JP2009244112A - Stability prediction method for medical product - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately perform stability prediction of a medical product by analyzing a near-infrared spectrum thereof. <P>SOLUTION: The stability prediction method comprises: generating a calibration model (S14) by performing multivariable analysis using a near-infrared spectrum before storage measured with respect to a prepared specimen (S10) and a specific index of the specimen actually measured after storage for a predetermined period in a specific condition (S12); and calculating a prediction value of the specific index (S18) by applying near-infrared spectrum data of a medical product that is a stability prediction object (S16) to the calibration model. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a nondestructive analysis technique using a near-infrared spectrum, and more particularly to a technique for predicting the stability of a pharmaceutical preparation.

  In recent years, a method using a near infrared spectrum obtained by irradiating a target with near infrared light has attracted attention as a chemical analysis method based on nondestructive analysis (Patent Documents 1 to 4). Near-infrared light has a longer wavelength than visible light and is an electromagnetic wave in the short wavelength range of infrared light. It is used for nondestructive analysis in a wide range of fields such as food and agriculture.

  For example, Patent Documents 1 and 2 disclose that operation control of a production plant is performed with high accuracy using a near-infrared spectrum measured in real time. Patent Document 3 discloses that the properties and performances of an unknown lubricant can be estimated by measuring the near-infrared spectrum of the lubricant and applying it to a predetermined calibration model. Patent Document 4 discloses that a grain sample is irradiated with a near infrared spectrum to evaluate the internal quality of the grain.

JP 2000-298512 A

Special table WO02 / 012969

JP-A-11-194124

Japanese Patent Laid-Open No. 06-144264

  However, in any of Patent Documents 1 to 4, although the evaluation of the object by near infrared analysis is performed, these are merely “evaluation of the object at the time of spectrum measurement”, and over time. There was no one that predicted changes.

  On the other hand, the stability of pharmaceutical preparations is generally evaluated by subjects that are susceptible to various environmental factors such as humidity, temperature, and light due to storage, or that affect quality, safety, or effectiveness. However, when it is determined by the ratio of the amount of related substances in the entire vial (for example, a standard for related substances is set in advance, and a preparation in which the amount of related substances after storage for a certain period exceeds the standard does not satisfy the stability. There are standards such as defective products.) Regarding the pharmaceutical preparations, the amount of related substances immediately after production may not be directly related to the subsequent stability, and it seems difficult to predict stability. There were various factors.

  FIG. 19 is a table comparing the initial values of the amounts of related substances actually measured from three lots in which a certain pharmaceutical preparation was manufactured, and the amounts of related substances actually measured after the preparations of each lot were stored under a certain condition. FIG. 20 is a graph showing the relationship between the storage period and the amount of related substances in the long-term storage test shown in FIG.

  As shown in FIG. 19 and FIG. 20, the initial values of the related substance amounts are almost the same in the preparation lots 2 and 3, but after 12 months of the long-term storage test, the related substance amounts (average value, maximum amount and (Range) is different. On the other hand, regarding the preparation lots 1 and 3, the initial value of the related substance amount is different, but the value of the related substance amount is equal in the long-term storage test. In the accelerated test shown in FIG. 19, the amounts of the related substances in the preparation lots 2 and 3 show almost the same value, but only the preparation lot 1 shows a different high value.

  As described above, the relationship between the amount of the related substance immediately after production and the amount of the related substance after storage cannot be found, and it may be difficult to predict the stability. On the other hand, the standards for the stability of pharmaceutical preparations are clearly defined by laws and regulations, and an effective means for accurately predicting the stability of pharmaceutical preparations after storage has been demanded.

(1) The stability prediction method of the present invention is:
A stability prediction method for predicting the stability of a pharmaceutical preparation,
Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time A calibration model generation step for generating a calibration model in advance by performing variable analysis,
A near-infrared spectrum measurement step for obtaining a near-infrared spectrum data by measuring a near-infrared spectrum before storage for a specimen of a pharmaceutical preparation that is a stability prediction target,
By applying the near-infrared spectrum data obtained in the near-infrared spectrum measurement step to the calibration model generated in the calibration model generation step, after storing the specimen under the same or similar conditions as the calibration model A specific index value calculating step of calculating a predicted value of the specific index in
Stability determination that determines whether or not a specimen satisfies stability after a predetermined period of time by comparing the predicted value of the specific index obtained in the specific index value calculation step and the stability evaluation criteria of the pharmaceutical preparation Process,
It is characterized by having provided.

  Thereby, the stability of a pharmaceutical formulation in the future can be accurately predicted using a specific index calculated by a nondestructive inspection using a near infrared spectrum.

(2) The stability prediction method of the present invention is:
A stability prediction method for predicting the stability of a pharmaceutical preparation,
Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time A calibration model generation step for generating a calibration model in advance by performing variable analysis,
A near-infrared spectrum measurement step for obtaining a near-infrared spectrum data by measuring a near-infrared spectrum before storage for a specimen of a pharmaceutical preparation that is a stability prediction target,
By applying the near-infrared spectrum data obtained in the near-infrared spectrum measurement step to the calibration model generated in the calibration model generation step, after storing the specimen under the same or similar conditions as the calibration model A specific index value calculating step of calculating a predicted value of the specific index in
It is characterized by having provided.

  This makes it possible to accurately calculate an index for predicting the stability of a pharmaceutical preparation in the future by a nondestructive inspection using a near infrared spectrum.

