QUALITY CONTROL - High-Temperature Stability | Do the Math for Shelf Life
By Michelle Duncan, PhD, and Irene Zaretsky, MS
Pharmaceutical scientists routinely predict long-term chemical stability at a lower temperature using data generated at a higher temperature over a shorter time period. The use of 40 degrees C chemical data (e.g., assay and related substances) to predict levels over long-term 25 degrees C storage has become such common practice that the underlying theory is overlooked or was never learned.
There is disagreement about whether three months at 40 degrees C indicates expected 25 degrees C levels for 12 months or for 24 months because of insufficient understanding or information about the kinetic model. More importantly, without some theoretical understanding, the practice is inappropriately applied to situations that do not fit the kinetic model upon which it is based, resulting in erroneous predictions.
In addition to providing the background of how these kinetic predictions work, this paper will provide the key to understanding the temperature relationships and the ability to more accurately predict the expected long-term levels for a specific product.
Arrhenius Kinetics: Where This All Began
Swedish scientist Svante Arrhenius provided the first kinetic model to interpret the effect of temperature on reaction rate given by Equation 1 (see info box).
1-3 The Arrhenius equation can be applied regardless of the order (zero-order, first-order, etc.) of the reaction kinetics.
4 Equation 2 presents the linear form of the Arrhenius equation for graphical presentation (y = mx + b). For many reactions, a linear relationship can be obtained between the inverse of temperature (in degrees Kelvin) and the natural log (Ln) of the measured rate constant (k), as shown in Figure 1.
continues below…
The Arrhenius Equation
K = A exp (-Ea/RT) (Equation 1)
k = reaction rate constant
A = Arrhenius factor (y-intercept constant)
Ea = the energy of activation for the reaction, cal/mole (1000 cal = 1 kcal)
R = the ideal gas constant, 1.987 calories/deg mole
T = the absolute temperature (degrees Kelvin), for 25° C T= 298° K and for 40°C T= 313° K)
This equation can be written in several equivalent forms as follows:
Ln k = -Ea/RT + Ln A (Equation 2, y = mx + b)
Ln k2/k1 = -Ea/R*(1/T2-1/T1) (Equation 3)
k2 = k1 * exp[-Ea/R*(1/T2-1/T1)] (Equation 4)
The constants k1 and k2 are the rate constants at temperature T1 and T2 (for example 25 degrees C and 40 degrees C), respectively.
Equation 3 presents a simplified form for use with two temperatures. Equation 4 expresses the relationship between the reaction rates and the corresponding temperatures when the activation energy (Ea) of the reaction is known. Equation 4 allows for the calculation of a reaction rate constant at a lower temperature when the activation energy and the reaction rate at the higher temperature are known.
Figure 1. Arrhenius plot of Ln K against 1/T. Slope = - Ea/R k = reaction rate constant, T = temperature in degrees Kelvin Ea = activation energy, R = ideal gas constant.
As shown by Figure 1, the Arrhenius model provides the ability to determine the reaction rate and, hence, predict stability at any temperature with knowledge of the activation energy (Ea) and the reaction rate at another temperature.
Limitations to Arrhenius’ Model
The first requirement is a reaction (ongoing) where the reaction rate constant at a given temperature can be determined. When monitoring the generation of a primary degradation product, the presence of a secondary degradation reaction can introduce error to the calculation of the primary rate constant and any attempted Arrhenius model predicted rates. Likewise, if a component is consumed to the extent that the reaction equilibrium changes, the reaction rate will not remain constant. The Arrhenius kinetic model requires a reaction rate constant (k).
The Arrhenius kinetic model can be utilized across the temperature range where a constant relationship between the effect of temperature on the reaction rate is maintained, where the graph is linear. The reaction mechanism should not change over the temperature range studied. Conformance to the model (linearity) is often lost when crossing a phase change, such as freezing and the glass-phase transition of proteins and peptides. Reactions where the rate is dependent upon oxygen, light (photochemical), diffusion, or microorganism-based decomposition may not demonstrate Arrhenius kinetics over any temperature range.
Figure 2. The zero-order kinetic model for 10% drug loss over 24 months at 25 degrees C shown as the percent of drug remaining as a function of time at 25 degrees C.
