERS-active molecule-specific reporter NP conjugates, which can be delivered into the skin and provide disease-specific Raman signatures [77]. Resonant dyes as conjugants enable the detection of very low reporter molecule concentrations, as the dyes deliver highly enhanced signals. These Raman signals can be recorded with the help of the deep Raman technique SORS through thick tissues of several tens of millimeters [116], pushing the LODs far beyond conventional SORS [116]. This method might find further application in monitoring drugs or drug delivery processes in different tissues.

The first step in this direction was already taken by Moody et al. They analyzed neurotransmitters in brain-mimicking agarose gel through a cat skull via SESORS [105]. Combining this method with principal component analysis, the authors could differentiate between the neurotransmitters melatonin, serotonin, and epinephrine [105].

Combining SORS with Raman imaging for the detection of explosives from a distance [117] allows the collection of information from different spatial offsets simultaneously. Thus, the imaging facilitates the process, as the optimal spatial offset is unknown [117]. Overall, SORS is a very promising method for applications in pharmaceutical, forensic, and security substance monitoring and enables tests through the packaging.

2.5 Raman optical activity and enantioselective Raman spectroscopy for drug monitoring

Many pharmaceutical drugs and biomolecules are chiral substances. These molecules can exist in at least two different steric formations, i.e. left-handed (l) or right-handed (d). The two versions are called enantiomers. The stereo-chemical structure of a pharmaceutical active ingredient is decisive for the binding to its specific target structure (often proteins or DNA). Enantioselective Raman spectroscopy (EsR) and Raman optical activity (ROA) are specific Raman techniques for the identification and quantification of chiral molecules. Both techniques take advantage of the optical activity of chiral molecules. Optical activity describes the phenomenon that the interaction of polarized light with left-handed or right-handed enantiomers leads to different effects.

EsR is a novel approach developed by Kiefer et al. [118] based on polarization resolved signal detection. The sample is illuminated with linearly polarized laser light, and a half-wave retarder in the detection arm of the setup enables the investigation of the optical activity of the sample. As a result, enantiomers can be distinguished and enantiomeric fractions can be determined [119]. EsR is thus promising, e.g. for the on-line monitoring of synthesis processes of chiral drugs [120]. The approach was recently advanced with an automatized polarization-resolved Raman setup and coupled with chemometric analysis [119], [121] which enabled enantiomeric differentiation of butan-2-ol [120], [121]. Also, the enantiomeric fractions of the chiral molecule (5,6)-diphenyl-morpholin-2-one could be determined [119].

ROA takes advantage of the fact that excitation of the sample by right- or left-circularly polarized light results in differences in the Raman spectra [122]. Employing this approach, Gąsior-Głogowska et al. [123] monitored the interaction of the anticancer drug cisplatin with the target DNA structure via the detection of conformational rearrangements within the chiral sugar moieties [123].

EsR and ROA could be especially interesting for the in situ quality assessment of novel biopharmaceuticals, particularly where glycoproteins play a pivotal role [124]. They could help to characterize the polypeptide backbone and the carbohydrate conformation at the same time. In the first study about the application of ROA for the investigation of a biopharmaceutical formulation, a therapeutic monoclonal antibody was characterized, looking into detail at the overall three-dimensional structure of the protein [125].

The implementation of these techniques in drug sensing is still very limited. EsR is just emerging, the nearest to real drug application was the reported proof-of-principle study with a morpholinone derivative [119].

The major challenge for the application of ROA is the extremely low signal intensity, which is three to five orders of magnitude lower than in conventional Raman scattering [126]. Therefore, enhancement techniques are necessary. Analog to other Raman techniques, ROA can be combined with SERS, exploiting the enhancement effect of surface plasmon resonance. Surface-enhanced Raman optical activity (SEROA) was tested using Ag nanowires for the detection of the peptide fmoc-glycyl-glycine-OH [126]. Furthermore, a special nanotag was developed for analyzing d- and l-tryptophan and d- and l-ribose. The nanotag is based on a Ag NP core coated in a silica shell, to which the benzotriazole reporter molecule and the analyte are attached [127]. Recently, another approach for SEROA has been published by Wu et al. [128]. Here, the authors used hybrid magnetic NPs and showed that the ROA to Raman ratio is enhanced by the magnetic properties of the NPs [128].

It is not known yet how these advanced ROA approaches will perform in investigations of structural and thus qualitative changes of real biopharmaceuticals. Nevertheless, the developments in this direction are promising for the sensitive detection of peptide and sugar structures and their moieties upon drug action or degradation processes.

3.1 Drug monitoring in the development of pharmaceuticals

In pharmaceutics, chemically selective methods are highly desired as PAT for on-line or in-line quality control. Raman spectroscopic techniques could meet the requirements. Currently available portable Raman spectrometers facilitate the application in the pharmaceutical industry [14], but their low sensitivity and spectral resolution limit the use. However, continuous improvements and newly developed technologies are expanding the range of applications.

Fiber-based systems can be used at-line or in-line and allow high flexibility. Especially for chemical imaging, fiber-based techniques are very promising. APIs in solid dosage forms can be monitored without dissolution. Data acquisition speed can be improved markedly while high spatial resolution is maintained [103].