(3) The stability prediction method of the present invention is:
The wave number region of the near-infrared spectrum applied to the calibration model in the specific index value calculation step is in the range of 7500 cm ^{−1 to} 4400 cm ⁻¹ .
It is characterized by that.

  As a result, an index for predicting the stability of a pharmaceutical preparation in the future can be calculated with general-purpose and high accuracy by nondestructive inspection using a near-infrared spectrum.

(4) The stability prediction method of the present invention is:
The wave number region of the near infrared spectrum applied to the calibration model in the specific index value calculation step is determined by selecting a wave number region having a high correlation coefficient.
It is characterized by that.

  Thereby, an index for predicting the stability of a pharmaceutical preparation in the future can be calculated effectively and accurately by a nondestructive inspection using a near-infrared spectrum.

(5) The stability prediction method of the present invention is as follows:
The pharmaceutical preparation is amrubicin hydrochloride lyophilized preparation,
The specific indicator is the amount of deglycosides that are related substances.
It is characterized by that.

  In this way, the amount of related substances (amount of deglycosides), which is an index for predicting the stability of amrubicin hydrochloride freeze-dried products in the future, should be calculated accurately by nondestructive testing using near-infrared spectra. Can do.

(6) The method for producing the pharmaceutical preparation of the present invention comprises:
A method for producing a pharmaceutical preparation using the stability prediction method,
In the stability determination step, samples that are determined not to satisfy the predetermined stability criteria are excluded from the production line.
It is characterized by that.

  This makes it possible to accurately predict the future stability of pharmaceutical preparations using a specific index calculated by nondestructive inspection using the near-infrared spectrum, and remove medicines that do not meet the stability criteria from the production line. .

(7) The stability prediction method of this invention is
Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time A stability prediction method for predicting the stability of a pharmaceutical preparation using a computer having a storage unit that stores a calibration model generated by performing a random analysis,
A near-infrared spectrum measurement step for obtaining a near-infrared spectrum data by measuring a near-infrared spectrum before storage for a specimen of a pharmaceutical preparation that is a stability prediction target,
By applying the near-infrared spectrum data obtained by the near-infrared spectrum measurement step to the calibration model stored in the storage unit, the sample after storing the specimen under the same or similar conditions as the calibration model A specific index value calculating step for calculating a predicted value of the specific index,
It is provided with.

  This makes it possible to accurately calculate an index for predicting the stability of a pharmaceutical preparation in the future by a nondestructive inspection using a near infrared spectrum.

(8) The stability prediction method of the present invention is:
A stability prediction method for predicting the stability of a pharmaceutical preparation,
Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time A calibration model generation step for generating a calibration model in advance by performing variable analysis,
A near-infrared spectrum measurement step for obtaining a near-infrared spectrum data by measuring a near-infrared spectrum before storage for a specimen of a pharmaceutical preparation that is a stability prediction target,
By applying the near-infrared spectrum data obtained in the near-infrared spectrum measurement step to the calibration model generated in the calibration model generation step, under the same conditions as the calibration model (that is, the same storage conditions (temperature , Humidity), a period different from the storage period in the calibration model generation process) a specific index value calculation step for calculating a predicted value of the specific index after the specimen is stored,
It is provided with.

  This makes it possible to accurately calculate an index for predicting the stability of a pharmaceutical preparation in the future by a nondestructive inspection using a near infrared spectrum.

  In this embodiment, “before storage” means a state immediately after manufacturing a pharmaceutical preparation until a chemical or physical change occurs that affects the near-infrared spectrum of an object. . “Stability” is the degree to which quality is maintained under the influence of various environmental factors such as temperature, humidity, and light.

  In this embodiment, the “same condition” means that the storage period and the storage conditions (temperature, humidity) are the same between the sample for generating the calibration model and the sample for which stability is predicted, “Similar conditions” means that the storage conditions (temperature, humidity) are the same and the storage periods are different between the sample for generating the calibration model and the sample whose stability is to be predicted.

[Hardware configuration of stability prediction device]
FIG. 1 is a diagram showing a hardware configuration of a stability prediction apparatus 100 that executes the stability prediction method of the present invention. As shown in FIG. 1, the stability prediction apparatus 100 includes a near-infrared spectrum measuring means 102 for measuring a near-infrared spectrum of an object (a sample of a pharmaceutical preparation) contained in a vial 103, and a near-infrared spectrum. The processing computer 104 is connected to the measuring means 102.

  The near-infrared spectrum measuring means 102 shown in FIG. 1 is a spectrophotometer capable of transmitting / receiving data to / from the processing computer 104 and generally available on the market. As a measurement method of the near-infrared spectrum, there are transmission measurement, diffuse reflection measurement, and the like, but in the case of a liquid agent, it is preferable to perform transmission measurement, and for a thick solid preparation, it is preferable to perform diffuse reflection measurement. .

  As shown in FIG. 1, the processing computer 104 includes a CPU 10, a RAM 12, a display 14, a hard disk 16, a keyboard / mouse 18, a communication circuit 20, and the like. It is an apparatus for performing various processes to be described later based on infrared spectrum data.

  Further, as shown in FIG. 1, a stability data calculation program 30 and a calibration parameter DB 32 are stored in the hard disk 16. The installation of the stability data calculation program 30 or the like on the hard disk 16 is performed by reading a program recorded on a recording medium (not shown).

