INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

DEVELOPMENT OF PAINLESS METHODS of measuring blood analyte concentrations has become an area of increasing interest, especially for the measurement of glucose (9). A number of different technologies have been investigated using invasive, minimally invasive, and noninvasive techniques (19). Examples of minimally invasive techniques include the use of lasers to make a small hole in the skin through which blood or interstitial fluid (ISF) can be withdrawn, lancets that penetrate only into the epidermis and extract a small sample of ISF (9), and removal of the stratum corneum (SC), followed by extraction of ISF using vacuum (8). Among the noninvasive techniques that have been studied, near-infrared spectroscopy has received much attention as a potential method for glucose sensing (5). For a detailed review of blood glucose monitoring, see Ref. 19.

Transdermal extraction of analytes offers an attractive method of noninvasive diagnostics. In this method, a sample of ISF is extracted transdermally and is subsequently analyzed for desired analytes (i.e., glucose). However, this approach is limited by the low permeability (P) of skin to glucose. The enormous barrier properties of human skin are attributed to the SC, the outermost layer of the skin (1). A variety of approaches have been suggested to enhance transdermal transport of molecules. These include 1) the use of chemicals to modify the skin structure, 2) the application of electric fields (3, 17), and 3) the application of ultrasound or sonophoresis (11, 13). Although most studies of these enhancers have been performed with the objective of transdermal drug delivery, the use of iontophoresis (20) and chemical enhancers (19) has been attempted for transdermal extraction of glucose. Our laboratory has recently shown that application of ultrasound can also be used for transdermal extraction of several analytes in humans (10). Specifically, ultrasound was used to enhance skin P and was followed by transdermal extraction of analytes using vacuum. In this paper, we present a detailed experimental and theoretical analysis of the correlation between transdermally extracted glucose and blood glucose levels.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

In vivo experiments were performed using Sprague-Dawley rats (10-16 wk of age). Rats were anesthetized with a mixture of ketamine (60 mg/kg) and xylazine (10 mg/kg), injected intraperitoneally or intramuscularly. After anesthesia was confirmed, a flanged glass cylinder (Crown Glass, diameter 15 mm, height 2 cm) was glued to the rat's shaved lateral flank using a minimal amount of cyanoacrylate adhesive (Permabond International or Vet Bond) applied to the outer edge of the flange. The chamber was filled with 2 ml of a 1% solution of sodium lauryl sulfate in phosphate-buffered saline (PBS). Ultrasound was at 20 kHz, and an intensity of 7 W/cm² was applied once for <1 min by immersing the horn (VCX 400, Sonics and Materials) into the liquid. The horn was positioned 1 cm above the skin inside the receiving chamber. The sonicators were operated in the pulsed mode (5-s on/5-s off, also referred to as 50% duty cycle). Ultrasound intensity was measured using a calorimetric method described elsewhere. The skin conductivity was measured between a subcutaneously inserted electrode (E242, Invivo Metric) and the metal sonicator horn. To measure the electrical resistivity of the skin, a 100-mV AC electric field (10 Hz) was applied across the skin for a short time using a signal generator (model HP 4116A). Current was measured with an ampmeter (Micronta, Tandy). The electrical resistance was then calculated from Ohm's law. The resistivity of skin was obtained by multiplying skin electrical resistance (measured experimentally) by skin area (1.7 cm²). Skin conductivity was calculated by taking the reciprocal of the resistivity. Ultrasound was terminated when skin conductivity reached a value of 0.6 (k-cm²)¹. At the end of sonication, the chamber contents were removed and replaced (after rinsing) with 1 ml of fresh PBS. Extraction was performed by a 15-min application of vacuum (10 in. Hg) using an Air Cadet pump (Cole-Palmer). At the end of extraction, the chamber content was collected and analyzed. Concentration of glucose (C) was measured using a Sigma assay kit (315). This kit is designed to measure physiologically relevant concentrations of analytes in serum. To increase the sensitivity of the assay, the sample-to-reagent ratio was modified to 0.4. Experiments were also performed to assess the dynamic correlation between serum C and transdermally extracted fluxes (f). In these experiments, animals were infused, via the jugular vein, with a solution of insulin (10 mU/min) to vary the blood glucose level. Extractions were performed using 5 min of vacuum (10 in. Hg) that was applied once every 20 min. Transdermal f increased with time due to structural changes in the skin and stabilized after 1 h. Transdermally extracted glucose f (measured every 20 min) collected after the 1-h stabilization period were correlated with the blood glucose values over this period, as explained in Correlation between transdermally extracted glucose and blood glucose.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Skin permeabilization and analyte extraction. Ultrasound-induced skin permeabilization was monitored using skin conductivity. Conductivity is an excellent indicator of skin P (15), because the lipid bilayers of the SC, which offer electrical resistance to the skin, retard transdermal transport of molecules. Conductivity of rat skin before ultrasound is ~0.01 (k-cm²)¹. Skin conductivity increases with application of ultrasound (20 kHz, 7 W/cm², 50% duty cycle). The increase in skin conductivity depends on the total energy (E) delivered to the rat skin [E = I, where I is ultrasound intensity (W/cm²) and  is ultrasound on time in seconds (14)]. Specifically, there exists a threshold ultrasound E (E_threshold) below which no significant change in the electrical conductivity of skin is observed. However, at energies higher than E_threshold, the electrical conductivity increases with increasing ultrasound E dose. For rat skin, E_threshold is ~1 J/cm² (12). Detailed discussion of E_threshold dose is provided in Refs. 12 and 15. Application of ultrasound was stopped when the conductivity achieved a value of ~0.6 (k-cm²)¹. Although the choice of this stopping point was arbitrary, we found that skin having this conductance allowed a 100-fold enhancement of transdermal f of various analytes. Under the typical conditions used (20 kHz, 7 W/cm², 50% duty cycle), the average application time required to reach the conductance was <1 min. Skin conductivity was maintained at an elevated state for the 3-h duration of the experiment. We measured transdermal glucose f through ultrasonically permeabilized skin by application of vacuum over this period. The measured glucose f was ~52 ± 30 µg · cm² · h¹ when the average serum C of the rat was 183 mg/dl, a f that was ~100 times higher than passive f across nontreated skin.

