Department of Chemical and Materials Engineering, University of Calabria, Arcavacata di Rende (CS), Italy

Transcription

1 European Journal of Parenteral Sciences 2000; 5(3): Parenteral Society Automated integrity testing of hydrophobic filters based on water intrusion measurements: comparative analysis of a refilling continuousflow and a pressure-decay batch device Peter Czermak 1, Gerardo Catapano 2 1 Institute of Biochemical Engineering and Membrane Technology, Department KMUB - Biotechnology, University of Applied Sciences Giessen and Biotechnologie Gesellschaft Mittelhessen mbh, Giessen, Germany 2 Department of Chemical and Materials Engineering, University of Calabria, Arcavacata di Rende (CS), Italy The water intrusion test (WIT) has become a routine procedure in the pharmaceutical industry as a tool for the in situ integrity testing of hydrophobic filters. During the test, the filter housing is submerged with water and automated devices provide an estimate of the evaporated water flow rate across the filter, i.e., water intrusion rate (WIR). WIR is generally correlated to the filter s bacterial retention rate. Commercially available WIT devices differ with regard to the technique used to measure WIR. In this paper, we investigated the capacity of a pressure-decay batch device and a refilling continuous-flow device to provide reliable measurements of WIR at various WIRs and initial gas volumes above the filter. Experiments were performed with a model filter system simulating typical conditions of WITs and with a full-scale commercial filter. The pressure-decay batch device consistently underestimated the actual WIR. The error increased with decreasing initial gas volumes above the filter and with increasing actual WIRs. Measured WIR values were as much as 25% and 31% lower than the actual WIR for the model filter system and the full-scale filter, respectively. The refilling continuous-flow device yielded ures in excellent agreement with the actual WIRs with both the model filter system and the full-scale filter, independently of the WIR value and the initial gas volume above the filter. Der Wasser-Intrusions-Test (WIT) hat sich in der pharmazeutischen Industrie als Routinetest zur in-situ Integritätsprüfung hydrophober Filter etabliert. Während des Testes wird das Filtergehäuse mit Wasser gefüllt und automatisierte Geräte liefern eine Bestimmung des Wasserdampfflusses über den Filter, die Wasser-Intrusions-Rate (WIR). Die WIR wird im allgemeinen mit der Bakterien- Rückhalterate des Filters korreliert. Kommerziell erhältliche Geräte zur Durchführung des WIT unterscheiden sich in der Technik, um die WIR zu messen. In der vorliegenden Veröffentlichung wurde die Leistungsfähigkeit eines Druckabfall-Batch- Systems und eines kontinuierlichen Nachfüll-Systems zur zuverlässigen Messung der WIR bei verschiedenen WIRs und Anfangsgasvolumen über dem Filter untersucht. Es wurden Experimente mit einem Modell-Filter-System, mit dem die typischen Bedingungen des WIT simuliert werden konnten, und mit einem kommerziellen Filter durchgeführt. Das Druckabfall-Batch-Gerät unterschätzte konsequent den tatsächlichen WIR. Der Fehler stieg mit geringer werdenden Anfangsgasvolumen über dem Filter und mit steigenden tatsächlichen WIRs an. WIR Messungen waren mehr als 25% bzw. 31% niedriger als der tatsächliche WIR für das Modell-Filter-System und für den kommerziellen Filter. WIR Messungen mit dem kontinuierlichen Nachfüll-System lieferten sowohl für das Modell-Filter-System als auch für den kommerziellen Filter Ergebnisse, die in exzellenter Übereinstimmung mit den tatsächlichen WIRs waren, unabhängig vom WIR Wert und dem Anfangsgasvolumen über dem Filter. Corresponding author: Prof. Dr.-Ing. Peter Czermak, University of Applied Sciences, Institute of Biochemical Engineering and Membrane Technology, Department KMUB Biotechnology, Wiesenstrasse 14, D Giessen, Germany Tel: Fax: peter.czermak@tg.fh-giessen.de 59

2 60 P CZERMAK AND G CATAPANO Introduction The water intrusion test (WIT) has become an established procedure in the pharmaceutical industry as a tool for the in situ integrity testing of hydrophobic gas filters. In WITs, the filter housing already assembled in the process line is submerged with water and subjected to a pressure higher than atmospheric but lower than the water breakthrough pressure for the given filter. Membrane filter hydrophobicity prevents liquid water from permeating its pores. Instead, water vapour crosses the filter at a rate that initially decreases in time as a result of effects 1 such as compaction of membrane pleats and elimination of air trapped in the filter membrane. After this transient phase (often termed stabilisation phase), evaporated water crosses the filter at a constant flow rate whose actual value depends on the filter surface area and on operating conditions 1,2 (e.g., pressure, temperature, water conductivity, etc.). The evaporated water flow rate, or water intrusion rate (WIR), across the filter is generally correlated to its bacterial retention rate 2, WIRs lower than a filter-dependent set value ensuring successful bacterial retention rate 3. Today, WITs are routinely performed for the