Finding Ea for the Reaction
The activation energy (Ea) for a reaction can be determined by conducting stability studies at several different temperatures and applying the Arrhenius kinetic model. The slope of the line formed with Equation 2 contains Ea, as demonstrated in Figure 1. Higher temperatures (e.g., 55 degrees C, 70 degrees C) and corresponding shorter times (e.g., weeks, days) can be employed for this determination provided the Arrhenius kinetic model remains valid (see Limitations to Using Arrhenius Kinetic Model). The Ea for drug decomposition will usually fall in the range of 12 to 24 kcal/mole, with a typical value of 19 to 20 kcal/mole.
5 The activation energy can be approximated based upon prior knowledge of the drug decomposition kinetics. Once the Ea is known, it usually remains valid for use through small concentration changes or slight formulation changes.
Importance of Ea: Theoretical Model for Drug Loss
The following discussion demonstrates the relationship of drug degradation kinetics at 40 degrees C and 25 degrees C, where drug loss is the shelf life-determining parameter. For this illustration, acceptable product stability is based upon a lower limit of 90% label claim and the expiration date set at exactly 90% label claim. For this exercise, a zero-order degradation kinetics model (Δdrug/Δtime = —k) is applied to determine the rate for 10% of drug loss occurrence over 24 months at 25 degrees C (Figure 2). The slope was calculated and is equal to —0.4167 drug %label claim/ month, which is the reaction rate constant at 25 degrees C (k25). Hence, for the drug to remain within acceptance criteria for 24 months at 25 degrees C, the rate of degradation at 25 degrees C must be less than 0.4167 drug %label claim/month.
Figure 3. Zero-order kinetic models for 10% drug loss over three months and over six months at 40 degrees C.
In a similar manner, drug loss can be modeled at 40 degrees C accelerated temperature for the two scenarios of 10% drug loss occurring at three months and at six months (Figure 3). The slopes, which represent the reaction rate constant at 40 degrees C (k40), were calculated to be —1.6667 drug %label claim/month for the six months limit scenario and —3.3333 drug %label claim/month for the three months limit scenario.
Using the Arrhenius equation (Equation 3) and the reaction rates at 25 degrees C and 40 degrees C, the Ea can be calculated as shown in Table 1. The Arrhenius equation can be applied regardless of the order of the reaction kinetics. For the kinetic model that assumes 10% drug loss at 24 months at 25 degrees C (k25 = —0.4167 drug %label claim/month) and 10% drug loss at three months at 40 degrees C (k40 = —3.3333 drug %label claim/month), the calculated Ea is 26 kcal/mole. For the model that assumes 10% drug loss at 24 months at 25 degrees C (k25 = —0.4167 drug %label claim/month) and 10% drug loss at six months at 40 degrees C (k40 = —1.6667 drug %label claim/month), the calculated Ea is 17 kcal/mole.
Knowledge of the Ea is key to the interpretation of accelerated data for predictions at lower temperatures. If the Ea is low (less than or equal to 17 kcal/mole), the accelerated 40 degrees C drug concentration must remain greater than 90% label claim for six months to achieve at least 90% label claim for a shelf life of 24 month at 25 degrees C. If the Ea is high (more than or equal to 26 kcal/mole), the accelerated 40 degrees C drug concentration must remain greater than 90% label claim for only three months (and be 80 %label claim at six months), and yet the product will remain at or above 90% label claim for a shelf life of 24 month at 25 degrees C. This application of Ea and Arrhenius kinetics to predict shelf life can be applied to any drug level (limit) specified.