  The stability data calculation program 30 includes a calibration model (calibration formula) generation process based on the chemometrics method, a stability data calculation process using the calibration model, and whether or not the sample satisfies a predetermined stability criterion. Commercially available software for performing the process of determining

The calibration parameter DB 32 shown in FIG. 1 stores parameters for generating a calibration model as shown in FIG. The data1 (X1, X2,... Xk) shown in FIG. 2 is the absorbance value of the near infrared spectrum at each predetermined wave number (V1, V2,... Vk), and the near infrared spectrum for each of the sample numbers 1 to N. A value measured by the measuring means 102 is stored. data2 (M ₁ , M ₂ ,... M _n ) is an actual measurement value M of a specific index actually measured after each sample is stored. data3 (W11, W21, ... Wk1, etc.) is a coefficient (loading spectrum value) indicating the weight at each wave number (V1, V2, ... Vk), and data4 (B0, B1, ... Bn) is a calibration model (calibration formula) ) Is a partial regression coefficient. Note that the number of principal components (described later) that is optimal for regression is set in n of Bn. Details of each parameter item and calculation method shown in FIG. 2 will be described in the following flowchart.

[Flowchart of stability prediction processing]
FIG. 3 shows a flowchart of processing performed by the program 30 of the stability prediction apparatus 100 shown in FIG. The target of stability prediction in the present invention is a pharmaceutical preparation. In particular, in Examples 1 to 3 to be described later, an experiment was conducted with an amrubicin hydrochloride freeze-dried preparation for injection as a specific example.

First, when a sample for generating a calibration model is prepared in the near-infrared spectrum measuring means 102, a near-infrared spectrum before storage is measured for the prepared sample based on an operation input by an operator (step). S10). The near-infrared spectrum data data1 measured here is transmitted from the near-infrared spectrum measuring means 102 to the processing computer 104, and the absorbance X (FIG. 5) is associated with each preset wave number (V1, V2,... Vk). 2 (X ₁ 1, X ₁ 2,... X ₁ k, etc.) are stored in the calibration parameter DB 32 of the hard disk 16.

  FIG. 4 shows the original spectrum of the near-infrared spectrum data data1 measured in step S10. In FIG. 4, a plurality of near-infrared spectrum data data1 measured from a large number of vials 103 are displayed in an overlapping manner. In order to create an appropriate calibration model, a sample for generating a calibration model can be one lot, but it is preferable to use a plurality of lots, for example, using three or more lots. . Moreover, it is preferable that the number of specimens (vials) per lot is large, and it is usually preferable to use 5 vials or more.

  Next, after the specimen for generating the calibration model is stored under a specific condition for a predetermined period, the specific index of the specimen is actually measured (that is, the conventional analysis method is measured), and the actually measured values M and data2 (FIG. 2) are obtained. Are stored in association with each sample (step S12).

  Specifically, after measuring the near-infrared spectrum before storage in step S12, the operator temporarily moves the specimen to another location, during a predetermined storage period (from the first time point to the second time point), and Store under specific storage conditions (eg, temperature 40 ° C./75% RH). Thereafter, the operator inputs the actually measured value data obtained by actually measuring the specific index through the keyboard / mouse 18 and stores it in the calibration parameter DB 32 of the hard disk 16. As an actual measurement method (measurement by a conventional analysis method) of the specific index used here, for example, HPLC or the like can be used for the amount of related substances.

  In addition, as a specific index that is a target of actual measurement (measurement by a conventional analysis method) in the present invention, a numerical value that changes with the passage of time of storage and is a quality index, that is, an index that is a standard for stability evaluation Good. A specific example of the specific index is, for example, the amount of a related substance. Particularly, when the pharmaceutical preparation is an amrubicin hydrochloride freeze-dried preparation, the amount of deglycosides which is one of the related substances is preferable. These specific indicators need to be appropriately selected for each target pharmaceutical preparation.

  It should be noted that a series or each process of storing the sample vial, actual measurement after the storage, and data input may be automatically performed mechanically without an operator. In addition, the actual measurement (step S12) of the specific index can be performed in parallel with the measurement of the near infrared spectrum (step S10) or before the measurement of the near infrared spectrum (step S10). .

  The stability data calculation program 30 (FIG. 1) of the processing computer 104 further shows the near-infrared spectrum data data1 (FIG. 2) before storage obtained in step S10 and the specific index after storage obtained in step S12. By performing multivariate analysis using the actual measurement data data2 (FIG. 2), a calibration model (calibration formula) is generated and stored in the hard disk 16 (step S14).

The calibration model is the difference between the near-infrared spectrum data created by analyzing with the chemometrics method (multivariate analysis) based on the slight difference between the measured near-infrared spectra and the analytical value of the measured component. It is a mathematical relational expression. Specifically, the absorbance data X (X1, X1, FIG. 2) at each wave number V (V1, V2,... Vk) of the analysis region determined according to the pharmaceutical preparation and purpose from the near infrared spectrum obtained by the near infrared analysis. X2, ... Xk), and multivariate analysis (for example, data processing by PLS method) using these data and measured values M of conventional analysis method (M ₁ , M ₂ , ... M _{N in} FIG. 2) This is a correlation equation (composed of parameters data3 and data4 in FIG. 2) obtained by performing this, and is expressed by the following equation (β).

