article

Developments in stability testing and evaluation

The Joint Pharmaceutical Analysis Group (JPAG) held a stability meeting at the Royal Society of Chemistry’s headquarters in London earlier this year. Attended by 65 delegates, its focus was on developments in stability testing and evaluation. Here follows a summary of the presentations made at the meeting.

In-use stability: the BSAC antimicrobial drug stability programme; an innovative model for collaborative data sharing

Dr Conor Jamieson, Pharmacy Team Leader – Antimicrobial Therapy, Sandwell and West Birmingham NHS Trust

Dr Jamieson explained the role of outpatient parenteral antimicrobial therapy (OPAT) in the management of infection. OPAT is a model of treatment that can facilitate early discharge from hospital or admission avoidance, reducing pressure on the NHS and improving the experience for patients who might otherwise need a long hospital admission.

The challenges of OPAT are finding antimicrobial agents with suitable characteristics for once- or twice-daily administration, which are convenient to administer, safe and meet the needs of the service as well as the antimicrobial stewardship agenda. He contrasted the treatment of Staphylococcus aureus infection using either ceftriaxone or flucloxacillin, to demonstrate the challenges posed by once-daily dosing with ceftriaxone versus the narrow spectrum, preferred agent, flucloxacillin, which required multiple daily dosing.

Administering antibiotics as continuous infusion via an elastomeric device is also a useful option for OPAT, but there must be assurance that the drugs are stable over the 24-hour infusion period. There were no data meeting the NHS guidance on stability available for key antibiotics for OPAT, so the drug stability-testing programme was launched to make data available for open access by the OPAT community.

Dr Jamieson then discussed the findings for flucloxacillin stability in citrate buffer in two commercially available, ambulatory devices; INfusor LV (Baxter) and Accufuser (Woo Young Medical). Initial stability evaluation found that the product was too unstable in WFI, 0.9 percent NaCl, MCP buffers, 5.25 percent citrate buffer (buffer pH ranges were pH 5-8); showing both high degradation and precipitation. Therefore, 0.3 percent citrate buffer was chosen at two clinically‑relevant concentrations, which were stable for 24 hours when heated to 37°C. In addition, stability in the two elastomeric devices at two different concentrations was evaluated. In both cases good stability data were generated. A similar study was conducted with meropenem, but stability for only six hours could be demonstrated, limiting its usefulness in OPAT. Work had been due to commence on ceftazidime, but significant challenges with its stability and the formation of the neurotoxic degradant pyridine were recognised. Future projects include ceftazidime/avibactam and possibly benzyl penicillin, amoxicillin, cefepime and temocillin. The working party is keen to work in partnership with the pharmaceutical industry and device manufacturers to generate and disseminate open access stability data.

IQ Lean Stability Working Group update

Susan Smith, AstraZeneca

Susan Smith provided an update on the IQ (International Consortium for Innovation & Quality in Pharmaceutical Development) working group on lean stability. She introduced the concept of lean stability, which constitutes science and risk‑based strategies that utilise all available product knowledge and understanding to develop the right stability plan for a product, resulting in either a ‘lean’ protocol or an enhanced protocol.

Smith presented the following key lean stability concepts:

  • Stability-related quality attribute (SRQA) Typically, these constitute a subset of quality attributes that either directly/ indirectly impact stability but may or may not impact other manufacturing parameters, or are attributes whose values may change depending on stability.
  • Shelf life-limiting attributes (SLLA) These are a subset of stability-related quality attributes that are shelf-life limiting under label conditions.
  • Stability risk assessment (SRA) This compiles existing stability knowledge to identify which stability aspects require further study or monitoring. SRAs can be used throughout a product’s lifecycle.
  • Accelerated predictive stability (APS) This entails accelerated stability testing at a range of temperatures and relative humidities, in conjunction with statistical fitting to a modified Arrhenius equation, to determine product shelf life as a function of storage conditions and packaging.