Table 1. Relationship of reaction rates at 40 degrees C and activation energy (Ea) for drug loss
Theoretical Model for Impurity/Degradant Generation
The same approach can be used to understand and predict the generation of impurities/degradants. For this exercise, a zero-order degradation kinetics model (ΔImpurity/Δtime = —k) is applied to determine the rates for an arbitrary 1% impurity growth at 25 degrees C and at 40 degrees C. The calculated slope for growth from 0 to 1.0% w/w over 24 months at 25 degrees C is equal to 0.0417% w/w per month, which is the impurity generation rate constant at 25 degrees C (k25). In the same manner, the impurity generation rate at 40 degrees C can be calculated for reaching 1.0% w/w after three months (0.3333% w/w per month) and for reaching 1.0% w/w after six months storage (0.1667% w/w per month). Table 2 summarizes the corresponding rates and Eas:
Example calculation
Where:
Ln k2/k1 = -Ea/R*(1/T2-1/T1) (Equation 3)
Ln (k40/k25) = - [Ea/1.987*(1/313-1/298)]/1000cal/kcal
k40 = 0.1667%w/w degradant/month
k25 = 0.0417%w/w degradant/month
Solving for Ea:
Ea = - [Ln (0.1667/0.0417)*1.987]/[(1/313-1/298)*1000]
Ea = 17 kcal/mole
Table 2. Relationship of reaction rates at 40 degrees C and activation energy (Ea) for impurity generation
As the Ea increases, the reaction rate at 40 degrees C increases. The degradant level at six months at 40 degrees C is predictive of the degradant level to be reached after 24 months at 25 degrees C when the Ea is 17 kcal/mole. Degradants with reaction mechanisms with high activation energies (greater than 17 kcal/mole) can exhibit levels exceeding 1.0% w/w by six months at 40 degrees C, yet have 25 degrees C values below the 1.0% w/w limit.
Shelf Life, Rates, and Activation Energy
To demonstrate the importance of Ea in predicting long-term shelf life at 25 degrees C from 40 degrees C rates, the Arrhenius kinetic equation was used to demonstrate the relationship of reaction rates and Ea. The relationships are demonstrated in Table 3 for the scenario where the degradant reaches the 1.0% w/w limit at six months at 40 degrees C and the zero-order rate constant (k40) is equal to 0.1667% w/w degradant/month.
Table 3. Relationship of rates and activation energy (Ea) for shelf-life stability when limit is reached after six months at 40 degrees C.
Table 4 demonstrates these relationships for the scenario where the degradant reaches the 1.0% w/w limit at three months at 40 degrees C and the zero-order rate constant (k40) is equal to 0.3333% w/w degradant/month. The zero-order degradant generation rates at 25 degrees C (k25) are calculated for reaching 1.0% w/w at different expiry time intervals: 12, 18, and 24 months (Tables 3 and 4). For these examples, the initial time zero starting level is set to 0% w/w. As shown in Tables 3 and 4, each of these different shelf-life scenarios has a reaction mechanism with a different Ea.
As the Ea decreases, the reaction rate at 25 degrees C increases relative to the rate at 40 degrees C. The temperature sensitivity of the reaction decreases with decreased activation energy, as indicated by the decreased proportionality between the rates (k40 α k25). When the activation energy is 9 kcal/mole (very low), even when remaining within the limit of 1.0% w/w degradation through six months at 40 degrees C, a maximum of 12 months at 25 degrees C shelf life can be projected.
In comparison, when activation energy is 17 kcal/mole with 1.0% w/w degradation through six months at 40 degrees C, 24 months elapses before the same 1.0% w/w degradant level is reached at 25 degrees C. If the Ea is known to be 17 kcal/mole or greater, then 40 degrees C values at six months of 1.0% w/w (or less) can kinetically support a shelf life, for this degradant limit, of 24 months at 25 degrees C. (This article does not address data accuracy—variability of the analytical methodology and/or the product samples—or the use of confidence intervals, which should also be incorporated when establishing product shelf life.)
Table 4. Relationship of rates and activation energy (Ea) for shelf-life stability when limit is reached after three months at 40 degrees C.
Because many drugs demonstrate Eas of 19-20 kcal/mole, this is the basis for the practice of comparing six month 40 degrees C values against the specification limit as a predictor of meeting that specification through a shelf life of 24 months at 25 degrees C. The proportionality of six months storage at 40 degrees C as predictive of 24 months at 25 degrees C is predicated upon an activation energy of at least 17 kcal/mole.
Some reactions proceed relatively fast at higher temperatures, as demonstrated in Table 4 and Figure 4. In this example, the degradant growth at 40 degrees C reaches the product limit (assigned here as 1.0% w/w) after three months storage (k40 = 0.3333%w/w degradant per month). The usual response is to assume that only a 12-month shelf life at 25 degrees C can be achieved. In reality, the levels of degradant reached at 25 degrees C are dependent upon the Ea of the reaction as shown in Figure 5.
