Web.mit.eduReal Time and Noninvasive Monitoring of
Dry Powder Blend Homogeneity
Chee-Kong Lai, David Holt, James C. Leung, and Charles L. Cooney
Dept. of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139 Gokaraju K. Raju
Sloan School of Management, MIT, Cambridge, MA 02139 Peter Hansen
Union Biometrica, Inc., 19 Ward Street, Somerville, MA 02143 One of the most common unit operations in preparation of resentative pattern for the mixture. This approach can pro- pharmaceutical solid dosage forms is the physical blending of vide a basis for monitoring the convergence of the expected the active drug substance with one or more excipients. The aggregate spectra to establish blend homogeneity. Because end point of this process is material homogeneity as mea- reflectance NIR is usually a weak signal, except for water, sured by sampling and off-line analysis of the powder. Re- there is a limitation to the sensitivity of this method for high moval of samples is currently done with a sampling probe potency drug formulations where drug content may be below called a thief to withdraw samples from different locations of 1% wrw in the mixture.
a blender. A sample thief is a probe designed to extract and As an alternate optical method, we developed a laser-in- collect small volumes of powder from a chosen representative duced fluorescence ŽLIF. technique to monitor homogeneity cross-section of a blender. The resulting samples are then of solid powder mixtures during component blending. LIF has assayed using the same methods used to analyze the finished been successfully deployed in diverse applications such as the product. Content uniformity is established if the drug content automotive ŽBeer, 1995 and biomedical industries ŽAlfano, of the samples conform to predetermined criteria ŽBerman 1998 to monitor fluorescence of gases and liquids. Recently, and Blanchard, 1995; Muzzio et al., 1997 . This method is Unger and Muzzio employed the LIF technique for quantifi- influenced by the skill of the operator and often provides false cation of mixing liquids in impinging jets ŽUnger and Muzzio, representation of the sample due to desegregation and dis- 1999 . The operations of LIF involve irradiating samples at a ruption of the powder bed during sampling and transport.
suitable wavelength for excitation and evaluating the emis- Thus, both sampling and analytical errors are likely to incur sion at another wavelength. An examination of many drugs in in these procedures ŽBerman et al., 1996; Harwood, 1964; the marketplace suggests that a majority of them are likely to Harwood and Ripley, 1977; Schofield, 1976 . Furthermore, fluoresce when excited at the proper wavelength. Hence, sampling and off-line analysis causes long cycle time for op- there is a broad opportunity for employing LIF for monitor- erating and optimizing the blending process. Because blend- ing dry powder blend homogeneity of pharmaceutical prod- ing validation is mandatory, due to the FDA's 1996 proposal ucts. The analysis is rapid and usually on the order of mi- to amend the good manufacturing practice regulations ŽFDA, croseconds. If a continuous light source is used, then the limit 1996 , commercial-batch final blends need to be tested rou- to data acquisition is the limit of computing speed. This rapid, tinely for blend homogeneity. For this reason, there is an op- on-line method allows one to examine quickly, in real time, portunity for new technology to fulfill and perhaps replace the details of blending kinetics, and thus the effect of blend- the conventional thieving method with a more rapid and con- ing conditions, such as blender type, physical particle charac- sistent technique of measuring blend homogeneity.
teristics, and order of component addition.
An interest in noninvasive monitoring of powder blend is The primary goal of this research is to develop a method to seen in the work with near-infrared Ž NIR spectroscopy to implement both process and product verification in real time monitor blend homogeneity ŽHailey et al., 1996; Sekulic, 1996; for blending of the dry active pharmaceutical ingredient Ž Wargo and Drennen, 1996 . NIR relies on the use of a com- with excipients. It is expected that success will facilitate pro- plex reflectance spectra specific to the substance analyzed and cess development and validation, open the possibility to re- register extensive data analysis to reduce the spectra to a rep- duce cycle time by eliminating the need for routine off-lineanalysis of blend homogeneity, and further assure product Correspondence concerning this article should be addressed to C.-K. Lai.
quality by measuring component homogeneity in a way that Vol. 47, No. 11
DataMax software. When excited at 488 nm, triamtereneproduced two distinct emission peaks at 526 nm and 561 nm.