Correlation between transdermally extracted glucose and blood glucose. Because the objective of sonophoretic extraction is to correlate f values with blood glucose values, additional experiments were performed to assess the dynamic correlation between serum C and transdermally extracted f. In these experiments, rat skin was exposed to ultrasound, as described in MATERIALS AND METHODS. Multiple extractions were performed using vacuum (10 in. Hg for 5 min, applied every 20 min) over a period of 2 h. The first transdermal f [f₁, nmol · cm² · h¹; after an initial 1-h stabilization period] was used to calculate the calibration factor [(K_c = B₁/f₁), where B₁ is the blood glucose value (mg/dl) for each animal, measured at the same time as f₁ was extracted]. Blood glucose levels (B_i) of rats were varied by infusing insulin (Humulin, Eli Lily) intravenously at a rate of 10 mU/min for 2 h. B_i were predicted, based on K_c and the measured glucose f (f_i), as B_i = K_cf_i. K_c ratio was calculated individually for each animal. The ratios for four animals were 0.30, 0.29, 0.13, and 0.20 mg · dl¹ · nmol¹ · cm² · h, respectively. The average calibration factor was 0.23 ± 34%. The variability in the average K_c (34%) suggests that individual calibration is necessary to achieve an accurate correlation between the glucose values predicted from transdermally extracted f and reference glucose values. This is essential because the allowed error in the predictions of blood glucose in a clinical setting is <20% (4). Figure 1 shows typical variations in blood glucose levels of a rat and the corresponding changes in transdermally extracted glucose f. Transdermally extracted glucose f correlated well with the changes in the blood glucose level in the hypo- and hyperglycemic range. Figure 2 compares the predicted blood glucose values, based on the transdermally extracted f, and those directly measured in blood. The predictions are based on a one-point calibration. The relationship between the predicted and measured glucose values is linear (r = 0.97). We also assessed the relationship between transdermal glucose f and blood glucose values using an error grid (Fig. 2). Predictions in zones A and B are clinically acceptable, whereas those outside zones A and B would lead to clinically significant errors (2). All but 1 of the 26 predictions based on transdermal glucose f are in zones A and B. The accuracy at the low glucose levels is particularly important because the requirements of the accuracy in predictions in these regions are stricter than those at higher glucose levels. The predicted blood glucose values compare well with the measured values in this region. Another criterion for accuracy is the mean relative error between the reference and calculated glucose value defined by mean relative error {[absolute(serum glucose  calculated glucose)/serum glucose] × 100}. A mean relative error of 17% was obtained for all measurements (n = 26). The data presented to this point show that transdermally extracted glucose can be measured and correlates well with blood glucose values. In the next section, we present a mathematical model describing various fluid flow mechanisms that play an important role in determining the correlation between extracted f and blood glucose values.

