Predictive Ppk calculations for biologics and vaccines using a Bayesian approach - a tutorial.

In pharmaceutical manufacturing, especially biologics and vaccines manufacturing, emphasis on speedy process development can lead to inadequate process development, which often results in less robust commercial manufacturing process after launch. Process performance index (Ppk) is a statistical measurement of the ability of a process to produce output within specification limits over a period of time. In biopharmaceutical manufacturing, progression in process development is based on Critical Quality Attributes meeting their specification limits, lacking insight into the process robustness. Ppk is typically estimated after 15-30 commercial batches at which point it may be too late/too complex to make process adjustments to enhance robustness. The use of Bayesian statistics, prior knowledge, and input from Subject matter experts (SMEs) offers an opportunity to make predictions on process capability during the development cycle. Developing a standard methodology to assess long term process capability at various stages of development provides several benefits: provides opportunity for early insight into process vulnerabilities thereby enabling resolution pre-licensure; identifies area of the process to prioritize and focus on during process development/process characterization (PC) using a data-driven approach; and ultimately results in higher process robustness/process knowledge at launch. We propose a Bayesian-based method to predict the performance of a manufacturing process at full manufacturing scale during the development and commercialization phase, before commercial data exists. Under Bayesian framework, limited development data for the process of interest at hand, data from similar products, general SME knowledge, and literature can be carefully formulated into informative priors. The implementation of the proposed approach is presented through two examples. To allow for continuous improvement during process development, we recommend to embed this approach of using predictive Ppk at pre-defined commercialization stage-gates, for example, at completion of process development, prior to and completion of PC, prior to technology transfer runs (Engineering/Process Performance Qualification, PPQ), and prior to commercial specification setting.

Use of Stability Modeling to Support Accelerated Vaccine Development and Supply.

Stability assessment of pharmaceuticals in specific storage and shipment conditions is a key requirement to ensure that safe and efficacious products are administered to patients. This is particularly relevant for vaccines, with numerous vaccines strictly requiring cold storage to remain stable. When stability evaluation is exclusively based on real-time data, it may represent a bottleneck for rapid and effective vaccine access. Stability modeling for vaccines represents a key resource to predict stability based on accelerated stability studies; nevertheless, this approach is not fully exploited for these kinds of products. This is likely because of the complexity and diversity of vaccines, as well as the limited availability of dedicated guidelines or illustrative case studies. This article reports a cross-company perspective on stability modeling for vaccines. Several examples, based on the direct experience of the contributors, demonstrate that modeling approaches can be highly valuable to predict vaccines' shelf life and behavior during shipment or manipulation. It is demonstrated that modeling methodologies need to be tailored to the nature of the vaccine, the available prior knowledge, and the monitored attributes. Considering that the well-established strategies reported in ICH or WHO guidelines are not always broadly applicable to vaccines, this article represents an important source of information for vaccine researchers and manufacturers, setting the grounds for further discussion within the vaccine industry and with regulators.

Analytical performance of a single epitope B-type natriuretic peptide sandwich immunoassay on the Minicare platform for point-of-care diagnostics.

Point-of-care B-type natriuretic peptide (BNP) testing with adequate analytical performance has the potential to improve patient flow and provide primary care givers with easy-to-use advanced diagnostic tools in the management of heart failure. We present the analytical evaluation of the Minicare BNP immunoassay under development on the Minicare I-20 platform for point-of-care testing. Analytical performance was evaluated using EDTA venous whole blood, EDTA plasma and capillary whole blood. Method comparison with a lab-testing system was performed using samples from 187 patients. Normal values were determined based on 160 healthy adults, aging from 19 to 70 years. Limit of blank (LoB), limit of detection (LoD) were determined to be 3.3 ng/L, 5.8 ng/L. Limit of quantitation (LoQ) in whole blood at 20% and 10% coefficient of variation (CV) was found < 9 ng/L and <30 ng/L respectively without significant differences between EDTA whole blood and EDTA plasma. Total CV was found to be from 6.7% to 9.7% for BNP concentrations between 92.6 and 3984 ng/L. The sample type comparison study demonstrated correlation coefficients between 0.97 and 0.99 with slopes between 1.03 and 1.09 between the different samples. Method comparison between Minicare BNP and Siemens ADVIA Centaur BNP demonstrated a correlation coefficient of 0.92 with a slope of 1.06. The 97.5% URL of a healthy population was calculated to be 72.6 ng/L. The Minicare BNP assay is a robust, easy-to-use and sensitive test for rapid determination of BNP concentrations that can be used in a near-patient setting.