Y = B0 + B1 * Z1 + B2 * Z2 +... + Bn * Zn (β)
Here, the objective variable Y is a value to be predicted, that is, an amount of a related substance. The partial regression coefficient B (B0, B1, B2,... Bn, data4 in FIG. 2) is a weight applied to the explanatory variable Z (Z1, Z2,... Zn), and is a difference (residual) of 2 from the actual measurement value M. It is determined by the method of least squares so that the sum of multipliers is minimized.

The explanatory variable Z (Z1, Z2,... Zn) is expressed by the Z value of the n-th principal component in the principal component analysis, Z = X1 * W1 + X2 * W2 + ... + Xk * Wk (X: Absorbance of near infrared spectrum, W: coefficient). Here, the coefficient of the first principal component (W1, W2,... Wk) is the first principal component, with the axis in the direction of the largest data dispersion (the direction of the largest difference in absorbance) as the first principal component. Z1 = X1W11 + X2W21 + ... + XkWk1 and the values of absorbance X (X1, X2,... Xk) are substituted, the values of Z _{1 1} to Z _N 1 calculated for each sample (sample numbers 1 to N), [ It can be obtained by calculating a value that maximizes the variance of equation (1).

  Further, the coefficient W for each principal component up to the n-th principal component is obtained as necessary and stored in the hard disk 16. Note that the calibration parameter DB 32 shown in FIG. 2 shows only the coefficients W (W11, W21,... Wk1) calculated for the first principal component.

  Steps S10 to S14 described above are steps until a calibration model (calibration equation) is generated based on near-infrared spectrum data measured before storage and actually measured values measured after storage. In subsequent steps, the stability of the pharmaceutical preparation specimen is predicted based on the calibration model.

  The CPU 10 of the processing computer 104 controls the near-infrared spectrum measuring means 102 so as to measure the near-infrared spectrum for the specimen of the pharmaceutical preparation for which the prepared stability is to be predicted (step S16). As a result, the processing computer 104 receives the near-infrared spectrum data, and the value of the absorbance X ′ (X′1, X′2,... X′k) associated with each wave number (V1, V2,... Vk). Is stored in the hard disk 16.

  Further, the CPU 10 of the processing computer 104 applies the near-infrared spectrum obtained in step S16 to the calibration model generated in step S14, thereby storing the specific index after storing the specimen under the same conditions as the calibration model. A predicted value M ′ is calculated (step S18). Specifically, the absorbance X ′ values (X′1, X′2,... X′k) of near-infrared spectrum data measured from specimens of pharmaceutical preparations whose stability is to be predicted are calculated using a calibration model (see above). Substituting into the calibration formula (β)), the value of Y is calculated. The value of Y becomes the predicted value M ′ of the specific index.

  Finally, the CPU 10 of the processing computer 104 determines whether or not the predicted value M ′ of the specific index satisfies the stability criterion (step S <b> 19), and the pass / fail result is displayed on the display 14. Specifically, data indicating a predetermined stability criterion input by an operator or stored in advance in the hard disk 16 is read, and when the predicted value M ′ of the specific index exceeds a predetermined reference value, the stability is increased. It is determined that it is not satisfied, and if it is within a predetermined reference value, it is determined that the stability is satisfied. Note that the processing computer 104 can perform the processing up to step S18, and an operator who sees the predicted value M ′ of the specific index calculated in step S18 can make the determination.

  As described above, once the calibration model is generated, the predicted value Y can be easily calculated by applying the near-infrared spectrum data X of the object to be predicted to the calibration model, and the stability of the pharmaceutical preparation in the future. Prediction can be performed with high accuracy.

In general, the wave number region of near-infrared light is 12500 cm ^{−1 to} 3600 cm ⁻¹ , but a part of the calibration model may be used for evaluation, and in particular 7500 cm ^{−1 to} 4400 cm ^−1. This range is considered preferable because it is versatile and accurate. This is considered to be because the prediction accuracy is improved by omitting it because the noise is relatively large outside the range of 7500 cm ^{−1 to} 4400 cm ⁻¹ . Also, as shown below, it is possible to select two or more specific wave number regions as the specific wave number region.

The difference in correlation (correlation coefficient) when the specific wavenumber region (analysis region) of the near-infrared spectrum to be analyzed is different will be described with reference to FIG. 6, every sample of a same pharmaceutical formulations stored for 12 months at 20 ℃ / 60% RH, the analysis region of the near infrared ^{spectrum, 7500cm -1 ~4400cm -1, 7200cm -1} ~6900cm -1, 5400cm - ^{1 ~4400cm -1,} when the ^{7200cm -1 ~6900cm} -1 and ^{7200cm -1 ~6900cm} -1 (2 single wavenumber region) is a table showing the difference in the correlation coefficients obtained as an experimental result.

As shown in FIG. 6, as a result of creating a calibration model from each of the wavenumber regions, the correlation coefficient R was the highest at 0.8606 when the evaluation was performed in a wide analysis region 7500 cm ^{−1 to} 4400 cm ⁻¹ . It was confirmed that higher correlation was obtained. As shown in FIG. 6, the next highest correlation coefficient R is 5400 cm ^{−1 to} 4400 cm ⁻¹ , and many factors that affect stability prediction are included in the wave number region of 5400 cm ^{−1 to} 4400 cm ^−1. It is thought that. ^However, the wave number peak in particular greater contribution in 5400cm ^-1 ~4400cm -1 (e.g. FIG. ^5A, 5240cm -1 as shown in B, 5120cm ^-1) near but I went similar experiments alone, the correlation coefficient R The value was about 0.5, and it was found that good results as in the case of analyzing the above wide range of wave numbers could not be obtained. 5A is a diagram showing an example of data obtained by differentiating the near-infrared spectrum (original spectrum) of the object shown in FIG. 4, and FIG. 5B is a graph showing a loading spectrum corresponding to FIG. 5A. The analysis region was selected by extracting a portion having a large absolute value of the loading spectrum shown in FIG. 5B as a portion contributing to the correlation of the calibration model.