View larger version (16K): [in this window] [in a new window]   Fig. 1.   Variation of blood glucose levels () and transdermal glucose flux () with time. Typical data for 1 rat are shown.

View larger version (24K): [in this window] [in a new window]   Fig. 2.   Relationship between reference glucose level (measured in tail vein) and glucose levels predicted using transdermally extracted glucose fluxes after 1-point calibration (n = 4). Values are presented on an error grid comprising 5 zones (zones A-E).

Formulation of the mathematical model. A schematic representation of analyte transport (Q) across skin is presented in Fig. 3. Figure 3A shows the various layers of skin involved in modeling transport during extraction, whereas Fig. 3B shows a schematic representation of various transport processes. A description of the model parameters is provided in Table 1. The general formulation of the model is based on a model developed by Reed et al. (18) for determining transcapillary protein transport. The schematic representation (Fig. 3) consists of four layers: capillaries, epidermal ISF (e), SC, and the receiving chamber. We assume that because the first layer of capillaries exists near the bottom of the epidermis, we may not need to account for Q across the dermis. Q of a given analyte (i.e., glucose, chosen as a model molecule) from the capillary layer into the receiving chamber is determined by a balance of various processes. Q from capillaries into the epidermis occurs through three pathways: 1) convective plasma leak, which occurs through large pores in the capillary wall and is insensitive to analyte size, 2) convective filtration, which occurs through small pores in the capillary wall and is size-selective, and 3) diffusion across the capillary wall (18). A fraction of the amount transported from capillaries to the epidermis is recovered through convective reflux. The remaining amount is either 1) cleared by lymph, 2) transported to the SC through convection, or 3) transported to the SC through diffusion. Similarly, the amount transported into the SC is further transported to the receiving chamber through convection or diffusion. For each Q term, there is a corresponding fluid velocity (J) term, except in the case of diffusion. Each layer is characterized by an average glucose concentration (C), pressure (P), and volume (V). The capillary layer is characterized by the arterial and venous pressures. A summary of all model parameters and their typical values is provided in Table 1. Estimation of model parameters is discussed below. Considering that the diffusive f is proportional to the C gradient and the convective f is proportional to the P gradient, this description of Q can be mathematically written as shown by Eqs. 1-15 in Table 2 (assuming steady-state conditions). The rate of accumulation of the total amount of analyte, as well as the total amount of fluid in the epidermal ISF (e) and the SC (which is equal to zero at steady state), can be described by Eqs. 16-19 in Table 2. The assumption of steady state is validated in the DISCUSSION.

View larger version (28K): [in this window] [in a new window]   Fig. 3.   A: graphic representation of various layers of skin [blood capillaries, epidermis, and stratum corneum (SC)] with extraction chamber. B: schematic representation of skin layers considered in the mathematical model. B also shows the various transport processes involved in transdermal glucose extraction. Q, transport of analytes; J, fluid velocity; V, volume; C, concentration; P, pressure. Subscripts: a, arterial; b, blood; c, convection; cham, receiving chamber; d, diffusion; e, epidermal interstitial fluid (ISF); f, filtration; l, cleared by lymph; pl, plasma leak; SC, stratum corneum; r, reflux; v, venous.

3

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Application of ultrasound significantly enhanced skin P to glucose. Glucose f measured during the extraction corresponds to an ISF extraction rate of 25.6 µl · cm² · h¹. In this respect, it is important to note that vacuum-based extraction of ISF across tape-stripped skin has been previously measured. Specifically, Kimura et al. (8) have shown that the rate of ISF extraction by application of vacuum across tape-stripped skin is ~24 µl · cm² · h¹ in rabbits and ~12 µl · cm² · h¹ in humans (6). The pressure used in their studies was 400 Torr. These numbers are similar to the ones we obtained. Specifically, the flow rates obtained in our experiments were ~25.6 and 12 µl · cm² · h¹ in rats and humans (10), respectively. The vacuum used in our experiments was 10 in. Hg. Although the rates of ISF extraction are similar in both methods, these methods differ significantly in that 1) ultrasound does not remove SC (as revealed by histology studies) and 2) skin recovers to its baseline P relatively quickly.