Direct compaction anhydrous lactose ŽSheffield Products,New York of a mean particle size of 100 m and bulk den- sity of 0.6 grcc is used as the primary bulk filler for all subse-quent blending studies.
LIF sensor operation parameters
For instrumentation optimization, it was essential to ex- plore parameters to maximize the measured signal over thebackground. We would explore the effect of sample dis-tances, sample angles, laser power, detector sensitivities, andprocess variable such as bulk density during powder blend-ing. Signal responses to sample distances from the sensor weredetermined with triamterene in a 1-cm quartz cuvette at 60W laser power and detector sensitivity at 300 V. The rangeof the measured distance was "15 mm from the focal dis-tance of the lens. Evaluation of sample angle variation effectswas conducted at "10⬚ intervals from a reference point 90⬚to the laser beam. Laser-power output was reduced from 60 Figure 1. LIF sensor.
W to 16 W using a neutral filter to determine the effectson the API signal over the background with respect to detec-tor sensitivity using a 50-V stepwise increase from 350 V up avoids the possibility of sampling error. There is also the ben- to 700 V. Laser-power output was measured at the start of efit of permitting better equipment utilization by shortening each set of experiments and may vary due to changes in cycle time and reducing labor costs associated with this unit alignment to the optic fiber.
operation. Although the work here described processing ofpharmaceutical powders, this technology is applicable for alltypes of powder mixing processes in other industries.
Powder bulk density variations were achieved by gradual compaction of a fixed sample of 10% wrw triamterene pow-der mixture. LIF signal response was determined correspond- ing to each change in density. Blend homogeneity experi- The sensor ŽFigure .
1 uses an argon laser as the excitation ments were conducted in a 20-mL glass microblender with light source ŽOmnichrome Air Cooled Argon Laser Model 60% filled capacity Ž ;8 g of a formulation containing 10% 532 . The laser beam is directed into a fiber-optic cable con- triamterene in anhydrous lactose. This is a typical case for nected to the Photosensor Module containing the lens and mixing of a low dose and fine cohesive powder with a fast- the detector ŽHamamatsu HC120 Photomultiplier Tube. In- flowing bulk filler. All blends were run for a duration of 20 side the photosensor module, a dichroic mirror ŽOmega Opti- min at 20 rpm and raw data were acquired as distinct data 505DRLPO2 reflects the laser beam at 90 degrees to the points and presented without further manipulation. The laser sample. The fluorescent signal that is emitted from the sam- beam was directed either one-third from the top or bottom of ple is collimated by the lens and passes through the dichroic the powder level to determine consistent signal stability. An mirror and an emission filter ŽOmega Optical 530DF30. to infrared switch was used to control and synchronize data ac- the detector. Signals are converted to voltage and recorded quisition only when powder was present.
by a multimeter ŽHewlett-Packard 34401 A , which is inter- faced to a computer via the RS-232 and hyperterminal. Re- Powder mixing kinetics
flected light Žat 488 nm is prevented from reaching the de- tector by the dichroic mirror and the emission filter.
The same triamterene-lactose formulation as just cited was used in two separate experiments to demonstrate the sensitiv-ity of the LIF sensor toward the kinetics of mixing when the API was either placed at the top or bottom locations of the Triamterene Ž2,4,7-triamino-6-phenylpterine. is the active blender. Mixing proceeded under conditions identical to these pharmaceutical ingredient Ž API used in all experiments. It is an oral diuretic and antihypertensive agent typically formu-lated as Dyazide capsules and is a gift from SmithklineBeecham. This API is a cohesive powder of particle size 10᎐30 Monitoring of the mixing process of formulations of
m and bulk density of 0.3 grcc. The solid-state fluorescence spectra of triamterene is obtained at its tap density using Several batches containing 10%, 5%, 1%, 0.5%, 0.1%, and front-face optics in a FluoroMax-2 ŽISA Jobin 0% triamterene ᎐lactose formulations were prepared as ear- equipped with modified Czerny-Turner spectrometers and lier. Their blend kinetics profiles and homogeneity endpoints Vol. 47, No. 11
Figure 2. Effects of powder bulk density on LIF signals.