Smith then highlighted the pros and cons of these approaches. The pros included:

  • reduction of testing burden
  • focusing resources on the right testing
  • allocation of shelf life based on stability‑indicating attributes only
  • enhanced efficiency
  • using scientific understanding to fully characterise products with reduced likelihood of future surprises.

The cons included:

  • less collected data
  • nontraditional
  • the need to understand reliability of methods and uncertainty together with a full information package.

…most companies would like to see additional ICH harmonised guidance relating to stability testing and more data to support a post‑approval change”

Smith then discussed feedback from a recent IQ benchmarking survey that encompassed 23 companies. The findings were as follows: SRAs are utilised by the majority of responders (70 percent); SRAs have been used by over half of respondents in regulatory submissions (56 percent), typically at registration or post-approval stage; SRAs are leveraged for several different purposes over the course of development.

Risk-based predictive stability (RBPS) is used to customise protocols. There is heavy usage in early development (94 percent), typically customising temperature and packaging parameters. An overwhelming majority of surveyed companies felt that science and risk-based strategies should also apply to annual stability commitments in production. Most responders felt they should only test stability-limiting attributes and that testing frequencies should be decreased, the number of batches should be limited and the test conditions should be reduced. Responders felt that statistical modelling, moisture vapour transmission (MVTR) studies and risk-based approaches, including accelerated predictive stability, should be allowed during stability protocol reduction discussions.

However, regulatory endorsement was extremely varied. On this basis most companies would like to see additional ICH harmonised guidance relating to stability testing and more data to support a post‑approval change and new ICH guidance on Science and Risk-based Approaches to Stability Testing for Post‑approval CMC Changes.

Stability prediction modelling – statistical considerations

Dr Jonathan Bright, AstraZeneca

The time until a drug product stored at 25°C/60%RH breaches an impurity limit is often shelf-life limiting; you can either place the sample on real-time stability and wait, or use six or so harsher combinations of T and RH before using statistical modelling to predict what will happen at 25°C/60%RH. A key question is: for any given T, RH and timepoint, how many replicates are required?

Dr Bright then reviewed several different statistical designs but surprisingly demonstrated that the outputs were often very similar. He concluded that the experiment is usually designed to efficiently estimate the relation between the measured response of interest and the explanatory variables. However, if the purpose of the experiment is not the relationship per se, but a prediction outside of the design space, then other – perhaps surprising – designs may be equally good alternatives. Dr Bright then reviewed some typical data outputs. He concluded that it is important to “picture the raw data” and to reflect on the importance of the zero time value – its re‑use and high leverage – by using repeats and then estimating a single zero time value that is applicable to all conditions. He concluded that it was also necessary to “picture the data” being modelled against the fitted model. Be aware that model diagnostics do not speak directly to the quality of predictions outside the design space and extrapolations should be minimised as far as possible.

Finally, Dr Bright concluded that if the purpose of an experiment is a prediction outside the design space, you must be aware that some unusual-looking designs may be good options. It was necessary to “picture both the raw data and the modelled data” (if different). The zero time value is critical – it should be based on several measurements and may be estimated as a single value applicable to all conditions. Good model diagnostics (R-squared etc) are necessary, but not sufficient, for a good prediction.

Advanced modelling from highly accelerated stability testing to determine drug product shelf life

Dr Maria Krisch, FreeThink Technologies

An accelerated stability assessment programme (ASAP) study involves exposing a product to a range of open (high stress) conditions; eg, 40-80°C and 0-80 percent relative humidity. The shelf life-limiting characteristics of the drug substance or drug product are assessed, such as related substance, assay, colour or dissolution changes. The experimental data are then modelled using the ASAPprime® software programme, using a humidity-corrected Arrhenius equation. The resultant model gives a fundamental understanding of the temperature and humidity dependence of the product and allows for shelf-life determination, with calculation for the appropriate packaging. In addition, the impact of storage and shipping excursions can be calculated explicitly. An ASAP approach enhances the development speed; a traditional ICH approach takes in the order of six months minimum whereas an ASAP study takes only five to six weeks.

