EBE Psitin Paper Draft: Final: Authr: EBE BiManufacturing WG Date: 13/09/2016 EBE Psitin Paper: A Risk-Based Apprach t Setting Sterile Filtratin Biburden Limits Executive Summary The ratinale behind the EMA recmmended biburden limit befre sterile filtratin f NMT 10 CFU / 100 ml (EMA Guidelines f 1996 (1), 2012 (2) and Draft EMA Guideline 2016 (3)) has n clear rigin. The limit has been taken frm the pharmacpeial specificatin fr water fr injectin t prduce bulk, but has n scientific basis when applied t drug prduct. Overall, we wuld instead prpse a "Risk-based Apprach" t biburden cntrl which justifies alternative biburden specificatins that take the prduct manufacturing prcess int cnsideratin, in cntrast t defaulting t the histrical 10 CFU / 100 ml limit. This risk-based apprach cnsiders the tw main risks at the sterile filtratin step f drug prduct manufacture: (a) capability f the micrbilgical biburden methd and, (b) micrbial breakthrugh during sterile filtratin. These risks have been assessed fr their interactin by means f statistics and pssible risk-mitigating measures are described. Table f Cntents 1. Risk: the biburden determinatin methd... 2 2. Risk: Bacteria breakthrugh during sterile filtratin... 2 3. Cmbinatin f bth main risks... 3 4. The hlistic view cnsidering a prtfli f risk-mitigating measures... 4 5. Cnclusins... 6 6. References... 7 7. Addendum... 7 Page 1 f 12
1. Risk: the biburden determinatin methd As Jrnitz et al. (5) described, it is impssible due t the inherent variability f micrbilgical analyses, t determine an exact biburden in the rder f 10 CFU / 100 ml. The true biburden culd likely be belw r abve and thus the batch incrrectly accepted r falsely rejected. It is therefre mre accurate t specify the acceptable limit f 10 CFU plus a margin f errr. This inaccuracy in micrbilgical analyses is taken int accunt in USP Chapter <61> and als in the Ph.Eur. (Chapters 2.6.12 and 5.1.4): an acceptance criterin fr micrbilgical quality f 10 CFU wuld therefre be crrespnding t a maximum acceptable cunt f 20 CFU when the variability f analysis is cnsidered. We have mdeled the sensitivity f the Ph.Eur. 2.6.12. biburden assay methd fr an acceptance criterin f 10 CFU / 100 ml (6) while varying the test sample vlume. The Pissn mdel assumes a unifrm distributin f the bacteria in the slutin prir t sterile filtratin and is therefre nt cnsidered a suitable mdel. Bacteria can clump tgether t prduce a nn-unifrm dispersin which leads t either verestimatin r underestimatin f detected biburden when determined by the number f cunts in the slutin. This ver-dispersin can be described by a negative binmial distributin with a crrespnding prbability density functin and an ver-dispersin factr. Based n a negative binmial mdel with an ver-dispersin factr f 2, it can be estimated that using a limit f 10 CFU / 100 ml, a slutin with a true biburden level f 10 CFU / 100 ml wuld have a 41.2% prbability f batch rejectin, althugh the specificatin wuld still be met. Cnversely, a true biburden f 11 CFU / 100 ml wuld still lead t a prbability f batch acceptance f 50% (see figure 2, Addendum). The prbability curve als demnstrates that - given an acceptable risk bund f 5% - a batch passing with a measured biburden f nt greater than 10 CFU/100 ml culd reveal a true biburden f up t 20 CFU/100 ml (fr a 100 ml test sampling plan). The prbability curve shifts t the right as the vlume tested decreases such that a 10 ml test vlume may have up t 63 CFU/100 ml actual biburden fr the same 5% prbability f false negatively passing and the same cntrl limit. This increase in acceptable actual biburden can be cmpensated by adjusting parameters fr Risk 2, as described in the next sectin. We cnclude that the specificatin f the biburden at the level 10 CFU / 100 ml is t tight, leading t unreasnably high rejectin prbability f acceptable batches. Furthermre, the micrbilgical quality f the slutin prir t sterile filtratin shuld be cnsidered hlistically and the ttal risk cmpsed f many individual risks shuld be cnsidered in rder t avid falsely accepting unacceptable batches due t the limited sensitivity f the test. 2. Risk: Bacteria breakthrugh during sterile filtratin Accrding t the FDA guideline (4) and industry standards, filters used fr the final filtratin shuld be validated t