  As described above, the reason for obtaining good results when the spectrum is evaluated over the wider wavenumber region is that the wider wavenumber region makes the physicochemical characteristics such as chemical structure and heterogeneity affecting stability. This is considered to be because a lot of information can be used for evaluation.

  Next, the reason why the near-infrared spectrum was measured before storage (immediately after manufacture) but not after storage in the present invention will be described with reference to FIG. FIG. 7 is a table showing the correlation between the amount of related substances predicted from the near-infrared spectrum measured after storage and the correlation between the amounts of related substances predicted from the near-infrared spectrum measured before storage (immediately after production). It is. FIG. 7 shows three different conditions for the same pharmaceutical preparation (amrubicin hydrochloride freeze-dried preparation) (Example a is 6 months at 25 ° C./60% RH, Example a is 12 months at 25 ° C./60% RH). Example c shows the results when stored at 40 ° C./75% RH for 6 months).

  As shown in FIG. 7, the result was that the amount of related substances expected by measuring and analyzing the spectrum after storage had a significantly lower correlation coefficient than when the spectrum before storage was measured. This is presumably because the freeze-dried preparation is an unstable preparation, so it is desirable to measure the near-infrared spectrum immediately after production, and the correlation becomes low after storage. In addition, physicochemical characteristics such as chemical structure and heterogeneity appear in the near-infrared spectrum immediately after production, but after storage, such an original state is caused by factors such as an increase in moisture and an increase in other chemical substances. It is also speculated that will collapse.

[Example 1]
Example 1 is an example in which a calibration model is generated from 2 lots of amrubicin hydrochloride freeze-dried preparation (50 mg preparation, 3 lots) and the stability of the remaining 1 lot is predicted.

  The lyophilized preparation can be produced, for example, by the method described in WO2004 / 050098. That is, amrubicin or a salt thereof, L-cysteine or a salt thereof and, if necessary, an excipient, etc. are dissolved in distilled water for injection, pH is adjusted with a small amount of base and / or acid, and a sterile filtered liquid is then added to a vial. It is filled into a bottle and freeze-dried to prepare a preparation in a powder state. The injection is stored in this state, dissolved in water at the time of use, and used for administration. A freeze-dried preparation containing amrubicin or a salt thereof can be used as a cancer chemotherapeutic agent for the treatment of various cancer diseases.

The near-infrared spectrum was measured with a near-infrared spectrophotometer for a total of 18 lots for a calibration model of the pharmaceutical preparation, 18 in total. NIR analysis by Bruker Optics Ltd. "near infrared spectrophotometer MPA" (TM), carried out in the range of analysis areas ^{7500cm -1 ~4400cm} -1 diffuse reflection measurement, near infrared obtained First-order differentiation processing and normalization processing were performed on the outer spectrum.

  The sample was stored at 25 ° C./60% RH for 6 months, and the amount of related substances was measured (measurement of the amount of deglycosides by the conventional analysis method) by liquid chromatography, which is a conventional analysis method. Of the 3 lots prepared, 2 lots were used.

  From the above, using the near-infrared spectrum obtained by near-infrared analysis and the amount of related substances obtained by liquid chromatography (HPLC), the “OPUS spectroscopic software” (trademark) manufactured by Bruker Optics, Inc. A calibration model was generated by the PLS method. FIG. 8 shows a calibration curve drawn by the calibration model generated in Example 1 and a plot of the predicted value and the actual measurement value.

  Furthermore, the near-infrared spectrum was measured before storage for a total of 18 specimens of unknown stability of the pharmaceutical preparation (remaining 3 prepared lots), and the above-mentioned near-infrared spectrum data and the above-mentioned were prepared. Based on the calibration model, the amount of the related substance after “25 ° C./60% RH-6 months” storage of the specimen was predicted. For the purpose of evaluating the amount of the related substance after storage and the predicted value, the above-mentioned sample with unknown stability is stored at “25 ° C./60% RH—6 months” and HPLC (liquid chromatography) which is a conventional analysis method. Thus, the amount of related substances was measured, and the amount of related substances and measured values were obtained.

  FIG. 9 and FIG. 10 show the predicted value and the measured value of the related substance amount after storage and the correlation thereof, respectively. As a result of the experiment, as shown in FIG. 10, the correlation was high as a whole (correlation coefficient R = 0.7459), and the stability prediction was performed with high accuracy.

[Example 2]
In the same manner as in Example 1, the storage period was changed to 12 months, and the calibration curve drawn by the created calibration model and the evaluation values and the actual measurement values were evaluated in FIG. The result of the predicted value and the actual value of the substance amount and the correlation thereof are shown in FIGS. 12 and 13, respectively.

  As shown in FIG. 13, in Example 2, as in Example 1, the overall correlation is high (correlation coefficient R = 0.8668), and the result that the stability prediction is performed with high accuracy is obtained. It was.