Dr Krisch asked why traditional accelerated ageing approaches have not been predictive of real-time stability, deducing it is because over 50 percent of products show complex kinetics. She then questioned whether kinetic modelling can be performed in the absence of detailed kinetic studies, suggesting the solution is isoconversion, which focuses on a set level of change – time to ‘edge of failure’ eg, time to specification limit (specific degradant, total degradant, etc) is determined. These isoconversion times under different stress conditions constitute the input values for the humidity corrected Arrhenius equation, which is then used to predict the time to reach the specification limit under real‑time storage conditions.

ASAP determines the actual activation energy (Ea) for each drug product. Thus, for a low Ea substance (eg, 17kcal/mol), six months at 40°C/75%RH is equivalent to a two-year shelf life at 25°C/60%RH, while for a high Ea substance (eg, 40kcal/mol), six months at 40°C/75%RH is equivalent to a 12-year shelf life at 25°C/60%RH. The higher the B factor (humidity sensitivity factor) in the modified Arrhenius equation, the more product stability can be controlled with packaging. Based on an assessment of many drug products and drug substances, the average Ea is 27kcal/mol and the average B value is 0.04. As with all extrapolations based on short-term data, the output is accurate but imprecise. A statistical treatment incorporates the error in the measurements to determine the probability of achieving a certain shelf life.

In order to assess shelf life inside packaging, the relative humidity inside packaging can be calculated precisely using straightforward input information. The moisture vapour transmission rate (MVTR) characterises moisture ingress into packaging at known external storage conditions (%RH and T). Moisture sorption isotherms (MSIs) characterise how much moisture is held by the drug product and any desiccant in the packaging. Using either experimental data or values in the ASAPprime® database, these inputs are used to calculate the internal relative humidity over time. The stability model, which gives the temperature and %RH dependence of stability over time, is then combined with the calculated relative humidity inside the packaging as a function of time to give stability over time within a certain packaging configuration.

In conclusion, accelerated stability studies can be effectively applied to establish shelf life with ASAPprime® using the concept of isoconversion, statistical treatment and a model that accounts for moisture sensitivity. Short‑term experimental measurements under open conditions can be used to assess a wide range of packaging configurations. Consequently, ASAPprime® enables better decisions, faster.

Risk-based predictive stability – an industry perspective

Rachel Orr, GlaxoSmithKline

Rachel Orr discussed how utilising a ‘risk-based approach’ is common practise within the pharmaceutical industry and gave examples of it; such as the modelling of process parameters to develop design spaces, quality by design (QbD) and technical risk assessments, sunset testing and animal testing models to name a few. In addition, Orr noted that risk-based approaches are encouraged by regulators, citing numerous publications from the FDA, WHO, EMA and ICH to demonstrate this. It was, however, acknowledged that taking risk-based approaches in stability has lagged behind other areas of the industry and that this area needs considerable improvement.

Risk-based stability can be subdivided into two approaches: lean stability, eg, testing, bracketing and matrixing, or predictive stability (RBPS), such as stress testing and accelerated stability monitoring. Orr commented that there has been an evolution in the stability of science since ICH Q1A was first published with improvements to both modelling tools and protocol development. However, due to a lack of updated guidance, RBPS has been implemented inconsistently across the industry. In recognition of this, in 2015 the IQ consortium launched a working group to focus on RBPS tools to optimise pharmaceutical development. The working group has approximately 50 members from 18 companies across the pharmaceutical industry and 18 pharma companies were surveyed in 2016. The results were first published in 2017. The key findings of this survey were that although RBPS was regularly used in regulatory submissions by most of the surveyed companies (70 percent), this was primarily in clinical phases (80 percent) and the majority of companies (85 percent) continue to generate full ICH stability data despite several companies having successfully leveraged these advanced stability approaches within their development paradigm.