reprducibly remve micrrganisms frm a carrier slutin cntaining biburden f a high cncentratin f at least 10 7 CFU/cm 2 f effective filter area (EFA). The validatin shuld be cnducted under the wrst-case prductin cnditins fr the material t be filtered, and challenge experiments shuld result in n passage f the challenge micrrganism. Thus the retentin capacity f a validated sterilizing-grade filter with an EFA f A (cm 2 ) is at least A x 10 7 CFU. Hwever, the currently used 0.22 µm sterile filter membranes can withhld much higher micrbial challenges (unpublished cmpany results fr 0.22 µm plyvinylidene fluride membranes a bacterial challenge cncentratin as high as 10 9 CFU/cm 2 can be validated). In practice, accumulatin f biburden n the final filter might even cause partial clgging. As a result, a Page 2 f 12
biburden CFU that reaches the filter early in the prcess may have a higher prbability t penetrate the filter than thse that reach a partly clgged filter later in the prcess. Therefre it is cnservative t assume that all biburden CFUs in the slutin have the same prbability t g thrugh the final filter. Similar t terminal sterilizatin methds, fr sterile filtratin a Sterility Assurance Level (SAL = degree f assurance with which the prcess renders a ppulatin f items sterile) can be calculated based n the bacteria retentin capacity f the filter. Here, wrst case assumptins have been made, fr example during the validatin using a small micrrganism (typically Brevundimnas diminuta). Since the pre-sterile filtratin biburden is determined fr each lt and nt nly as part f a validatin prcedure, the SAL can als be calculated as a LOT-SPECIFIC parameter frm the actual, measured biburden lad (with a factr included fr variability), the applied ttal filter area and the validated effective bacteria retentin capability. Thus, the decisive factr f pre-filtratin biburden is knwn (within assay variability) and nt assumed. Fr prducts subjected t terminal sterilizatin, the true biburden is nt determined (the Draft EMA Guideline n the Sterilizatin f the Medicinal Prduct, Active Substance, Excipient and Primary Cntainer suggests a maximum biburden limit f 100 CFU / 100 ml r 100 g befre) as the manufacturing prcess des nt take place under aseptic cnditins, and this must be cmpensated by incrprating apprpriate safety factrs. Since the filling prcess is run under class C cnditins, a relatively high biburden must be assumed which then has t be killed by the terminal sterilizatin prcedure. In cntrast, the sterile filtratin prcess prir t aseptic filling can be cnsidered a clsed system, fllwed by manufacture under class A cnditins. Finally, the bacteria are mechanically retained instead f being inactivated, and therefre d nt enter the final prduct. In summary, fr sterilizatin f prduct thrugh sterile filters SAL values lwer than used fr terminal sterilizatin (<10-6 ) shuld be acceptable e.g. 10-4 r 10-5. 3. Cmbinatin f bth main risks The tw types f main risks assciated with the sterile filtratin prcess can be statistically described (6): Risk 1 ( pre-sterile filtratin risk ): risk due t biburden test methd insensitivity (risk f false negative = passing a batch with unacceptable biburden), i.e. the drug slutin with an unacceptable biburden level befre sterile filtratin passes the pre-sterile filtratin biburden test, either due t inherent test methd variability r a sampling methd that des nt have sufficient statistical pwer t detect drug slutins with unacceptable levels f biburden Risk 2 ( pst-sterile filtratin risk ): risk f breakthrugh f biburden thrugh the final sterile filter with 1 CFU entering the sterile filtered slutin due t an inapprpriate prcess, i.e. prcess-related risks and micrbial breach acrss sterile filter These tw risks are inter-dependent as a high pre-sterile filtratin risk wuld require a mre stringent cntrl f the pst-sterile filtratin risk and vice versa. Therefre an effective verall cntrl strategy shuld take int accunt this inter-crrelatin. Yang et al (6) linked maximum batch size, sample vlumes f 10 ml, 30 ml and 100 ml (negative binmial mdel, specificatin limit 10 CFU / 100 ml and acceptance limit f pre-filtratin biburden 1 CFU/10 ml, 3 CFU/30 ml, 10 CFU/100 ml) t the Risk 1 prbability f incrrectly accepting a batch (5%, 1% and Page 3 f 12