Figure 3. Synchronized LIF data acquisition for blend-
ing a 10% wrw triamterenerlactose mixtures.
were monitored in the usual manner. Triamterene was addedat the bottom Žnear the laser beam. in all cases, so that wecould track the reduction in signal even at the low drug pow- initial signal was low since the sensor was responding only to der concentration.
lactose during the first few rotations. In both cases, the LIFsignals reached an equilibrium state of blend homogeneity atthe same value with signal deviations of less than 5% Žsee LIF operation parameters
The LIF signal response of the sensor was maximum at the Monitoring of the mixing process of different powder
focal length of the lens with a first-order exponential decay in signal " from the focal distance. Since fluorescence emissionis radial in nature, a "30⬚ tolerance to the incident angle of Blend homogeneity was reached after 70 rotations for all the excitation beam to the sample was observed without loss the triamterene powder concentrations investigated, with of signal. The fluorescent signal from triamterene saturated each endpoint at signal values proportional to their API con- at a detector sensitivity of 400 V at 60-W laser output power.
tents. An example of the sensitivity of the LIF technology is Signal saturation was extended by reduction of the output illustrated in Figure 5 for blending profiles of API contents power to 16 W with the aid of a neutral-density filter. The of 0.5%, 0.1% and 0.05%. The results are summarized in signal to background fluorescence response was more favor- Table 1, where the relative standard deviation for all mixing able at lower laser power and lower detector sensitivity.
experiments was less than 5%.
A correlation curve generated from these data was fitted by a polynomial equation with R of 0.9984 Žsee Figure .
The nonlinearity of the correlation is believed to result from Changes in bulk density of the powders in the blender cor- saturation of the PMT response at the higher concentration responded to a proportional change in LIF signals. For an level. Linearity was observed at the lower concentration evenly mixed sample, there will be more API for each unit ranges from 0 to 1% triamterene with an equation of y s surface volume excited by the laser beam with increasingpacking density Žsee Figure .
2 . This suggested that bulk-den- sity variation during the blending process may elevate back-ground noise. Hence, it is favorable to consider monitoringblend homogeneity within the blender at a location wherebulk-density changes were minimal. Synchronized data acqui-sition was made at the selected location of the blender whenpowder was present and where the bulk density remainedrelatively constant. Data collected at a position one-third thedistance from the bottom of the powder bulk ŽFigure .
sulted in a clean first-order mixing kinetics profile.
Powder blend kinetics
Distinct differences in the early mixing phases were demonstrated when the API was loaded at different locationsof the blender. When the API was loaded at the bottom, the Figure 4. Sensitivity of LIF blend kinetic profile to the
initial signal was high due to close proximity of the LIF sen- location of adding active pharmaceutical in-
sor near the bottom. When loaded on top of the blender, the gredient in the blender.
Vol. 47, No. 11
during the mixing. These issues are handled effectively withthe introduction of an infrared on-off switch to synchronizedata acquisition only at positions when powder is present andby selecting a location for data acquisition close to the bot-tom of the vessel where the powder bulk density is relativelyconstant.
Each API concentration provides a unique mixing profile with endpoints of less than 5% deviations. A correlation ofthe LIF signal at homogeneity to the API concentration wasdetermined to a triamterene powder concentration of 0.1%.
This low level of detection demonstrated potential usefulnessto monitor the blending process in the manufacture of high-potency drug formulations.