[Example 3]
A calibration model created using two calibration sample lots, a total of 30 specimens and changing the storage conditions and period to “40 ° C./75% RH-1 month” in the same manner as in Example 1. FIG. 14 shows evaluation results of the calibration curve and predicted values and actual values drawn by Fig. 15, and FIG. 15 and FIG. 16 show the results of the predicted values and actual values of related substances after storage and their correlations, respectively.

  As shown in FIG. 16, in Example 3, as in Examples 1 and 2, the overall correlation is high (correlation coefficient R = 0.7443), and the stability prediction is performed with high accuracy. was gotten.

[Other embodiments]
As a method for evaluating the near-infrared spectrum, the original spectrum is evaluated as it is, or after performing preprocessing such as differentiation, normalization, and smoothing in order to reduce unwanted spectrum fluctuations such as noise. Analysis by the metrics method may be performed, and an optimal processing method may be set according to the sample. However, if the reproducibility of measurement is not good or it is easily affected by the external environment, differentiation processing or normalization processing Is preferred.

  FIG. 5A is a diagram showing an example of data obtained by first-order differentiation of the near-infrared spectrum data (original spectrum) shown in FIG. 4, and FIG. 5B is a loading spectrum (stable) in which each wave number in FIG. 5A indicates the contribution to the calibration model. The sex data calculation program 30 adds a weight by a partial least square method). In particular, the wave number region considered to contribute to the correlation of the calibration model is indicated by a broken line in FIG. 5B. Note that the above-mentioned Z value is the sum obtained by multiplying the absorbance X (FIG. 5A) of the near-infrared spectrum of each wave number by the numerical value of the loading spectrum and the coefficient W (FIG. 5B).

  In the above embodiment, the partial least square regression analysis method (PLS analysis method (Partial least squares)) is used as the chemometrics method, but the principal component analysis method (PCA method (Principal component analysis)), linear multiple regression An analysis method (MLR analysis method), a principal component regression analysis method (PCR method (Principal component regression)), a Fourier transform regression analysis method (FTR method (Fourier transform regression)), and the like can also be used.

  In the above embodiment, amrubicin hydrochloride lyophilized preparation, which is an injection, has been described as a specific example, but the pharmaceutical preparation that is subject to stability prediction is not limited thereto. Examples of the pharmaceutical preparation for which the stability is predicted according to the present invention include tablets, capsules, granules, pills, powders, solutions, syrups, troches, suspensions, emulsions, injections (lyophilized preparations) , Powder-filled preparations, aqueous injection preparations, etc.), suppositories, ointments or patches, and the like, and various commonly used pharmaceutical preparations. An injectable preparation is preferable, and a lyophilized preparation is particularly preferable.

  In the above embodiment, the target of actual measurement (measurement by the conventional analysis method) is the amount of the related substance. However, the specific index that is the target of actual measurement (measurement by the conventional analysis method) in the present invention is the time course of storage. It is sufficient that the value is a numerical value serving as an index of quality, that is, an index serving as a reference for stability evaluation. Specifically, for example, in addition to the amount of related substances, the content of active ingredients, the amount of water and the like can be mentioned. These specific indicators need to be appropriately selected for each target pharmaceutical preparation.

  In the above embodiment, HPLC or the like is used for the amount of the related substance as the method of measuring the specific index (measurement by the conventional analysis method) in step S12 of FIG. 3, but the method is not limited to this. In addition, use HPLC etc. for the active ingredient content, Karl Fischer method etc. for water content, paddle method, rotating basket method, free through cell method etc. for elution, and hydrogen electrode method etc. for pH. Can do.

  In the above embodiment, the storage condition is 40 ° C./75% RH (step S12 in FIG. 3) or the like. However, the storage condition is not limited to this, but 30 ° C./65% RH, 25 ° C./60% RH. Etc. Examples of the storage period include 1 week, 2 weeks, 1 month, 3 months, 6 months, 9 months, 12 months, 18 months, 24 months, 30 months, 36 months, and the like. Moreover, as these combinations, 40 degreeC / 75% RH-2 weeks, 40 degreeC / 75% RH-1 month, 25 degreeC / 60% RH-6 months, and 25 degreeC / 60% RH-12 months are mentioned. . Note that it is preferable that the specific storage conditions and the predetermined storage period are appropriately changed according to the properties of the specimen.

  In the above embodiment, the stability of the measured specimen after storage for the same period (the stability after storage under the “same conditions”) is predicted by applying the calibration model. Is known, stability after storage for a period different from the storage period actually measured when the calibration model is generated (stability after storage under “similar conditions”) may be predicted. For example, based on one calibration model or two calibration models with different storage periods, (1) multiply the prediction value obtained in Example 1 by the ratio of the storage period, and (2) the prediction value obtained in Example 1. By interpolating from the predicted values obtained in Example 2, the stability in an arbitrary storage period between the storage period of Example 1 (6 months) and the storage period of Example 2 (12 months) is predicted. (3) As shown in FIG. 19 and FIG. 20, the relationship between the storage period and the specific index, which is not simply proportional, is estimated from the relationship between the measured values of the specific index in different storage periods, and differs from the relational expression. It is also possible to predict the stability after a lapse of a period different from the storage period measured at the time of calibration model generation by a method such as prediction of the storage period.

  In the above embodiment, the stability of the pharmaceutical preparation is predicted by the processing computer 104. However, the stability prediction method can also be used in a pharmaceutical preparation production line in a factory or the like.