…risk-based approaches in stability has lagged behind other areas of the industry and… this area needs considerable improvement”

The FDA guidance for post-approval changes subdivides changes into ‘minor’, ‘moderate’ and ‘major’, based on what has been changed and the perceived risk that this introduces. From a stability perspective, these categories are allocated on the following basis: (i) minor – no stability data is required, (ii) moderate – some additional stability is required as supportive data and (iii) major – a complete stability dataset will be required post change. However, in place of this generic classification for changes, the option to use RBPS tools to assess the impact of the change for appropriate categorisation was proposed. Orr discussed a decision tree, which would be useful in such discussions, where the two key questions would be: (i) are there any changes to site, route, process or packaging before file? and (ii) does modelling suggest any change in stability performance?

Orr concluded by indicating that risk-based approaches are actively encouraged across various areas of pharmaceutical development. Unfortunately, stability has lagged behind other areas with respect to taking up these opportunities. The IQ consortium has lobbied companies in an attempt to harmonise approaches and initial scoping of the landscape has demonstrated a lack of confidence in subsequent regulatory acceptance. Consequently, the group has defined some proposed regulatory templates for future use. However, continued advocacy with regulators is still required.

Predicting the long-term dissolution performance of an immediate-release tablet using accelerated stability studies

Dr Gary Scrivens, Pfizer

Dr Gary Scrivens indicated that, based on 12-month real-time and intermediate storage and six-months accelerated data in three packs (2x bottles with desiccant and foil-foil blisters), the dissolution of an instant-release tablet was slowing down during registration stability testing. The concern was whether this slow down would continue up to 24 months or whether it had “plateaued out”. Dr Scrivens questioned whether long‑term dissolution performance could be predicted from short-term accelerated stability studies. To examine this proposition, they exposed tablets to a range of different higher temperatures (T) and relative humidity (RH) conditions for short time periods. They then modelled the effects of T, RH and time (t) on the dissolution performance and extrapolated to long-term conditions, comparing it to registration stability data. Then they conducted accelerated testing on fresh batches. The predictive model assessed the following:

  • the influence of T and RH on the long‑term stability risk in different climatic zones
  • the influence of packaging types, including desiccants
  • whether stability trends will continue or reach a plateau
  • the maximum extent of the problem at future timepoints
  • sufficient parameters to provide a rapid screening tool for future lots.

During the short-term accelerated study, the dissolution was observed to slow down in a similar manner to the registration stability. The slowdown in dissolution could be conveniently measured as a single parameter referred to as the acceleration factor (AF); which is the degree by which the slower dissolution curve needs to be compressed on the time axis to overlay with the initial dissolution curve. AF values less than 1 indicate that the dissolution has slowed down. The Weibull function (in conjunction with the Excel Solver tool) provided a rapid means of calculating AF. The AF approach has the advantage of predicting the percent dissolution at any dissolution time point. Very similar (but not identical) predictions would be obtained if ‘percent release after X min’ was modelled instead of AF.

In the accelerated studies, the dissolution slowdown was observed to tend towards a ‘limit’ value (AFinf); with T and RH appearing to affect both AFinf and the rate constant (k) at which the dissolution tends towards this limit. Both AFinf and k were modelled as functions of T and RH using an empirical (data-led) approach. The best-fit models were found to have the same structure as the humidity-modified Arrhenius equation, often used for modelling chemical degradation rates in the solid state. Humidity was found to be the most significant factor affecting the plateau level (AFinf).

This model was found to accurately predict the dissolution of freshly made batches stored at various T and RH levels. Moreover, this model in combination with simulations of RH conditions inside packaging was also found to be fairly accurate at predicting the dissolution profiles observed in the long-term registration stability studies across all assessed packaging types.