The proportionality of the level measured after three months at 40 degrees C as predictive of the level for 12 months at 25 degrees C is valid only for an Ea of 17 kcal/mole. If the activation energy is known to be 22 kcal/mole or greater, then 40 degrees C values at three months up to 1.0% w/w can kinetically support a shelf life, for this degradant limit, of 18 months at 25 degrees C. This application of Ea and Arrhenius kinetics to the prediction of shelf life can be applied to any degradant level specified.
continues below…
CASE STUDY: Predict Dextrose Degradation
Figure A. Possible mechanism of 5-HMF formation in dextrose solutions (reference 10).
Figure B. Simplified reaction scheme for 5-HMF formation.
Table 5. Stability predictions based upon the ratio of rates (Ea) at 40 degrees C and 25 degrees C.
It is widely reported that dextrose in solution degrades to form 5-hydroxymethylfurfural (5-HMF) during heating (terminal sterilization) and over time.
6-8 According to the dextrose injection monograph in the USP, the limit for 5-HMF and related substances is not more than 0.25 absorbance units at 284 nm wavelength.
9
The multiple pathway formation of 5-HMF by dehydration of dextrose is depicted in Figure A.
10 The overall kinetic scheme, involving an intermediate, with definable rate constants for formation and reaction, has been published by at least three groups.
11 The Eas for k
1 and k
2 of the reaction scheme shown in Figure B, as determined by Sturgeon and colleagues, are 32.6 kcal/mole and 12.3 kcal/mole, respectively. Previously, Heimlich and Martin determined the Ea for 5-HMF formation from dextrose to be 31.2 kcal/mole and 31.8 kcal/mole, dependent upon the method used to determine the first-order rate constant.
11
Before appreciable formation of 5-HMF can occur, a reasonably high steady state level of the intermediate must first be established. Thus, the overall generation of 5-HMF will be dependent upon the k
1 with an activation energy of 32.6 kcal/mole. Using Table 5 and the Ea of 31 kcal/mole listing, the last column indicates that 5-HMF levels measured after two months at 40 degrees C will be predictive of 24 months at 25 degrees C. The production of
5-HMF is proceeding at least 12 times faster at 40 degrees C than at 25 degrees C. The level of 5-HMF after six months at 40 degrees C will be three times higher than the level reached at 24 months at 25 degrees C.
The U.S. Food and Drug Administration/International Conference on Harmonisation Guidance for Industry, Q1A(R2), Stability Testing of New Drug Substances and Products, requires long-term testing over 12 months at room temperature (25 degrees C or 30 degrees C) and over six months at accelerated conditions of 40 degrees C at the time of submission. The document’s Objections of the Guidance section states: “The guidance exemplifies the core stability data package for new drug substances and products, but leaves sufficient flexibility to encompass the variety of different practical situations that may be encountered due to specific scientific considerations and characteristics of the materials being evaluated. Alternative approaches can be used when there are scientifically justifiable reasons.”
12
The known chemistry of dextrose degradation to 5-HMF as presented here represents a justifiable reason for submitting a shorter time period of accelerated data. The kinetic model with known activation energy indicates that submission of two months at 40 degrees C data is sufficient and more appropriate than six months at 40 degrees C data to estimate the level of 5-HMF at 25 degrees C and to support a requested 24 months at 25 degrees C expiration dating period.
Figure 4. Kinetic model for 1% w/w degradant formation after three months at 40 degrees C.
Stability Prediction Made Easy
Now that the relationship of reaction rates to Ea is understood, it becomes easier to predict values over longer time periods at lower storage temperatures, like 25 degrees C, from values obtained at higher (accelerated degradation) temperatures, such as 40 degrees C.
The Ea is the proportionality factor between reaction rates at different temperatures (k40 α k25). The Arrhenius equation (Equation 4) can be solved for the exact relationship between reaction rates at 40 degrees C and 25 degrees C for any activation energy Ea. This proportionality can be used to predict levels and shelf life, as presented in Table 5. When the activation energy is 17 kcal/mole, the same degradant limit is reached at six months at 40 degrees C and at 24 months at 25 degrees C.
Table 5 can be used as a guide to interpret 40 degrees C kinetic prediction of 25 degrees C shelf life. As data is collected over time at 40 degrees C, the results at each test interval can be used to predict the level and, hence, the shelf life at 25 degrees C. When the activation energy is greater than 17 kcal/mole, samples stored at 40 degrees C and tested at six months will exhibit levels greater than what will be actually reached over 24 months at 25 degrees C. If the activation energy is 26 kcal/mole, the level measured at three months at 40 degrees C represents the expected level for 24 months at 25 degrees C, and the level measured at six months at 40 degrees C will actually represent twice the expected level for 24 months at 25 degrees C.