  FIG. 17 is a diagram showing a hardware configuration of the stability prediction apparatus 100 when the stability prediction method of the present invention is applied to a method for producing a pharmaceutical preparation. The stability prediction apparatus 100 is the same as that shown in FIG. 1, that is, a near-infrared spectrum measuring means 102 for measuring the near-infrared spectrum of an object, and a processing computer connected to the near-infrared spectrum measuring means 102. However, it is different in that it is connected to a production line control unit 106 for controlling the production line. Note that the near-infrared spectrum measuring means 102 shown in FIG. 17 is arranged so that the near-infrared spectrum of each vial containing a pharmaceutical preparation produced in the production line can be measured.

  Further, the stability data calculation program 30 stored in the hard disk 16 shown in FIG. 17 includes a calibration model (calibration curve) generation process based on the chemometrics method, a stability data calculation process based on the calibration model, In addition to processing for determining whether or not a predetermined stability criterion is satisfied, there is also processing for transmitting a command signal for excluding the sample to the production line control unit 106 when the predetermined stability criterion is not satisfied. It differs in the point to do.

  FIG. 18 shows a flowchart of processing performed by the stability prediction apparatus 100 shown in FIG. In this embodiment, it is assumed that the calibration model has already been generated and stored in the hard disk 16.

  As shown in FIG. 18, as in step S <b> 16 of FIG. 3, the CPU 10 of the processing computer 104 that has received the signal from the production line control unit 106 controls the near-infrared spectrum measuring means 102 for each specimen on the production line. As a result, the near-infrared spectrum before storage is measured and the near-infrared spectrum data is received and stored in the hard disk 16 (step S20).

  The CPU 10 of the processing computer 104 applies the near-infrared spectrum data obtained in step S16 to the calibration model generated in step S14, thereby predicting the specific index after storing the specimen under the same conditions as the calibration model. A value is calculated (step S22).

  Further, the CPU 10 of the processing computer 104 determines whether or not the predicted value of the specific index satisfies the stability standard of the pharmaceutical preparation stored in the hard disk 16, for example, the predicted value of the related substance amount is the stability standard of the pharmaceutical preparation. It is determined whether or not it is within the standard (step S24), and those that do not satisfy the standard (No in step S24), the production line control unit 106 (for example, remove the vial from the production line). A compressor (not shown) for operating the cylinder pushed out to the waste line is controlled (step S26).

  When the CPU 10 of the processing computer 104 determines that the stability criterion is satisfied (Yes in step S24), it determines whether or not the next sample exists in the production line (for example, sensed by the sensor 108 in FIG. 17). Further, if it exists, the process from step S20 is repeated for the next sample (Yes in step S28). On the other hand, if the next sample does not exist, the process ends (No in step S28).

It is a figure which shows the hardware constitutions of the stability prediction apparatus 100 of this invention. It is a figure which shows the example of the parameter memorize | stored in calibration parameter DB32. It is a flowchart of the process which the stability prediction apparatus 100 of this invention performs. It is a figure which shows the original spectrum of the near-infrared spectrum data measured. 5A is a diagram showing an example of data obtained by differentiating the near-infrared spectrum (original spectrum) of the object shown in FIG. 4, and FIG. 5B is a graph showing a loading spectrum corresponding to FIG. 5A. It is a table | surface which shows the difference of the correlation in case the specific wavenumber area | region (analysis area | region) to analyze differs. FIG. 7 is a table showing the correlation between the amount of related substances predicted from the near-infrared spectrum measured after storage and the correlation between the amounts of related substances predicted from the near-infrared spectrum measured before storage (immediately after production). It is. In Example 1, it is the figure which drew the analytical curve by evaluating the data of the predicted value and actual value which were used for calibration model generation. In Example 1, it is a table | surface which shows the result of the predicted value and measured value of the amount of related substances after a preservation | save. In Example 1, it is a graph which shows the correlation of the predicted value of the amount of related substances after a preservation | save, and an actual measurement value. In Example 2, it is the figure which drew the analytical curve by evaluating the data of the predicted value and actual value which were used for calibration model generation. In Example 2, it is a table | surface which shows the result of the predicted value of the related substance amount after a preservation | save, and the measured value. In Example 2, it is a graph which shows the correlation of the predicted value of the amount of related substances after a preservation | save, and an actual measurement value. In Example 3, it is the figure which drew the analytical curve by evaluating the data of the predicted value and actual value which were used for calibration model generation. In Example 3, it is a table | surface which shows the result of the predicted value of the related substance amount after a preservation | save, and a measured value. In Example 3, it is a graph which shows the correlation of the predicted value of the amount of related substances after a preservation | save, and an actual measurement value. It is a figure which shows the hardware constitutions of the stability prediction apparatus 100 at the time of applying the stability prediction method of this invention to the manufacturing method of a pharmaceutical formulation. It is a figure which shows the flowchart of the process which the stability prediction apparatus 100 shown in FIG. 17 performs. FIG. 19 is a table comparing the initial values of the amounts of related substances measured from three lots in which a certain pharmaceutical preparation was produced, and the amounts of related substances after the preparations of each lot were stored under certain conditions. It is a graph which shows the relationship between the storage period and the amount of related substances in the long-term storage test of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Stability prediction apparatus 102 ... Near-infrared spectrum measurement means 103 ... Sample vial 104 ... Processing computer 106 ... Production line control part