0.1%) and the Risk 2 prbability f a breakthrugh f 1 CFU in ne f 10,000 cases (0.01% prbability, 10-4 ) r 1 f 100,000 cases (0.001% prbability, 10-5 ). A validated bacterial retentin capability f the sterile filter f 10 7 CFU per cm 2 was used, fr a 1000 cm 2 filter, t calculate the maximum vlume t be sterile filtered. A mdified versin f the risks-linking table is attached t this psitin paper (see table 1, Addendum). By example, a 10 ml sample size using an actin limit f 1 CFU/10 ml, with a 5% risk f passing a batch thrugh a false negative result and a filter breakthrugh risk f 10-4, allws a maximum filtratin batch vlume f 424 L when using a 1000 cm 2 (0.1 m 2 ) filter. Nte that when e.g. δ 0 = 5% (pre-filtratin risk, i.e. risk f biburden testing insensitivity) and δ = 10-4 (pst-filtratin risk, i.e. breakthrugh risk f biburden thrugh the sterile filter), there is a 95% prbability that the pre-sterile filtratin biburden test (fr any given test scheme) will successfully detect unacceptable levels f biburden and nly 1 batch ut f 10,000 wuld have biburden breakthrugh (see Table 1). This implies that in a facility prducing 100 batches a year, this event wuld nly ccur every 100 years. By limiting the maximum filtered batch size, accepted pst-sterile filtratin risk levels can achieve the same level f sterility quality assurance, fr any given pst-filtratin risk (δ), as the biburden testing with 100 ml samples and a 10 CFU/100 ml acceptance limit. Indeed, it is prpsed that a Risk 1 prbability f 5% shuld be acceptable. Alternatively, t manipulating the filtratin batch size, ther parameters may als be varied t reduce risk, e.g. increasing filter area r validated maximum CFU. The relatinship between sample vlume and the batch size varies accrding t filter area, as illustrated in Figure 3 f the Addendum. A 10 ml sample size using a 1000 cm 2 filter allws up t 424 L t be filtered t retain a 5% prbability f passing a batch with biburden exceeding the acceptance limit and filter penetratin f less than 1 in 10,000 batches. Fr a 2000 cm 2 filter the filtered vlume culd be increased t 849 L at the same risk level (see table 1). The relatinship between filtratin batch size and filter area is tabulated and prvides the Spnsr with the pprtunity t select the sterile filtratin, minimum filter area accrding t filtratin batch size t prvide a desired level fr Risk 1 (0.1% t 5%) and Risk 2 (10-4 t 10-5 ). Therefre, fr a given risk bund, the pre-filtratin risk can be cntrlled with a prperly selected batch size r the pst-filtratin risk can be cntrlled thrugh filter area and filter validatin. The biburden level D 0 may be even greater than 10 CFU/100 ml since its impact n the final pst-filtratin risk can be cntrlled thrugh placing a limit n the size f a batch fr the sterile filtratin S 0 r adjusting the filter area r the validated range s as t ensure the ttal biburden in the batch wuld nt exceed the retentin capacity f the final sterilizing filter. The presented calculatins supprt use f biburden test sample vlumes 10 ml using a filter area f 1000 cm 2, a maximum filtratin batch size f 400 L and an actin limit f up t 2 CFU/10 ml that includes a 3-fld safety factr fr the biburden determinatin precisin (calculated frm 63 CFU / 100 ml). These limits ensure a pre-filtratin biburden risk f 5% and a pst-filtratin risk f 10-4. Further risk mitigatin is built int the prduct manufacturing prcess as summarized in the next sectin. 4. The hlistic view cnsidering a prtfli f risk-mitigating measures Mitigating risk factrs are essential fr the final biburden risk assessment f the sterile filtratin prcess. This includes examinatin f the related prcess steps and understanding f the prduct characteristics and ptential prduct related risk factrs like viscsity, ph r grwth prmting prperties f the frmulatin. Fr example, if the frmulated prduct has grwth prmting prperties, this cnstitutes a risk factr per se. It is als imprtant t evaluate available biburden data, including the manufacturing facility histrical Page 4 f 12