Figure 5. Predicted degradant formation at 25 degrees C when 1% w/w degradation is reached after three months at 40 degrees C for different activation energies (Ea).
The concept presented in Table 5 is shown in Figures 6 and 7. Figure 6 shows three different rate scenarios for the months at 40 degrees C to reach 1% w/w degradant. When the Ea for the reaction is 26 kcal/mole or greater, although the 1% w/w level is reached by three months at 40 degrees C (k40 = 0.3333% w/w degradant/month), the degradant growth at 25 degrees C will not reach 1%w/w until 24 months or beyond, as represented in Figure 7.
When the Ea for the reaction is 22 kcal/ mole and at four months at 40 degrees C the 1% w/w level is reached (k40 = 0.2500% w/w degradant/month), the degradant growth at 25 degrees C will again not reach 1%w/w until 24 months, as represented in Figure 7. An Ea of 17 kcal/mole and a six months 40 degrees C degradant level of 1% w/w (k40 = 0.1667% w/w degradant/ month) likewise corresponds to a projection of 24 months at 25 degrees C to reach 1% w/w. Thus, as demonstrated, different rates at 40 degrees C can project to the same or similar stability at 25 degrees C.
Additionally, as indicated in Table 5, for the same activation energy of 17 kcal/mole, when the degradation rate at 40 degrees C is faster, reaching 1% w/w by three months (k40 = 0.3333% w/w degradant/month), the 1%w/w level is projected to be reached by 12 months at 25 degrees C (Figures 4 and 5). A shelf life of 12 months is kinetically supported. Consequently, if the activation energy is lower than 17 kcal/mole for this same k40 rate of reaching 1% w/w by three months (k40 = 0.3333% w/w degradant/ month), levels at 25 degrees C will reach 1% w/w before 12 months, and a shelf life of 12 months is not kinetically supported. The information in Table 5 can be used to estimate shelf life based upon the Ea and stability at 40 degrees C.
Figure 6. Degradant reached 1%w/w at 40 degrees C; each model predicts the same level at 25 degrees C.
The practice of predicting long-term chemical stability at a lower temperature using data generated at a higher temperature over a shorter time period is based upon application of the Arrhenius kinetic model. The Arrhenius equation can be applied regardless of the order of the reaction kinetics. By using the Ea with the experimental level of drug or degradant obtained at a higher temperature (40 degrees C), the expected level of drug or degradant over long-term storage at a lower temperature (25 degrees C) can be more accurately predicted and the kinetic shelf life estimated. The proportionality of six months storage at 40 degrees C as predictive of 12 months at 25 degrees C is predicated upon an activation energy of 9 kcal/mole. The proportionality of six months storage at 40 degrees C as predictive of 24 months at 25 degrees C is predicated upon an activation energy of 17 kcal/mole. As the activation energy increases, the temperature sensitivity of the reaction increases, resulting in a greater difference in the rates at different temperatures. The information in Table 5 can also be used to estimate shelf life at 25 degrees C based upon the activation energy and the stability at 40 degrees C. As presented, the Arrhenius kinetic model can be applied to specific drug chemistry to scientifically justify appropriate months at accelerated 40 degrees C condition for submission.
Figure 7. Degradant level at 25 degrees C for the three activation energy models in Figure 6.
Michelle Duncan is an associate director in global research and development with Baxter Healthcare. She holds a bachelor’s in biochemistry and molecular biology from Northwestern University and earned her PhD in pharmaceutics from the University of Texas at Austin. She specializes in the formulation development of parenteral products and has led numerous injectables through FDA approval (NDA and ANDA routes) and to successful commercialization. Irene Zaretsky is a research associate with Baxter, supporting analytical and formulation development. She earned her master’s in chemistry from the Institute of Technology in Minsk, Belarus, and has worked more than 13 years in the U.S. pharmaceutical industry.
REFERENCES
- Martin A. Physical Pharmacy: Physical Chemical Principles in the Pharmaceutical Sciences. 4th ed. Philadelphia: Lea & Febiger; 1993: Chapter 12, Kinetics, pages 284-323.