Claims (10)

A stability prediction method for predicting the stability of a pharmaceutical preparation,
Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time A calibration model generation step for generating a calibration model in advance by performing variable analysis,
A near-infrared spectrum measurement step for obtaining a near-infrared spectrum data by measuring a near-infrared spectrum before storage for a specimen of a pharmaceutical preparation that is a stability prediction target,
By applying the near-infrared spectrum data obtained in the near-infrared spectrum measurement step to the calibration model generated in the calibration model generation step, after storing the specimen under the same or similar conditions as the calibration model A specific index value calculating step of calculating a predicted value of the specific index in
Stability determination that determines whether or not a specimen satisfies stability after a predetermined period of time by comparing the predicted value of the specific index obtained in the specific index value calculation step and the stability evaluation criteria of the pharmaceutical preparation Process,
A stability prediction method characterized by comprising: A stability prediction method for predicting the stability of a pharmaceutical preparation,
Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time A calibration model generation step for generating a calibration model in advance by performing variable analysis,
A near-infrared spectrum measurement step for obtaining a near-infrared spectrum data by measuring a near-infrared spectrum before storage for a specimen of a pharmaceutical preparation that is a stability prediction target,
By applying the near-infrared spectrum data obtained in the near-infrared spectrum measurement step to the calibration model generated in the calibration model generation step, after storing the specimen under the same or similar conditions as the calibration model A specific index value calculating step of calculating a predicted value of the specific index in
A stability prediction method characterized by comprising: In the stability prediction method of Claim 1 or Claim 2,
The wave number region of the near-infrared spectrum applied to the calibration model in the specific index value calculation step is in the range of 7500 cm ^{−1 to} 4400 cm ⁻¹ .
A stability prediction method characterized by that. In the stability prediction method in any one of Claims 1-3,
The wave number region of the near infrared spectrum applied to the calibration model in the specific index value calculation step is determined by selecting a wave number region having a high correlation coefficient.
A stability prediction method characterized by that. In the stability prediction method in any one of Claims 1-4,
The pharmaceutical preparation is amrubicin hydrochloride lyophilized preparation,
The specific indicator is the amount of deglycosides that are related substances.
A stability prediction method characterized by that. A method for producing a pharmaceutical preparation using the stability prediction method of claim 1,
In the stability determination step, samples that are determined not to satisfy the predetermined stability criteria are excluded from the production line.
A method for producing a pharmaceutical preparation characterized by the above. Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time A stability prediction method for predicting the stability of a pharmaceutical preparation using a computer having a storage unit that stores a calibration model generated by performing a random analysis,
A near-infrared spectrum measurement step for obtaining a near-infrared spectrum data by measuring a near-infrared spectrum before storage for a specimen of a pharmaceutical preparation that is a stability prediction target,
By applying the near-infrared spectrum data obtained by the near-infrared spectrum measurement step to the calibration model stored in the storage unit, the sample after storing the specimen under the same or similar conditions as the calibration model A specific index value calculating step for calculating a predicted value of the specific index,
A stability prediction method characterized by comprising: A stability prediction method for predicting the stability of a pharmaceutical preparation,
Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time A calibration model generation step for generating a calibration model in advance by performing variable analysis,
A near-infrared spectrum measurement step for obtaining a near-infrared spectrum data by measuring a near-infrared spectrum before storage for a specimen of a pharmaceutical preparation that is a stability prediction target,
By applying the near-infrared spectrum data obtained in the near-infrared spectrum measurement step to the calibration model generated in the calibration model generation step, the identification after storing the specimen under similar conditions to the calibration model A specific index value calculating step for calculating a predicted value of the index,
A stability prediction method characterized by comprising: A stability data calculation program for predicting the stability of pharmaceutical preparations,
Using near-infrared spectrum data measured before storage for a sample for generating a calibration model and actual measurement data obtained by actually measuring a specific index of the sample after storing the sample under a specific condition for a predetermined period of time Calibration model generation processing for generating a calibration model in advance by performing variable analysis,
A near-infrared spectrum measurement process that obtains near-infrared spectrum data by measuring the near-infrared spectrum before storage for specimens of pharmaceutical preparations that are subject to stability prediction,
After applying the near-infrared spectrum data obtained by the near-infrared spectrum measurement process to the calibration model generated by the calibration model generation process, the specimen is stored under the same or similar conditions as the calibration model. Specific index value calculation processing for calculating a predicted value of the specific index in
A stability data calculation program for causing a computer to execute. It has a near infrared spectrum measuring means for measuring the near infrared spectrum of a pharmaceutical preparation, and is a stability prediction device for predicting the stability of a pharmaceutical preparation,
Measurement data obtained by measuring the near-infrared spectrum data measured by the near-infrared spectrum measuring means before storage for a calibration model generation sample, and measuring the specific index of the sample after storing the sample for a predetermined period under specific conditions Calibration model generation means for generating a calibration model in advance by performing multivariate analysis using value data,
The calibration generated by the calibration model generation means, the near-infrared spectrum data obtained by measuring the near-infrared spectrum before storage for the specimen of the pharmaceutical preparation to be stability-predicted by the near-infrared spectrum measurement means. A specific index value calculating means for calculating a predicted value of the specific index after storing the sample under the same or similar conditions as the calibration model by applying to the model;
A stability prediction apparatus comprising:

Images