A mathematical model for estimating residual transmission risk of occult hepatitis B virus infection with different blood safety scenarios.

BACKGROUND: If anti-hepatitis B core antibody testing is not mandated blood donors with occult hepatitis B infection (OBI) may transmit hepatitis B virus (HBV) to a recipient in spite of the use of nucleic acid amplification technology (NAT) or pathogen inactivation (PI). STUDY DESIGN AND METHODS: We developed a model to estimate OBI transmission risk based on three components: the probability distribution of the viral load (VL) in a randomly selected OBI donor, the probability that a given VL remains undetected, and the probability that this VL causes infection in the recipient. A subset of 217 South African OBI samples identified by individual donation (ID)-NAT screening were quantified by replicate testing using an ID-NAT assay (Ultrio Plus) against HBV DNA standard dilution series. The observed log VLs could be described by a Gumbel distribution. A correction was included to compensate for OBI samples missed by initial ID-NAT screening. RESULTS: The model estimates that 3.3% of all OBI donations are undetected by ID-NAT (Ultrio Plus) and cause infection by a blood component containing 20 mL plasma, going up to 8.7% when using minipools of 6 (MP6)-NAT. For 200-mL plasma transfusion these risks were estimated at 14 and 28%, respectively, while PI with modest (2 log) reduction capacity would reach 4.8% without NAT and 1.3 or 0.4% when combined with MP6- or ID-NAT. CONCLUSION: The model can be used to compare different screening and/or PI strategies in reducing viral transmission risk and could serve as a tool in evaluating efficacy of alternative blood safety scenarios.

Independent assessment of matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) sample preparation quality: A novel statistical approach for quality scoring.

Preparation of samples according to an optimized method is crucial for accurate determination of polymer sample characteristics by Matrix-Assisted Laser Desorption Ionization (MALDI) analysis. Sample preparation conditions such as matrix choice, cationization agent, deposition technique or even the deposition volume should be chosen to suit the sample of interest. Many sample preparation protocols have been developed and employed, yet finding the optimal sample preparation protocol remains a challenge. Because an objective comparison between the results of diverse protocols is not possible, "gut-feeling" or "good enough" is often decisive in the search for an optimum. This implies that sub-optimal protocols are used, leading to a loss of mass spectral information quality. To address this problem a novel analytical strategy based on MALDI imaging and statistical data processing was developed in which eight parameters were formulated to objectively quantify the quality of sample deposition and optimal MALDI matrix composition and finally sum up to an overall quality score of the sample deposition. These parameters can be established in a fully automated way using commercially available mass spectrometry imaging instruments without any hardware adjustments. With the newly developed analytical strategy the highest quality MALDI spots were selected, resulting in more reproducible and more valuable spectra for PEG in a variety of matrices. Moreover, our method enables an objective comparison of sample preparation protocols for any analyte and opens up new fields of investigation by presenting MALDI performance data in a clear and concise way.

Quantitative analysis of morphine in dried blood spots by using morphine-d3 pre-impregnated dried blood spot cards.

Two different internal standard dried blood spot (DBS) pre-impregnation procedures (prior to blood spotting) were investigated. In the first procedure DBS pre-impregnation is performed by immersing the DBS card fully into an internal standard solution. In the second procedure pre-impregnation is performed by pipetting a certain volume of an internal standard solution onto the DBS card. Morphine-d3 was used as the model compound for all experiments. The pre-impregnation procedure by immersing was further investigated with respect to homogeneity of impregnation, influence of different blood spotting techniques and the influence of spotting different blood volumes on the internal standard distribution, calibration and stability of pre-impregnated cards. Finally, the immersing procedure was used for the analysis of morphine in dried blood spots and the results were compared to the conventional procedure in which the internal standard morphine-d3 was added to the extraction solvent. The new pre-impregnated cards couple simplicity of operation and convenient use in the field to results equivalent to the conventional procedure.