integrity testing of sterile filters prior to the evacuation of decontaminated autoclaves (to minimise the system's environmental impact), on standard sterilisers and on freeze dryers (after sterilisation but prior to freeze drying), among others. A few automated devices are commercially available to perform WITs that differ in the technique used to estimate the evaporated water flow rate across the filter. Two out of three available commercial devices are batch operated 4,5. In fact, the pressurising gas line is shut off after the set pressure is established on the liquid water above the filter. From this moment on, gas pressure in the filter housing is recorded in time and its decay is used to estimate the evaporated water flow rate. In the first type of device 4, during the stabilisation phase, the high-pressure gas is allowed to expand into a low-pressure chamber of known volume to permit evaluation of the initial gas volume according to the ideal gas law. At the end of the stabilisation phase, the device records gas pressure decay as water evaporates and automatically estimates the water intrusion rate. In the second type of device, gas volume (hence, gas pressure) above the filter is held constant as water vapour crosses the filter by continuously feeding the filter housing with liquid water from an external tank until gas pressure stabilises (Figure 1). When gas pressure is stable, the external tank is disconnected and the device starts recording gas pressure. WIR is estimated from gas pressure decay over a fixed time interval in agreement with the ideal gas law 6,7. The gas volume needed for calculating WIR from gas pressure decay in time is estimated only once, when the system is set in place, and relies on a reproducible filling procedure for the subsequent measurements. In the only refilling continuous-flow commercial device, the filter is first submerged with pressurised water. The device then continuously feeds the filter housing with volumes of gas sufficient to hold gas pressure above the water level constant and to compensate for water vapour losses through the filter (Figure 2). Test device Tank Figure 1. Schematic of a batch automated device equipped with an external liquid water tank for water intrusion rate measurements based on gas pressure decay in time In this device, the actual gas volume flow rate fed to the housing provides a direct measure of the evaporated water flow rate across the filter throughout the entire test sequence 8. In this paper, we report on an investigation aimed at assessing the capacity of commercial batch and continuous-flow devices for WITs to provide reliable estimates of the evaporated water flow rate across filters of different sizes, in the WIR range typical of intact filters. To this purpose, a model filter system was developed which simulates the continuous and stable loss of water from a water reservoir pressurised with gas, at levels encountered during WITs. The system also permits the independent verification of the water flow rates estimated by the devices used. Materials and methods Pressure transducer Gas volume Filter Model filter system The model filter system consists of a vertical cylindrical tube connected at the bottom to a peristaltic pump (Amersham Pharmacia Biotech, Freiburg, Germany) via a stainless steel needle valve (Nupro Co., Willoughby, Ohio, USA). The former ensures continuous withdrawal of water from the tube. The latter minimises the pressure fluctuations on water delivery typical of volumetric pumps. Pressure-resistant tubing connects the top of the tube to a pressurised gas outlet via the automated WIT devices that were tested, as shown in Figure 3. Prior to the experiments, 100 ml of deionised water was poured into the tube and the tube was connected to the tested automated device. Gas volume above the liquid level, V, was varied by inserting additional tubing between the tube and the automated device being tested. Methods Each WIT was initiated by setting pump and needle valve to deliver a given water flow rate at the chosen operating pressure and by applying pressure to the model filter system. The actual rate at which water left the tube under given conditions (i.e., WIR) was measured by weighing

3 AUTOMATED INTEGRITY TESTING OF HYDROPHOBIC FILTERS BASED ON WATER INTRUSION MEASUREMENTS 61 Gas Filter membrane flow = V PT Flow measuring system Figure 2. Schematic of a refilling, continuous-flow automated device for water intrusion rate measurements (PT = pressure transducer) Test device, P 1, P 2 V Pressure supply PT P 0 > P 1 P 1 = P 2 = const flow rates (i.e., actual WIRs) were averaged over time, as carried out by the control software of each device, and were compared to the WIRs measured by the automated device being tested. WITs were performed with the model filter system by using a batch-operated Sartocheck 3 device (Sartorius AG, Goettingen, Germany) 4 and a continuous-flow Palltronic Flowstar FFS02 device (Pall Europe Ltd, Portsmouth, UK) 8. In the former device, water supply during gas pressure stabilisation was kept disabled at all times. Both devices were calibrated by the manufacturer, the former one month, the latter six