Skin P measured by vacuum remained within 20% of the value observed immediately after sonication over a period of 3 h. Thus these data suggest that a short pretreatment by ultrasound permeabilizes skin rapidly and elevated skin P is maintained for at least 3 h in animals. Skin P eventually recovers due to the recovery of the skin's lipid bilayers. Specifically, we have shown that application of ultrasound enhances skin P of diabetic human volunteers and P recovers to its baseline value in 20 h (10).

Predicted glucose values based on transdermally extracted glucose compared well with those measured directly in blood (Fig. 2). Similar results were obtained in the tests recently performed in human volunteers (10). Specifically, ultrasound was used to permeabilize skin of human volunteers. A short application of ultrasound permeabilized skin for ~15 h. During this period, ISF was extracted every 30 min. C_ISF was measured and compared with blood glucose values. The results showed good correlation between glucose in the ISF and in the blood. Furthermore, patients reported no pain with ultrasound application. These data demonstrate the applicability of the principles described in these studies on humans; however, further work is necessary before this method can be used in clinical applications.

We also developed a mathematical model to understand the role of various transport mechanisms in ultrasonic ISF extraction. Specifically, after the parameters from Table 1 were substituted into Eqs. 12 and 15, the predicted total outflux of glucose, at a blood glucose value of 183 mg/dl, is 38.5 µg · cm² · h¹. This value compares well with the experimentally measured value of 52 ± 30 µg · cm² · h¹. Of the predicted f, ~90% is contributed by convection and ~10% by diffusion. By substituting the appropriate parameters in Eq. 11, the predicted convective fluid flow out of the SC (J_ce), is calculated to be 18 µl · cm² · h¹. With the assumption that the contribution of convection to the experimentally measured glucose flux is 90%, the experimental flow rate of ISF is 25.6 µl · cm² · h¹ (based on the C_ISF of 183 mg/dl and total glucose flux of 52 µg · cm² · h¹). This value compares well with the predicted value of 18 µl · cm² · h¹.

We also evaluated the assumption of steady state by calculating the time constants (T) associated with glucose extraction. T of analyte extraction across the SC and the epidermis is approximately given by T = J_ce/ (V_SC + V_e), where V_SC and V_e correspond, respectively, to the volumes of the SC and the epidermis that are accessible to ISF per unit of skin area. Note that these values are less than the total volumes of the corresponding tissues, respectively. With the assumption that V_e available to the ISF corresponds to the cell free volume (~10%), V_SC + V_e is ~0.0005 cm³/cm², corresponding to a thickness of 50 µm between the epidermis and the SC (7, 16). Because J_ce is 0.018 cm/h (as predicted in Eq. 11), T is ~2 min. This time is less than the extraction time in our studies (5 min). Hence, this calculation supports our earlier assumption of the existence of steady state in our extraction experiments.

The extracted ISF was collected into a liquid reservoir with a volume of 1 ml; thus C_cham is much less than C_ISF. Because the model predicts the convective velocity of the ISF, it is now possible to predict the C_cham of various analytes, in addition to glucose, using the following equation

(20)

where  is the extraction time, A is the extraction area (cm²), and V_cham is the fluid volume (µl) in which the extracted analytes are diluted. Our laboratory previously measured ISF and C_cham of various analytes including glucose, urea, calcium, lactate, proteins, and triglycerides (12). Figure 4 shows a comparison of the prediction of Eq. 20 (shown by solid line) (A = 1.7 cm² and V_cham = 1,000 µl, C_SC = C_ISF, = 0.25 h) with the experimental data taken from Ref. 12. The predictions of the model compare favorably with the experimental data. Thus Eq. 20 can be used to predict C_cham once the C_ISF of a given analyte is known.

View larger version (15K): [in this window] [in a new window]   Fig. 4.   Comparison of chamber concentrations of various analytes predicted by Eq. 20 with experimentally measured ISF concentrations of analytes from Ref. 12.

Thus the results presented herein show that a short application of ultrasound increases skin P, so that ISF can be extracted through this skin at a rate of 25.6 µl · cm² · h¹. This rate is comparable with the theoretically predicted rate of 18 µl · cm² · h¹. Further work in this area should be focused on assessing the dynamic correlation between transdermally extracted f and blood levels of analytes other than glucose. Additional studies should also be performed to evaluate the safety of low-frequency sonophoresis and biosensor development.