Is there a suitable point-of-care glucose meter for tight glycemic control? Evaluation of one home-use and four hospital-use meters in an intensive care unit.

BACKGROUND: Implementation of tight glycemic control (TGC) and avoidance of hypoglycemia in intensive care unit (ICU) patients require frequent analysis of blood glucose. This can be achieved by accurate point-of-care (POC) hospital-use glucose meters. In this study one home-use and four different hospital-use POC glucose meters were evaluated in critically ill ICU patients. METHODS: All patients (n = 80) requiring TGC were included in this study. For each patient three to six glucose measurements (n = 390) were performed. Blood glucose was determined by four hospital-use POC glucose meters, Roche Accu-Check Inform II System, HemoCue Glu201DM, Nova StatStrip, Abbott Precision Xceed Pro, and one home-use POC glucose meter, Menarini GlucoCard Memory PC. The criteria described in ISO 15197, Dutch TNO quality guideline and in NACB/ADA-2011 were applied in the comparisons. RESULTS: According to the ISO 15197, the percentages of the measured values that fulfilled the criterion were 99.5% by Roche, 95.1% by HemoCue, 91.0% by Nova, 96.6% by Abbott, and 63.3% by Menarini. According to the TNO quality guideline these percentages were 96.1% , 91.0% , 81.8% , 94.2% , and 47.7% , respectively. Application of the NACB/ADA guideline resulted in percentages of 95.6%, 89.2%, 77.9%, 93.4%, and 45.4%, respectively. CONCLUSIONS: When ISO 15197 was applied, Roche, HemoCue and Abbott fulfilled the criterion in this patient population, whereas Nova and Menarini did not. However, when TNO quality guideline and NACB/ADA 2011 guideline were applied only Roche fulfilled the criteria.

Alignment and clustering strategies for GC×GC-MS features using a cylindrical mapping.

Comprehensive two-dimensional gas chromatography coupled to mass spectrometry is a powerful tool to analyze complex samples. For application of the technique in studies like biomarker discovery in which large sets of complex samples have to be analyzed, extensive preprocessing is needed to align the data obtained in several injections (analyses). We developed new alignment and clustering algorithms for this type of data. New in the current procedures is the consistent way in which the phenomenon referred to as wrap-around is treated. The data analysis problems associated with this phenomenon are solved by treating the 2D display as the surface of a three-dimensional cylinder. Based on this transformation we developed a new similarity metric for features as a function of both the cylindrical distance (reflecting similarity in chromatographic behavior) and of the mass spectral correlation (reflecting similarity in chemical structure). The concepts are used in warping and clustering, and include a protection against greedy warping. The methods were applied - for the purpose of an example - to the analysis of 11 replicates of a human urine sample concentrated by solid phase extraction. It is shown that the alignment is well protected against greedy warping which is important with respect to analytical qualities as robustness and repeatability. It is also demonstrated that chemically similar features are clustered together. The paper is organized as follows. First a brief introduction is provided addressing the background of the GC×GC-MS data structure followed by a theoretical section with a conceptual description of the procedures and details of the algorithms. Finally an example is given in the experimental section, illustrating the application of the procedures.

A stochastic model of the processes in PCR based amplification of STR DNA in forensic applications.

In forensic DNA profiling use is made of the well-known technique of PCR. When the amount of DNA is high, generally unambiguous profiles can be obtained, but for low copy number DNA stochastic effects can play a major role. In order to shed light on these stochastic effects, we present a simple model for the amplification process. According to the model, three possible things can happen to an individual single DNA strand in each complete cycle: successful amplification, no amplification, or amplification with the introduction of stutter. The model is developed in mathematical terms using a recursive approach: given the numbers of chains at a given cycle, the numbers in the next can be described using a multinomial probability distribution. A full set of recursive relations is derived for the expectations and (co)variances of the number of amplicon chains with no, 1 or 2 stutters. The exact mathematical solutions of this set are given, revealing the development of the expectations and (co)variances as function of the cycle number. The equations reveal that the expected number of amplicon chains without stutter grows exponentially with the cycle number, but for the chains with stutter the relation is more complex. The relative standard deviation on the numbers of chains (coefficient of variation) is inversely proportional to the square root of the expected number of DNA strands entering the amplification. As such, for high copy number DNA the stochastic effects can be ignored, but they play an important role at low concentrations. For the allelic peak, the coefficient of variation rapidly stabilizes after a few cycles, but for the chains with stutter the decrease is more slowly. Further, the ratio of the expected intensity of the stutter peak over that of the allelic peak increases linearly with the number of cycles. Stochastic models, like the one developed in the current paper, can be important in further developing interpretation rules in a Bayesian context.