months before use. Experiments were performed at room temperature, 2,500 mbar pressure and with initial volumes of gas above the water level, V o, ranging from 75 to 275 ml. flow rate ranged from 0.06 to 1 to simulate the WIRs exhibited by intact 10-inch filter cartridges. Deionised water was used throughout. The two automated commercial devices were also used to perform WITs on full-scale Sartofluor GA filters (Sartorius AG, Goettingen, Germany). Prior to the test, the filter was installed in the housing, the upstream volume was filled with deionised water and the housing was connected to the automated device under test. Experiments were performed at room temperature, 2,500 mbar pressure and with initial volumes of gas above the water level, V o, of 148 and 248 ml. The statistical significance of WIR differences under different conditions was determined using Student's t-test at a significance level of p=0.05. Figure 3. Schematic of the model filter system Variable gas volume 0.012g Scale the mass of water collected in a given time interval in a beaker placed on a scale (Mettler-Toledo GmbH, Giessen, Germany) connected to a computer. Steady state water Results Model filter system Figure 4 shows that the pressure-decay batch automated device consistently underestimated the actual WIR. The extent to which measured ( ) and actual (WIR actual ) WIRs disagree depends on the actual volume of gas above the liquid level V o and WIR. Performing WITs at high V o apparently reduced the error in the measured WIR over the whole range of WIR investigated. At a given V o, measured WIR values were affected by an error whose magnitude increased with increasing actual WIRs and reached 25% at WIR close to 0.9 and V o =75 ml. Figure 5 shows that the refilling continuous-flow automated device consistently yielded ures in excellent agreement with the actual WIRs, independently of the WIR value and volume of the gas above the liquid level. Full-scale filter Figure 6 shows that WITs performed on full-scale filters with the pressure-decay batch automated device yielded measured WIR values that increased by about 31% as V o increased from 148 to 248 ml. WITs performed with the refilling continuous-flow automated device consistently yielded the same WIR. Differences between WIR values measured in this way and those measured by the pressuredecay batch device at V o = 248 ml were statistically insignificant.

4 62 P CZERMAK AND G CATAPANO WIR actual Discussion The first steps of WITs performed with batch devices generally consist of filling the housing with water until the filter is completely submerged and in pressurising the gas above the water level. As the set pressure is established, the gas line is shut off. From this time on, the gas molecules trapped between the water surface and the closed valve cannot escape. They form a batch system (generally isothermal) whose volume increases as water vapour leaves the filter housing across the membrane. As the gas expands into this larger volume, pressure decreases in a fashion that can be described in terms of the ideal gas law, which states that at constant temperature: 1. p 1 = p 2 Figure 4. Comparison of actual WIRs with those estimated by a pressure-decay, batch automated device with the model filter system at increasing V o : ( ) 75ml; ( ) 175 ml; ( ) 275 ml. The dashed line (---) at unity is provided for reference Gas volume increase is readily obtained from equation 1 as follows: WIR actual Figure 5. Comparison of actual WIRs with those estimated by a refilling, continuous-flow automated device with the model filter system at increasing V o : ( ) 75 ml; ( ) 175 ml; ( ) 275 ml. The dashed line (---) at unity is provided for reference 2. V = (p 1 p 2 ) p2 where, in particular, (p 1, ) and (p 2, ) may be taken to be the pressure and volume of the gas above the water level at the beginning and at the end of the WIR measurement phase, respectively, when water vapour flow rate is constant. Under such conditions, mass balances around the gas and the water volume upstream from the filter in the housing, coupled with equation 2, give: dv V p 1 p 2 3. = = = WIR dt t t p 2 where WIR is the flow rate at which water vapour crosses the filter membrane, as noted above. The batch device used for this investigation automatically measures WIR from pressure decay over a ten-minute interval at the end of the stabilisation phase according to the following approximate relationship: Initial gas volume V 0 ml Figure 6. WIRs in a full-scale filter measured at increasing Vo using commercial automated devices: (unshaded bars) refilling continuousflow device; (shaded bars) pressure-decay batch device; () indicates statistically significant differences at p= WIR = (p 1 p 2 ) V m t p atm where p atm is the atmospheric pressure. V m is the gas volume above the water level measured after gas pressure stabilises at the value set for the test and is often considered constant for a given family of filters 4,6. Equation 3 shows that WIR should be properly estimated with reference to gas pressure at the end (i.e., p 2 ) and gas volume at the beginning of urement (i.e., ). The algorithm of the batch device refers to V m, possibly because variations in gas volume are attributed only to membrane pleat compression 6. In such a case, V m would indeed be quite close to. However, water evaporation across the filter membrane until the measurement phase begins adds to volume displacement caused by pleat compression. This may result in an increase in gas volume with respect to V m and a reduction in gas pressure during the whole test, which are not negligible, and whose extent