- Lachman L, Lieberman HA, Kanig JL. The Theory and Practice of Industrial Pharmacy. 3rd ed. Philadelphia: Lea & Febiger; 1986:765-767.
- Wigent RJ. Chemical kinetics. In: Gennaro AR, ed. Remington: The Science and Practice of Pharmacy. 20th ed. Philadelphia: Lippincott Williams & Wilkins; 2000.
- Carstensen JT. Drug Stability: Principles and Practices. New York: Marcel Dekker; 1990:29-34.
- Conners KA, Amidon GL, Stella VJ. Chemical Stability of Pharmaceuticals: A Handbook for Pharmacists. 2nd ed. New York: John Wiley & Sons; 1986:24.
- Brönsted JN, Guggenheim EA. Contribution to the theory of acid and base catalysis: the mutarotation of glucose. J Am Chem Soc. 1927;49(10):2554-2584.
- Taylor RB, Jappy BM, Neil JM. Kinetics of dextrose degradation under autoclaving conditions. J Pharm Pharmacol. 1972;24(2): 121-129.
- Sturgeon RJ, Athanikar NK, Hartison HA, et al. Degradation of dextrose during heating under simulated sterilization. J Parenter Drug Assoc. 1980;34(3):175-182.
- U.S. Pharmacopeia–National Formulary. Dextrose injection monograph. Available at: www.uspnf.com.
- Hryncewicz CL, Koberda M, Konkowski MS. Quantitation of 5-hydroxymethylfurfural (5-HMF) and related substances in dextrose injections containing drugs and bisulfite. J Pharm Biomed Anal. 1996;14(4):429-434.
- Heimlich KR, Martin AN. A kinetic study of glucose degradation in acid solution. J Am Pharm Assoc. 1960;49(9):592-597.
- U.S. Food and Drug Administration. Center for Drug Evaluation and Research. Center for Biologics Evaluation and Research. Guidance for Industry: Q1A(R2) Stability Testing of New Drug Substances and Products. Washington, D.C.: U.S. Food and Drug Administration; 2003.
Editor’s Choice
- Weiss WF IV, Young TM, Roberts CJ. Principles, approaches, and challenges for predicting protein aggregation rates and shelf life. J Pharm Sci. 2009;98(4):1246-1277.
- Zahn M, Kållberg PW, Slappendel GM, et al. A risk-based approach to establish stability testing conditions for tropical countries. J Pharm Sci. 2006;95(5):946-965.
- Waterman KC, Adami RC. Accelerated aging: Prediction of chemical stability of pharmaceuticals. Inter J Pharm. 2005; 293:101-125.
- Beaman J, Whitlock M, Wallace R, et al. The scientific basis for the duration of stability data required at the time of submission. J Pharm Sci. 2010;99(6):2538-2543.
- Bauer M. Stability of drug substances and drug products: Considerations on the stability of drug substances and formulation in the pharmaceutical industry domain. STP Pharma Pratiques 2005;15(3):232-246.
In the previous article, we discussed the background for laboratory out-of-specification (OOS) investigations. In this article, we will discuss Phase I of the OOS investigation as outlined in the OOS Guidance.
For the following discussion, refer to Figure 1, which shows a flow diagram of the OOS laboratory investigation process based on the phases indicated in the OOS Guidance. Phases IA, IB, IIA, and IIB can be correlated to discussions in the OOS Guidance. Numbered steps are added in the figure and identified in the discussion below. It should be noted that two steps are in dashed boxes. These are activities outside of the laboratory OOS investigation and are shown only to clarify their relationship to the laboratory OOS investigation.
As we will see, the repeat test activity occurs only after the laboratory investigation has identified an attributable laboratory error and has closed. Phase IIA, review of production activities, is conducted by the production or quality unit. Ideally, these activities occur simultaneously with the laboratory investigation. In reality, the timing of the review of production depends on organizational politics and dynamics.
Laboratory management and personnel should aggressively and proactively take actions that prevent OOS (and out-of-trend) observations. Management is responsible for providing qualified, maintained, and calibrated equipment and instruments; clear, current, and understandable procedures; and training for staff at all levels. Analysts must always be alert and on the lookout for potential problems. Problems with systems supporting equipment, procedures, and training should be brought to the attention of management.
When an analyst is performing a test and observes something unusual that might result in an OOS test result, the analyst should stop and discuss the situation with a laboratory supervisor before going on. Any decision to stop an analysis and start over should be recorded in the analyst’s notebook and justified.