Representative drug sampling: sample size calculations revisited.

Calculating the required number of samples to be tested from a consignment of pills suspected of containing drugs can be performed from a Bayesian perspective. Procedures in literature are based on the outstanding work of Aitken. However, in the mathematical treatment of the problem, the limitedness of the consignment is not systematically used. The current Technical Note addresses this problem. A suitable prior distribution for the number of positive pills is derived, being a betabinomial distribution with the consignment size as one of the parameters. A hypergeometric likelihood is used, as sampling generally proceeds without replacement. The betabinomial posterior distribution is mathematically identical to the predictive distribution as reported elsewhere. The currently used large consignment approximation can be derived from the betabinomial posterior, but the quality is not optimal when compared to the exact betabinomial-based results. A new approximation is derived, with better properties, as illustrated in some examples.

Refinement of a viral transmission risk model for blood donations in seroconversion window phase screened by nucleic acid testing in different pool sizes and repeat test algorithms.

BACKGROUND: In minipool nucleic acid test (MP-NAT) screening protocols, the donations implicated in reactive test pools are released for transfusion when they are nonreactive in a repeat test on the individual samples, but in individual-donation (ID)-NAT screening algorithms the release of nonrepeatable reactive (NRR) donations is under discussion. STUDY DESIGN AND METHODS: A previously developed window phase (WP) transmission risk model for NAT-screened blood transfusions has been refined to take the effect of repeat tests of initially reactive (IR) MP- or ID-NAT results into account. The model has then been applied to simulate the effect of different screening algorithms with ULTRIO and the new-generation ULTRIO Plus assay (Novartis Diagnostics) on transmission risk for hepatitis B virus (HBV), hepatitis C virus (HCV), and human immunodeficiency virus (HIV). RESULTS: We calculated WP risk-day equivalents for MP16-, MP8-, and ID-NAT with and without duplicate retesting of IR results of 3.1, 2.7, 1.5, and 1.3 days for HCV; 6.3, 5.5, 3.3, and 2.9 days for HIV; and 24.4, 22.2, 15.6, and 14.1 days for HBV, respectively. These latter infectious HBV WPs reduced to 20.4, 18.2, 11.6, and 10.3 days, respectively, with the more sensitive ULTRIO Plus assay. CONCLUSION: ULTRIO Plus ID-NAT screening reduces the virus transmission risk in the WP by 54% to 58% compared to ULTRIO MP16-NAT, while the incremental risk caused by releasing donations with duplicate ID-NAT NRR results is 5% to 6%. To achieve maximum safety and specificity a similar repeat test algorithm can be applied to ID-NAT as used for serologic assays.

A statistical approach to assess analytical sensitivity of diagnostic assays in the presence of false-positive results.

Analytical sensitivity is one of the performance characteristics of diagnostic tests. As always, a subtle balance has to be found between sensitivity and specificity, but given the clinical setting and the intended use of the diagnostic system, a less than 100% specificity can be perfectly acceptable. Probit analysis is a wide-spread statistical technique used to assess the concentrations corresponding to pre-set hit rates, but this technique assumes a 100% specificity. The current paper describes a statistical procedure, derived from the standard probit technique, to deal with the incidence of false-positive results in the assessment of the analytical sensitivity.

Sensitivity studies for quantitative assays: use of censored data analysis.