5 AUTOMATED INTEGRITY TESTING OF HYDROPHOBIC FILTERS BASED ON WATER INTRUSION MEASUREMENTS 63 depends on the actual filter WIR. Higher WIRs make greater than V m and p 2 closer to p atm. Both effects suggest that equation 4 yields estimates of WIR generally lower than its actual value, with an approximation that becomes worse as filter WIR increases. Indeed, Figure 4 shows that experimental (i.e., actual) WIRs were not in good agreement with measurements based on equation 4. The batch device yielded urements that are strongly dependent on the WIR value and on the actual gas volume at the start of the test, V o. Measured WIR becomes more accurate as V o increases. In fact, as V o increases, the volume changes in the gas above the water level, caused by the continuous loss of evaporated water across the filter membrane, and the consequent pressure decay, become smaller, less significant fractions of the initial gas volume and pressure, respectively. Dealing with V m as a constant for a given filter family might even add further bias to the measured WIR. In such cases, differences in the actual volume of individual filters would be likely to affect the measured WIRs in an unpredictable fashion. The continuous-flow device that we tested does not require any pressure-related evaluation of gas volume in the cartridge but rather permits a more direct measurement of WIR. Figure 5 shows that its control system successfully caught up with transient gas volume changes, ensuring a reliable match of the volume of evaporated water with that of the feed gas. As a result, the device yielded measured WIR values which were in excellent agreement with the actual evaporated water flow rates over the whole range of V o and WIR investigated. urements in a full-scale commercial filter were consistent with the results obtained with the model filter system (Figure 6). The batch device yielded WIR measurements strongly dependent on the actual V o that became consistent with those obtained with the continuous-flow device only when WITs were performed at the higher V o. These tests confirm that the continuousflow device provides consistent and reliable WIR measurements independently of the actual operating conditions. Conclusions The results reported in this paper suggest that the commercial batch device for WITs tested in this investigation is not entirely reliable and safe and should be used with caution. It provides urements that depend on the properties of the filter tested (i.e., its actual WIR) as well as on housing design and the particular setup used for the WIT (i.e., V o ). More importantly, the device generally underestimates the actual filter WIR by as much as 31%. This may occasionally lead to the overestimation of the bacterial retention properties of a filter, thus making validation of filters questionable. We conclude that safer WITs can be performed with the continuous-flow device used in this investigation. The device provides direct and reliable urements as long as the feed gas flow causes measurable pressure changes of the gas in the filter housing. However, the high sensitivity of the device calls for a strict control of environmental conditions during the WIT test. Acknowledgements The authors wish to express their gratitude to Heiko Bender and Uwe Schreck of Biotechnologie Gesellschaft Mittelhessen for their assistance in the experimental part of the investigations. Abbreviations p gas pressure, mbar t time, min V volume of gas above the water level, ml Subscripts atm atmospheric m value estimated after gas pressure stabilises at a set value o value estimated at the beginning of the test 1 value taken at the beginning of the WIR measurement phase 2 value taken at the end of the urement phase References 1. Jaenchen R, Schubert J, Jafari S, West A. Studies on the theoretical basis of the water intrusion test (WIT). Eur J Parenteral Sciences 1997; 2(2): Dosmar M, Wolber P, Bracht K, Troeger H, Waibel P. The water pressure integrity test A new integrity test for hydrophobic membrane filters. J Parenteral Science and Technology 1990; 46(4): The Pall Intrusion Test for Integrity Testing Sterile Gas Filters. Pall Publication STR1603, Installation and Operating Instructions Sartocheck 3. Sartorius AG SM16286, Operating and Maintenance Manual Integritest Exacta Instrument. Millipore Corporation Lit. No. XITXSP121, Gehne, I. Verfahren und Einrichtung zum Prüfen von Filterelementen durch einen Wasserintrusionstest. German Patent DE C 1, Bardat A, Chatenet E, Schmitthaeusler R. Automatic integrity testing of autoclave and freeze dryer inlet filters. Pharmaceutical Eng 1997; Nov/Dec: Palltronic Flowstar Integrity Test System Operating Instruction, Model FFS02. Pall Europe Ltd. USD 2005, 2000.