biburden detectin trends fr the prduct and acrss prducts when a multi-prduct site is used. Histrical biburden trend data are ften indicative f the effectiveness f the verall biburden cntrl, and need t be cnsidered in identificatin f risk factrs. The fcus f this paper is n parenteral bitech drugs that cannt be terminally sterilized in the final cntainer because they are therm-labile. Fr such prducts, hwever, the testing f the biburden prir t sterile filtratin and the subsequent sterile filtratin are nt the nly measures t cntrl the micrbilgical status f the slutins fr cntainer filling (see figure 2, addendum): The upstream and dwnstream prcesses are likely t have multiple 0.2 micrn r nan-filtratin steps fr reductin f adventitius micrbial, particulates r viral cntaminatin. The fermentatin itself takes place in clsed systems subject t a CIP / SIP prcess and under the additin f filtered r HTST-treated r even sterile media. The final frmulatin steps are carried ut in a Grade C envirnment and als use steamdecntaminated r sterilized cntainers int which the slutin is being transferred via 0.2 micrn filter. In the event f extended strage f the frmulated slutin, strage is usually at lw temperatures, typically -20, -40 r -80 C. In the frzen state bacterial grwth is extremely unlikely. Interim strage in the liquid state is time-limited fr reasns f prduct stability alne and perfrmed mstly als at reduced temperatures (2 8 C), further bacterial grwth is hereby als limited. Strage and hld times f bulk after final frmulatin r thawing are evaluated in studies and eventually validated. Rm temperature strage is kept at a minimum. Mnitring the cntainer integrity during such interim strage is ften dne by superimpsing pressure (nitrgen). Furthermre, interim bulk strage after thawing may be (re-)validated using grwth-prmting media in rder t uncver micrbilgical weaknesses (as part f the regular media fills). Increasingly, upstream, dwnstream and fill/finish prcessing use pre-sterilized single-use materials and aseptic cnnectrs ( dispsables prcessing cmpnents ). These are ften surced preassembled as maniflds, pre-sterilized and aseptically cnnected. In additin, these cmpnents represent clsed systems, which reduce the risk f biburden entering by nn-aseptic cnnectins. Thrugh increasing autmatin f manufacturing prcesses, individual persnnel errrs are further mitigated. The cmmercially available sterile filters have a larger bacteria retentin capacity than the underlying minimum expectatin f 10 7 CFU / cm 2. Fr early clinical phase manufacture, the rati between filter surface area and batch size is typically relatively high cmpared t cmmercial scale manufacture. Endtxin cntent (in case e.g. pyrgens are prduced by biburden) and sterility testing is part f the Drug Prduct release prcess An analysis f data frm the cmpanies invlved in this paper has shwn that it is a very rare event t find biburden ver the limit set by the EMA befre the sterile filtratin. Often it is an artifact f sampling r in the analysis, but bth are each difficult t prve with 100% certainty, s finally the batch is ften still discarded which is very cstly and may endanger the prduct develpment timelines. Each biburden fund, even if belw the limit f 10 CFU / 100 ml may trigger identificatin f the micrbial cntaminant, s that regardless f a limit set, investigatin and crrective measures are initiated. Page 5 f 12
Fr this reasn, the authrs are f the pinin that the prcessing f bitechnlgical active ingredients is typically under ample cntrl fr biburden, which allws the determinatin and justificatin f lwer sample vlumes and wider pre-defined levels f biburden befre sterile filtratin, using risk-based strategies. 5. Cnclusins Bitechnlgy-derived drug prducts are typically labile t methds f terminal sterilizatin such as irradiatin, chemical treatment r autclaving, and are manufactured thrugh aseptic prcessing which typically emplys sterile filtratin t remve micrbial cntaminants. High biburden in the drug slutin prir t sterile filtratin may increase the chance f breaching the sterilizing filter and cause prduct safety and quality issues. As a result, pre-filtratin biburden test is recmmended by regulatry guidelines. The current EMA guidelines stipulate that a maximum acceptable biburden level must be established befre the sterile filtratin step and that in general n mre than 10 CFU/100 ml wuld be acceptable. Hwever, a sample vlume f 100 ml ften represents a significant prprtin f the batch and a high cst fr bitech prducts, particularly taking int accunt that sufficient material wuld