The laboratory OOS procedure should detail the complete laboratory process, starting with an OOS observation. Unfortunately, the OOS laboratory investigation process is complex and includes a number of decision, or branching, points. Effective, timely, and compliant resolution of OOS observations requires a clearly written OOS procedure and a staff thoroughly trained and knowledgeable in that standard operating procedure (SOP).
When an analyst completes a test, compares the result with the predefined limits, and realizes that he or she has an OOS observation (step 1), it is immediately reported to the laboratory manager/supervisor and entered into the OOS logbook (step 2). An OOS logbook and the OOS observation logging procedure should be defined in, and required by, the OOS procedure. Although the OOS logbook and processes related to the logging of OOS observations are not discussed in the OOS Guidance, they are a regulatory expectation. This logbook is often one of the first things requested by an inspection team that includes the laboratory control system in the inspection. Because it is one place in which problems are recorded and will therefore be inspected closely, the logbook should be maintained meticulously.
Entry of the OOS observation into the log signals the beginning of the OOS investigation. The OOS procedure should define a maximum time for the completion of an OOS investigation. Many laboratory OOS procedures include a requirement that the laboratory investigation be completed within 20 working days. This expectation appears to be acceptable to the regulatory investigators and was originally derived from testimony in Judge Alfred M. Wolin’s 1993 decision in the Barr Laboratories case and an FDA inspection guidance. The clock starts running with the entry into the OOS log.
Immediately, the supervisor and analyst move to Phase IA (step 3). This step involves the supervisor’s thorough review of the testing process, the materials and equipment used in the testing process, and the analyst’s actions. Most organizations have a checklist for this review. The checklist may be a freestanding form, or it may be included in the OOS SOP. It is based on best industry practice and internal experience; there is no one-size-fits-all checklist. The supervisor should not create the checklist on the fly. Investigators evaluate checklists, as evidenced by the following observation:
The investigational checklist you currently use is insufficient to detect and evaluate instrument problems and standard/sample preparations errors.
In some laboratories, this review is treated as a mechanical, superficial operation without attention to detail, executed only for the purpose of completing the checklist. It is important that the review be thorough. This supervisor review includes the following steps:
- Discuss the method with the analyst; confirm analyst training and knowledge and performance of the correct procedure.
- Examine the raw data obtained in the analysis, including chromatograms and spectra, and identify unidentified peaks and anomalous or suspect information.
- Verify that the transcription of any data between records is accurate and that the calculations used to convert raw data values into a final test result are scientifically sound, appropriate, and correct; also determine if automated calculation methods are appropriately validated and if unauthorized or unvalidated changes have been made to automated calculation methods.
- Confirm the performance of the instruments, particularly that they have been qualified for use for this test and that they are in a current state of maintenance and calibration.
- Determine if appropriate reference standard, solvents, reagents, and other solutions were used, and that they meet quality control specifications and were properly controlled since receipt.
- Evaluate the performance of the test method to ensure that it is performing according to the standard expected based on method validation data and historical data. This includes comparing the method performance and spectra or chromatograms obtained during the test to examples in the test method or reported in the method validation report and determining that system suitability results meet acceptance criteria.
Throughout this step, the analyst and supervisor are looking for sound evidence that a laboratory error is the attributable cause of the OOS observation. They must record their actions and findings. Examples of attributable causes include the following:
- The analyst used the wrong mobile phase in the preparation of the samples;
- The analyst used the wrong pipette in a dilution step;
- The analyst set the wrong wavelength on the spectrophotometer;
- The oven controls for the gas chromatography column chamber malfunctioned;
- The spectrophotometer light source malfunctioned;
- Test procedure steps are not correct;
- There was a spike in the laboratory power at the time of the analysis;
- The reference material has expired;
- The wrong phosphate salt was used in the preparation of one of the required solutions; and
- A value was incorrectly transcribed from the instrument printout to the spreadsheet used for calculation.
Figure 1. An out-of-specification process chart.
The investigation must identify a specific attributable cause and be supported by data. If the analyst and supervisor are able to identify an attributable cause of the OOS observation (step 4), the investigation and conclusion are recorded in a format defined in the OOS SOP (step 8). The report should include what was done, the conclusion, and the corrective action and recommended preventive action(s) (step 7). In general, the corrective action requires the analyst to repeat the test (step A). The preventive action is an action that will prevent the attributable cause from occurring again. One preventive action that is often recommended is analyst retraining; this has been overused over the past decade and should be recommended cautiously.