Probit regression analysis is frequently used to study the relation between the concentration of an analyte in a sample and the probability that the assay used yields a positive test result. For these analyses only the qualitative classification 'positive' or 'negative' is used, whereas, particularly in the case where the assay is quantitative in nature, the results contain more information. In the current paper, we propose an alternative method, in which the negative test results are treated as being (left) censored. As such, more efficient use is made of the information in the data. The procedures are illustrated using two generations of NucliSens assays (BioMérieux), which are designed to quantify the viral load of HIV-1 in blood samples. Computer simulations are used to illustrate some properties of the estimated parameters.

Stochastic processes defining sensitivity and variability of internally calibrated quantitative NASBA-based viral load assays.

For quantitative assessment of virus particles in patient plasma samples various assays are commercially available. Typical performance characteristics for such assays are sensitivity, precision and the range of linearity. In order to assess these properties it is common practice to divide the range of inputs into subranges in order to apply different statistical models to evaluate these properties separately. We developed a general statistical model for internally calibrated amplification based viral load assays that combines these statistical properties in one powerful analysis. Based on the model an unambiguous definition of the lower limit of the linear range can be given. The proposed method of analysis was illustrated by a successful application to data generated by the NucliSens EasyQ HIV-1 assay.

Mathematic modeling of the risk of HBV, HCV, and HIV transmission by window-phase donations not detected by NAT.

BACKGROUND: Blood transfusion centers around the world have introduced minipool NAT to reduce the risk of HBV, HCV, and HIV transmission by blood donations drawn in the infectious window phase. What would be the reduction in the residual risk when minipool NAT would be replaced by single-donation NAT? STUDY DESIGN AND METHODS: A mathematic model was developed to estimate the probability of virus transmission by blood transfusion when NAT screening methods are used for virologic safety testing. The major assumptions used are threefold: 1) The viral nucleic acid concentrations in the early window phase of infection double in 2.8 (HBV), 0.74 (HCV), and 0.90 (HIV) days. 2) The detectability of low copy numbers of viral DNA or RNA by the screening assay can be described with a probit model. 3) The probability of infection depends linearly on the logarithm of the administered dose, with 50-percent infectivity rates at 10 (HBV and HCV) or 1000 (HIV) viral nucleic acid copies per transfusion unit (estimates based on NAT studies with samples of known infectivity in chimpanzees). RESULTS: A reasonably simple equation was obtained that allows studying the effect of the sensitivity of the NAT assay and of the pool size used for screening on the residual risk of transfusion-transmitted infection. The computations are illustrated by using observed sensitivity estimates of various NAT methods. By using epidemiologic data among European donors over 1997 as baseline, the calculations predict that the incidence of virus transmission per 10-million RBC transfusions reduces with the following numbers when lowering the test pool size from 96 to 1 (single-donation testing): HBV from 11 to 13 to 3.3 to 5.1, HCV from 1.7 to 2.0 to 0.5 to 0.8, and HIV from 0.47 to 0.62 to 0.010 to 0.045 (ranges for the different NAT screening methods). CONCLUSION: A proper mathematic model for the calculation of residual infection risk by blood transfusion helps understand the impact of introducing new NAT methods for blood safety testing.

Principles of quantitation of viral loads using nucleic acid sequence-based amplification in combination with homogeneous detection using molecular beacons.

For quantitative NASBA-based viral load assays using homogeneous detection with molecular beacons, such as the NucliSens EasyQ HIV-1 assay, a quantitation algorithm is required. During the amplification process there is a constant growth in the concentration of amplicons to which the beacon can bind while generating a fluorescence signal. The overall fluorescence curve contains kinetic information on both amplicon formation and beacon binding, but only the former is relevant for quantitation. In the current paper, mathematical modeling of the relevant processes is used to develop an equation describing the fluorescence curve as a function of the amplification time and the relevant kinetic parameters. This equation allows reconstruction of RNA formation, which is characterized by an exponential increase in concentrations as long as the primer concentrations are not rate limiting and by linear growth over time after the primer pool is depleted. During the linear growth phase, the actual quantitation is based on assessing the amplicon formation rate from the viral RNA relative to that from a fixed amount of calibrator RNA. The quantitation procedure has been successfully applied in the NucliSens EasyQ HIV-1 assay.