typically be drawn fr replicate testing, assay cntrls, verage and reserve samples. Therefre it is desirable t explre alternate sample vlumes and test methds that warrant the same r higher level f quality assurance as the EMA-recmmended methd. The guidelines allw the Spnsr t justify alternative sampling prcedures thugh n guidance is prvided n what might cnstitute an acceptable justificatin. Furthermre, the EMA guidelines d nt prvide scientific ratinale fr limit 10 CFU/100 ml, thereby cmpunding the difficulty t justify a test vlume smaller than 100 ml r ther limits. In this paper, a risk-based methd is prpsed t prvide a strategy and scientific methdlgy fr justifying alternate (smaller) biburden test vlumes than 100 ml and alternate specificatins (>10 CFU/100 ml r >1CFU/10 ml).. By mdeling biburden in the slutin prir t sterile filtratin, the relatinship(s) between pre-sterile filtratin risk (prbability f detecting biburden in the slutin prir t sterile filtratin) and pst-sterile filtratin risk (prbability f biburden breaching the filter), their cmbined risk as well as the assciated risk factrs such as test sample vlume r batch size are established. Such relatinships allw fr quantitative evaluatin f the impact f the risk factrs n the develpment f effective risk-based cntrl strategies, which include acceptable selectin f sample test vlume and maximum pre-filtratin biburden level. The risk-based methd is in accrdance with the quality by design (QbD) principles in ICH Q8 that enable manufacturers t define a manufacturing prcess design space that cnsistently prduces high-quality prducts thrugh increased understanding and knwledge f the prduct and prcess. It is als cnsistent with ICH quality initiative, Quality Risk Management, which achieves prduct greater quality assurance thrugh risk identificatin, analysis, and cntrl. The risk-based apprach als justifies use f actin limits fr cntrl f the pre-sterile filtratin biburden limit thrugh implementatin f an effective Quality Management System (QMS) that describes the actins fllwed in the eventuality f any detected biburden. The cnditins that initiate investigatin f biburden, determinatin f likely rt cause and any resulting CAPA shuld be available t the agency and may be verified thrugh GMP inspectin. Spnsrs may als emply tighter internal criteria per the QMS. Ultimately, the agency shuld be assured that any batch with a significant biburden breach ptentially impacting prduct quality r safety wuld be rejected. The nus shuld remain with the Spnsr t determine the definitin f significant. Page 6 f 12
Key findings f the paper include: Pre-filtratin and pst-filtratin risks are inter-crrelated. A hlistic apprach needs t be taken in develpment f risk mitigatin strategies s that a seemingly high risk in ne prcess step can be mitigated thrugh cntrlling risk factrs in ther step. Risk factrs are inter-dependent and shuld be cnsidered jintly when evaluating their impact f the pre-filtratin and pst filtratin risks. This includes selectin f sample test vlume, acceptable biburden limits, sterilizing filter area, in an integrated assessment; A biburden level higher than 10 CFU/ 100 ml in the unfiltered drug slutin may nt incur unnecessary risk as its impact can be mitigated thrugh effective sterilizing filtratin; Sample vlumes less than 100 ml and acceptance limits different frm 10 CFU/100 ml can be justified, thrugh cntrlling ther risk factrs such as batch size r filter area, withut increasing the risk f biburden breakthrugh in the final filtratin; Testing f biburden prir t final filtratin and the sterile filtratin are nt the nly measures t cntrl the micrbial status and t generate sterile prduct. Hlistic cncepts may be established that reduce risks prir t the final biburden testing and the sterile filtratin. Enhanced understanding f the manufacturing prcess and prduct attributes is key t successful biburden risk management. 6. References 1. EMA (1996). CPMP Ntes fr Guidance n Manufacture f Finished Dsage Frm. 2. EMA (2012). EMA Guideline n the Requirements fr Quality Dcumentatin Cncerning Bilgical Investigatinal Medicinal Prducts in Clinical Trials. (EMA/CHMP/BWP/534898/2008), currently under revisin 3. Draft Guideline f EMA n the sterilizatin f the medicinal prduct, active substance, excipient and primary cntainer, April 11, 2016 4. U.S. FDA Guidance fr Industry: Sterile Drug Prducts Prduced by Aseptic Prcessing Current Gd Manufacturing Practice, 2004. 