If the attributable cause is the fact that the analyst did not follow a procedure, the root cause is probably deeper; the procedure and the training the analyst received should be evaluated. The root cause may be buried in the knowledge the analyst has been provided. If the laboratory investigation checklist is designed well, the completion of this form may serve as the laboratory investigation report (step 6). When the report has been reviewed and approved by the quality unit (step 7), the laboratory investigation is complete and closed in the OOS log (step 18).
When Phase IA (steps 3 and 4) of the laboratory OOS investigation does not identify the attributable laboratory error, the investigation moves to Phase IB (step 5). This part of the investigation is accomplished by the analyst under the close supervision of the supervisor. While the focus in Phase IA is on the equipment, solutions, reagents, and procedures, the focus of Phase IB is on the samples and analytical preparations that resulted from the process. This part of the investigation depends on having the test sample and standard preparations available.
Many laboratories have a policy that the analyst will not discard test sample and standard preparations until the test results have been reviewed and approved by the designated reviewer. This policy supports the possible need for the sample preparations for a laboratory OOS investigation. Even if the test sample and standard preparations are available, the investigation must consider preparation stability. Standard and sample preparation stability should be identified in the test procedure. If it cannot be found there, it should be in the test method validation report. When the laboratory does not have data that supports the stability of the sample and standard preparations, it is difficult to draw any scientifically sound conclusions based on the examination of these preparations. Because test sample and standard preparation stability information usually covers a limited time, it is important that the investigation move through this step as quickly as possible.
In repeat testing, an analyst repeats the test exactly as described in the test method, taking the analytical sample from the original sample. Typically, the repeat test is performed by the original analyst.
The analytical preparation that yielded the OOS is reanalyzed. If the test is based on chromatography, this analytical preparation is reinjected into the chromatographic system under the conditions defined in the test procedure. If the test is based on spectrophotometry, the analytical sample is introduced into the spectrophotometer sample chamber at the test conditions. An instrument readout that differs from the one that produced the OOS result is an indication of an instrument malfunction and warrants further investigation.
Other intermediates of, or extraction residues from, the analytical sample preparation may be available. These should be examined or reprocessed for any clue that would point to an attributable cause of the OOS observation.
As with Phase IA, the objective of Phase IB is to identify an attributable cause for the OOS observation (step 6). If that attributable laboratory error is identified, the original laboratory results are invalidated, with appropriate notation made in the analyst’s notebook or worksheet. Results of the investigation and recommended preventive action (step 7) are identified in the investigation report (Step 8), which is reviewed and approved by the quality unit (step 9). The OOS investigation is closed in the OOS log (step 18), and the laboratory moves to repeat testing (step A).
When the laboratory identifies an attributable cause, it should be entered into an appropriate system and tracked to determine if there are any trends that justify a preventive action. As stated in the OOS Guidance:
As part of an effective quality system, a firm’s upper management should appropriately monitor these trends to ensure that any problematic areas are addressed.
In repeat testing, an analyst repeats the test exactly as described in the test method, taking the analytical sample from the original sample. Typically, the repeat test is performed by the original analyst. Some organizations, in their OOS SOP, require that a different analyst perform the repeat test. Other organizations define an extensive, multiple-analyst scheme for retesting. The OOS SOP should clearly define the repeat testing process for the laboratory, and, once established, the procedure should be faithfully followed.
When the laboratory cannot identify an attributable laboratory error to explain the OOS observation through Phase I (A and B, steps 3 and 4, and steps 5 and 6), there is still the regulatory expectation that the attributable cause be determined and a preventive action implemented to support process improvement. There is also still the organizational hope that the laboratory was wrong and the material will eventually be released. The laboratory investigation moves to Phase II, which will be discussed in part three of this series.
Por: John G. (Jerry) Lanese is an independent consultant with a focus on quality systems and the components of an effective quality system. In 1994, he formed The Lanese Group and consults on quality systems and cGMP compliance for small and large medical device and pharmaceutical companies, including companies under FDA Consent Decree, API and excipient manufacturers, electronic firms, device component manufacturers, and other manufacturing organizations.