Percutaneous Absorption in Man: In vitro-in vivo Correlation

Aims: To examine the existing literature to determine the degree to which percutaneous absorption data obtained using the excised human skin model match those obtained from living man. Methods: The scientific literature was reviewed to collect data on compounds whose percutaneous absorption through human skin had been measured under both in vitro and in vivo conditions. The in vitro-in vivo (IVIV) correlation was evaluated by computing the in vitro/in vivo ratio using total absorption (percent of applied dose) as the metric for comparison. Results: A total of 92 data sets were collected from 30 published studies. The average IVIV ratio across all values was 1.6, though for any single data set there could be a nearly 20-fold difference between the in vitro and in vivo values. In 85% of the cases, however, the difference was less than 3-fold. The correlation was significantly improved when data were excluded from studies in which the protocols for both studies were not fully harmonized. For harmonized data sets the average IVIV ratio was 0.96 and there was a less than 2-fold difference between the in vitro and in vivo results for any one compound, with IVIV ratios Received: May 31, 2010 Accepted after revision: January 22, 2011 Published online: March 31, 2011 Thomas J. Franz 10716 South East Forest View Lane Happy Valley, OR 97086 (USA) Tel. +1 503 698 3992 E-Mail franzcell @ frontier.com © 2011 S. Karger AG, Basel 1660–5527/11/0244–0224$38.00/0 Accessible online at: www.karger.com/spp In vitro-in vivo Correlation Skin Pharmacol Physiol 2011;24:224–230 225 Methods The published scientific literature available in English was reviewed and 30 studies identified in which the percutaneous absorption of the same compound had been measured in human skin under both in vitro and in vivo conditions [2–31] . In some cases the in vitro and in vivo studies on a given compound were conducted by the same laboratory and similar conditions used for both. However, in other cases only an in vitro study was conducted and the results compared with those available in the published literature from an in vivo study. For that reason the protocols for each of the in vitro studies were carefully reviewed to identify the extent to which they matched those utilized in living man. The following potentially critical parameters were examined: (1) skin source/anatomical region; (2) compound concentration (micrograms per milliliter) and dose (micrograms per square centimeter); (3) vehicle dose (microliters per square centimeter or milligrams per square centimeter); (4) vehicle composition; (5) chamber conditions (temperature, humidity, occluded/unoccluded); (6) length of application, protected/unprotected site, wash-off time. Two analyses were performed: (a) a comparison of the data from all studies irrespective of whether the protocols were fully harmonized, and (b) a comparison of the data from those studies in which full harmonization was documented. The in vitro-in vivo (IVIV) correlation was examined using only the data on total absorption (percent of applied dose) into the receptor solution (in vitro), ignoring residual compound contained within the skin at the time of study conclusion. The IVIV ratio for each compoundvehicle combination was calculated using the mean values for total absorption determined in the two studies being compared. In cases where there were multiple in vitro values and/or multiple in vivo values for the same compound, IVIV ratios were calculated independently for all combinations. Thus, where 9 in vitro values from a multicenter study were available for caffeine [29] , and 2 in vivo values [3, 6] , a total of 18 IVIV ratios resulted. An overall IVIV ratio was also calculated by averaging across all individual ratios. Results and Discussion All Data: Harmonized and Nonharmonized A total of 92 data sets involving 30 organic compounds were collected from 30 published studies ( table 1 ) 1 . Some of the studies contained only in vivo [2–4, 10–12, 14, 23, 26, 31] or in vitro [5, 8, 13, 17, 19–21, 24, 29] data, whereas others contained both sets of data [6, 7, 9, 15, 16, 18, 22, 25, 27, 28, 30] . In a number of cases the choice of which in vivo study to use as the comparator for the in vitro study was problematic since dissimilar vehicles were often utilized in vitro. For example, a significant body of in vitro data resides in the multicenter study of van de Sandt et al. [29] , in which the absorption of benzoic acid, caffeine and testosterone were measured from a vehicle consisting of 1: 1 ethanol/water. However, none of the studies conducted on living man utilized that same vehicle, making it necessary to choose the most logical match. In this case, in vivo data obtained using an acetone vehicle were chosen as being closer to ethanol/water than data obtained using either petrolatum, ethylene glycol gel or water gel as vehicles. An additional disqualifying factor against the use of in vivo data obtained from the water gel vehicle was its use at an infinite dose (400 mg/cm 2 ), which did not match up favorably with the 25  l/cm 2 dose used in vitro. Thus, where data were available from multiple studies in living man, the comparator data were selected on the basis of which vehicle and dose most closely matched that used in vitro. With three compounds, benzene, caffeine and testosterone, 2 in vivo studies were found equally suitable as comparators for the in vitro data, and both were used. Examination of the IVIV ratios for the entire data set of 92 demonstrates a clear tendency for the observed values to follow the line of a perfect 1: 1 correlation ( fig. 1 ). The average IVIV ratio for the group as a whole was 1.6. However, examination of the data from any individual data set reveals that variability can be great, with ratios ranging from 0.18 to 19.7. Nonetheless, although there is a single example where the in vitro value differs from in vivo by nearly 20-fold, in 85% of the data sets the difference is less than 3-fold. A review of the methods employed in the studies from which the data of table 1 were taken revealed that there were very few instances in which the in vitro protocol design perfectly matched that utilized in vivo. This lack of harmonization included differences in: (a) the anatomical site from which the skin was obtained, (b) test compound dose, (c) vehicle dose, (d) composition of vehicle, (e) length of exposure/wash time, and (f) the temperature at which the in vitro study was run as well as the use of isothermal rather than anisothermal conditions. Although the quantitative impact on absorption of some of these differences may have been trivial, two factors that unquestionably influenced the in vitro results were the use of skin from an anatomical site other than that used in vivo, as well the use of a vehicle whose composition did not match that used in vivo. Since regional variation in the permeability of skin is well documented [32, 33] , as is the ability to alter absorption through vehicle modification [34–38] , it is evident that substantial error may exist within the presently available data sets. 1 A more complete table listing essential parameters of individual in vitro and in vivo studies is available for electronic download on request.