In summary, long-term dissolution can be reasonably predicted from short-term accelerated data. A useful screening protocol has been devised (using less elevated temperatures). The model predicted that the dissolution profile should not significantly further slow down at future time points; this was later confirmed by real-time data. The model quantifies the effects of temperature and humidity on the dissolution stability and thus the long‑term dissolution performance can be predicted for any packaging type in any climatic zone.

Stability management regulations: global stability requirements – development of a CMC regulatory intelligence solution

John Cleverley, Clarivate Analytics

John Cleverley indicated that one of the major hurdles faced when launching a new drug is the approval of chemistry, manufacturing and controls (CMC) data. CMC modules are uniquely challenging because they must be compiled for individual countries and they depend on regulatory requirements that can be difficult to find and verify. In particular, access to low- and middle-income countries (LMICs) is negatively impacted by diverse regulatory requirements and limited CMC requirement visibility. Across multiple countries each requirement can result in a cumulative impact on costs and delays in product introduction; a global health product is typically introduced in over 30 countries. CMC issues include:

  • some countries have specialised excipient requirements; eg, restrictions on porcine products
  • some Latin American, African and ASEAN (Association of South East Asian Nations) countries require 30ºC/75%RH stability studies, while most will accept 30ºC/65%RH.

A successful stability study will establish the shelf life date of the drug product for the specified climatic zone under the product label-stated storage conditions. However, even where the US and EU have incorporated most ICH guidelines, requirement differences remain. Whereas EMA stability requirements are aligned with ICH guidelines, in the US the required number of batches and storage time depend on product dosage form and stability.

Clarivate Analytics has developed Cortellis CMC Intelligence, with the support of a grant from the Bill & Melinda Gates Foundation, with the intention of establishing a long-term, sustainable database of CMC regulatory requirements for small molecule products, with an initial focus on LMICs.

Cleverley went on to explain that Cortellis CMC Intelligence provides information on the following stability parameters: number and type of batches required, stability commitment, acceptance of bracketing/ matrixing, need for in-use stability data, whether the country follows ICH, whether statistical extrapolation is acceptable and whether stress testing and photostability are required.

In summary, CMC requirements are challenging because they must be compiled for individual countries; national regulatory authorities require stability data to ensure a product maintains consistent quality and shelf life and there is a need to manage stability requirements across multiple territories and address variations in requirements.

Case study: streamlining the management of stability studies

Dr Paul Fullerton, AMRI

Dr Paul Fullerton explained that the Glasgow AMRI site had been through a series of acquisitions. In 2015 it utilised entirely paper-based systems, but as the capacity of the site increased, electronic systems became necessary. The site now uses “Stratas”, an electronic QMS system that handles investigations, deviations, change controls, etc.

In parallel, Oracle EBS (E-Business Suite) has been introduced. Primarily brought in for warehouse/inventory, etc; it has since been adapted for other uses, including stability. Its benefits are as follows:

  • stability samples can be given an intermediate ‘hold’ status until placed in stability chambers and assigned a further status upon reception
  • samples have standard, automatically‑assigned labels that facilitate inventory control
  • inventory numbers are automatically updated and can be reviewed against stability documentation.

Staff could be assigned reviewer, approver or entry capabilities. The system also allows for routine reports to be created and viewed, either directly or through export to Word, Excel, etc. The system can also query by product, storage location, shelf life, etc. This aids to resolve customer queries and helps with stock control and capacity planning. The most involved aspect of the project was the actual definition of the process to ensure stock control.

In summary, although not specifically designed for stability control, Oracle EBS was found to provide suitable control and oversight. The system did not fully align with required procedures; however, adjustment of working practices remedied those deficiencies. Ultimately, the system promoted greater involvement between teams, rather than being solely a QC function. The system does not currently allow for flagging/oversight of individual time points; however, this is currently being assessed through use of batch attributes.

 

The session convenors were Malcolm Dash, MHRA and Dave Elder, JPAG chair