Going deeper, subsurface Raman techniques, such as SORS, provide a solution for the measurement of the drug content through the package [104], [111]. The low signal intensities can be significantly enhanced by combining SORS with SERS (SESORS). However, the applicability of SORS is limited when the sample is highly absorbing at the laser or Raman signal wavelengths. Absorption greatly reduces photon propagation distances and penetration depths [115]. Raman signals from deeper layers require larger spatial offset detection which results in lower signal intensities from the target. Consequently, penetration depth and needed signal intensity have to be balanced.

For in situ quality assessment of biopharmaceuticals, advanced ROA methods, such as SEROA, are highly interesting. However, due to the very low intensity of ROA (even with plasmonic resonance enhancement), the technique requires higher sample concentrations and longer data collection periods. This can induce adverse laser light-derived degradations of sensitive APIs. FERS could be an option for further enhancement [47], [92].

Furthermore, FERS could also be exploited for monitoring dissolution processes in low sample volumes in a highly sensitive and chemically selective way. Both flow-through and static chambers could be realized for FERS. To advance FERS towards these applications, further developments in instrumentation are needed to ensure good sensitivity, reproducibility, and easy handling.

All Raman techniques still face difficulties with fluorescent samples. This problem can be balanced by adapting the excitation wavelength to the sample characteristics. In the visible range, spectral windows are very specific for each analyte and matrix. Choosing one wavelength would thus restrict the universal use of a device. To avoid fluorescence signals in general, both NIR and deep-UV light could be used for typical samples in drug monitoring. However, both wavelength ranges have their specific challenges. The Raman scattering effect is intrinsically weaker in the NIR range, leading to unfavorably low signal intensities. For UV, a significant signal enhancement can be achieved, but currently available optical components for the deep-UV range are not as efficient as in the visible range.

Elaborate data processing methods are needed in all Raman spectroscopic applications in the field of pharmaceuticals. This is especially true when complex mixtures are investigated. Multivariate data analysis is often applied [129]. Advances in chemometrics and data-driven classification approaches will lead to automation and easy interpretation of results.

3.2 TDM and forensic sensing

Methods for fast on-site, highly selective, and ultra-sensitive drug sensing are sought after for clinical and forensic applications. TDM of anticancer agents and antibiotics improves the outcome of the patients. Yet, the long sample-to-result times of standard techniques hinder swift dose adaptions. Therefore, new methods are essential to achieve continuous and real-time drug monitoring at the point of care. Raman spectroscopic techniques are chemically selective and fast and can be miniaturized and applied at the point of care. However, the sensitivities required for the above-mentioned applications are challenging for conventional Raman sensing. Several enhancement strategies were successfully developed and advanced to meet this challenge.

In SERS, various advances were achieved for TDM [52], [75], [76], [79], [81], [83], [84], [97] of antibiotics [41], [75] and anti-cancer drugs [32], [35], [38], [81], [130]. Innovative SERS substrates [82], [84] represent a breakthrough towards clinical application as they overcome difficulties related to proteins in body fluids. Several approaches combine sample pretreatment methods with SERS detection [73], [74], [76], [79], [81], [86], [88], [90]. Some of the reported separation techniques result in extensive sample-to-result times (e.g. with 45 min [81] or more than 1 h [88]). These methods are not suitable for near-real-time drug monitoring. Other methods based on liquid extraction [77], filtration [59], [79], [80], [88], or purification columns [76] can be applied easily within 5 [77] to 15 min [79]. These preprocessing methods are compatible with future point-of-care use and could be automatized and included in a measurement procedure.

Complex nanophotonic technologies also show potential for the tracing of illegal or adulterant drugs [73], [74], [85], [87], [89], [131], [132], [133], [134], [135]. With the developed techniques, drugs can even be traced in fingerprints [89] or in body fluids other than blood or urine such as saliva [87], [90], [136]. D’Elia et al. recently reported about the extremely sensitive detection of cocaine in saliva via SERS, without any sample preparation [47].

There are two remaining major drawbacks that keep SERS away from widespread application: on the one hand, the complicated procedure for the synthesis of sophisticated plasmonic substrates and, on the other hand, the low reproducibility of SERS measurements (relative standard deviations about 20% are considered reasonable [72]). However, recent developments are promising. Novel substrates with shelf lives of up to 12 months [69], [72], [89] guarantee short-notice availability. Other materials achieved high reproducibility [77], [78] and low relative standard deviations (around 5% [71]). These results show that clinical or forensic standards are within reach of SERS methods.

FERS is another very promising Raman technique for continuous drug monitoring. Recent advances demonstrate the potential of FERS for TDM of antibiotics [97], [98], [101]. Different types of micro- and nano-structured sensor fibers, such as HC-PCFs [97], [98] and airclad fibers [101], [102], are developed, and their potential for clinical applications is explored. The first step towards clinical application was recently taken, as a selectively filled airclad fiber was used for the monitoring of the antibiotic cefuroxime in human urine [101].

The reviewed scientific reports demonstrate that Raman spectroscopic techniques greatly advanced in recent years. Several promising techniques for pharmaceutical drug monitoring take advantage of new nano-photonic developments, such as improved SERS substrates and tailor-made micro- and nano-structured optical fibers. Expected improvements and further steps – like the combination of different Raman techniques – will open up new fields of application, i.e. on-site or point-of-care drug monitoring in body fluids.