5. Jrnitz MW, Akers JE, Agallc JP, Madsen RE, and Melzer TH (2003). Cnsideratins in sterile filtratin. Part II: the sterilizing filter and its rganism challenge: a critique f regulatry standards. PDA Jurnal f Pharmaceutical Science and Technlgy. March/April. Vl. 57, N. 2, p 88 96. 6. Yang H, Li N and Chang S (2013). A risk-based apprach t setting sterile filtratin biburden limits. PDA J. f Pharm. Science and Technlgy. Vl. 67: 601-609 7. Addendum Manufacture f drug prducts (chemical synthesis r frm cell/micrbial cell cultures) generally fllw a similar set f prcess steps illustrated in Figure 1. Each f the nine steps includes cntrl f either biburden r sterility. They als include several filtratin and chrmatgraphic steps that may deplete any adventitius biburden frm the prduct. The prduct cntact prcess materials are cleaned and validated t minimize risk n cntaminatin. Page 7 f 12
Surce f material Material generated by bipharmaceutical prcess (prtein, Mab, DNA) N. f prcess step 1 Prcess Step Fermentatin Reductin/ cntrl f biburden Sterile cnditins (clsed systems, sterile feeds) Biburden checks Check fr freign rganisms 2 CFF (Cell free filtrate) after harvest Depth filtratin r Centrifugatin Pre- and pstfiltratin check 3 Purificatin (Catin exchange, Anin exchange, HIC, nanfiltratin, thers) Nanfiltratin, clumn resins trapping micrrganisms (+ cleaning cycles at unfavurable cnditins and validated re-use f clumns/ filter), qualified hld times Pre- and pstfiltratin checks Material generated by chemical synthesis (peptide, sirna, small mlecule) 4 Final Frmulatin after buffer exchange e.g. by UF/DF 0.2 micrn filtratin int sterilized cntainers, Class C envirnment Pre- and pstfiltratin check 5 DS strage cld r frzen in varius cntainers (bags, cryvessels, bttles, cans) Strage temperature slwing dwn r impeding micrbial grwth, qualified DS hld time (expiry) including biburden Biburden limit (specificatin) 6 Thawing f DS (if frzen) 0.2 micrn filtratin (ptinal), limited and qualifiied hld time (Pre- and) pst-filtratin check (ptinal) 7 Mixing/pling and dilutin/additin f excipients r full cmpunding Class C envirnment, 0.2 micrn filtratin (Pre-) and pstfiltratin check (ptinal) 8 Bulk fr sterile filtratin Limited and qualified hld time Pre-sterile filtratin check 9 Sterile Filtratin (inline during filling r ff-line befre filling) and dispensing int vials, syringes, cartridges etc. Sterile filtratin (ne r tw filters) Sterility check f final drug prduct and f sterile filtered bulk (if applicable) Figure 1. Typical prcess flw diagram fr manufacture f liquid bilgical drug prduct sterilised by filtratin. Page 8 f 12
Figure 2: Perfrmance characteristics f biburden testing using 100 ml samples and 10 CFU/100 ml acceptance limit, based n negative binmial distributin with a dispersin factr f 2 (6); dtted line: 5% acceptable risk bund. The Ph.Eur test fr biburden des nt have the capability t determine an accurate biburden cunt in a slutin as detectin depends n the vlume tested. The prbability f passing a batch fr biburden is best determined using a negative binmial distributin t accunt fr the clumping prperties f micrbes. Figure 2 illustrates such a prbability curve relatinship with the actual biburden level when 100 ml is tested. There is a significant prbability f failing and rejecting a gd batch r passing a bad batch n the basis f any single biburden test when cnsidered in islatin and applying an acceptance criterin f 10 CFU/100 ml. A hlistic apprach t biburden risk is recmmended that accmmdates the sterile filtratin prcess and assciated risk. Page 9 f 12
Figure 3: Design Space fr biburden test scheme (here: prbability f at least 1 CFU < 10-4 ) dependent n filtered vlume and filter area fllwing a risk-based apprach The relatinships demnstrate the breakthrugh risk fllwing biburden determinatin can be cntrlled thrugh the batch vlume filtered and/r filter area. The Spnsr may determine the biburden cntrl strategy that best fits the prcess and facility within the design space. A test sample vlume f 10 ml frm a 424 L filtratin batch size is illustrated (red circle) as acceptable when using a 1000 cm 2 filter. Page 10 f 12