In all of the cited experiments, data were obtained only from one location during each rotation. We made the as-sumption that monitoring one location of a mobile powder Figure 5. Monitoring of low-dose blending profiles.
mixture throughout the process would closely represent thebulk composition due to the constant turnover of powder atthat location. This was clearly demonstrated in Figure 4 by Table 1. LIF Mixing Endpoints
the experiment on blend kinetics where in one experiment, the active ingredient was placed on the top in the blender. In the second experiment, the same formulation was used where LIF Signal, Volts the active ingredient was placed at the bottom of the blender.
Both experiments were mixed similarly with LIF monitoring at a fixed location of the blender. Although the mixing kinet-ics was different during the mixing stage, both experimentsreached the same homogeneity steady state, indicating goodturnover of powders within the blender.
2.212 x q1.0493 and an R2 of 0.9947. Endpoint blend homo- This LIF technology can be applied to all types of mixing geneity was determined at as low a triamterene powder con- procedures, provided that the critical component of that mix- centration as 0.1% Ž1:1,000.
ture contained a measurable fluorescent signal at the concen-tration of interest. It can be used to study and monitor pow- der mixtures irrespective of the physical characteristic of eachparticle. Blend homogeneity is established when the LIF sig- We have optimized the design and operating parameters of nal at steady state is the same from one turnover of powder the LIF sensor. The laser power source, the detector sensitiv- to the next for each mixing rotation. The monitoring process ities, and the excitation wavelength can independently con- is considered as a snapshot of the powder content at each trol signal intensity. The laser output power needs to operate rotation. The signal derived from each snapshot is a quantita- at levels below the saturation limit that is necessary for quan- tive representation of the number of fluorescent particles dis- titative correlation of the fluorescence signal to the amount tributed within that area of analysis. Changes in that number of compound present.
of fluorescent particles within that area of analysis will result The primary process variables that impacted the signal in changes in fluorescence signals that indicate a nonhomo- quality are void volumes and differences in bulk densities geneous state. Any presence of a dead spot would result inan overall change in the API concentration. This change inconcentration would result in a change in signal when steadyis established.
The application of LIF to noninvasively monitor blend ho- mogeneity during dry pharmaceutical powder mixing allowsone to acquire real-time data on kinetics and the endpoint ofmixing. This approach eliminates errors introduced by the useof thief sampling and off-line assay techniques.
The authors acknowledge the financial support obtained from the Consortium for the Advancement of Manufacturing in Pharmaceuti-cals Ž CAMP . We also thank Union Biometrica, Inc. for helpful sug- gestions in the assembly of the LIF equipment.
Figure 6. Correlation of LIF signals to triamterene con-
Beer, et al., ‘‘Apparatus for the Detection and Control of Aromatic centration at mixing endpoint.
Compounds in Combustion Effluent, USP5425916 Ž Vol. 47, No. 11
Berman, J., and J. A. Blanchard, ‘‘Blend Uniformity and Unit Dose Schantz, S., et al., ‘‘In Vivo Native Cellular Fluorescence and Histo- Sampling,'' Drug De®. Ind. Pharm., 21, 1257 Ž
logical Characteristics of Head and Neck Cancer,'' Clin. Cancer Res., Berman, J., A. Schoeneman, et al., ‘‘Unit Dose SamplingᎏA Tale of 4, 1177 Ž
Two Thieves, Drug De®. Ind. Pharm., 22, 1121 Ž
Schofield, C., ‘‘The Definition and Assessment of Mixture Quality in FDA, ‘‘Current Good Manufacturing Practice: Amendment of Cer- Mixtures of Particulate Solids, Powder Technol., 5, 169 Ž
tain Requirements for Finished Pharmaceuticals; Proposed Rule,'' Sekulic, S., et al., ‘‘On-Line Monitoring of Powder Blend Homogene- 61 FR 20103 Ž ity by Near-Infrared Spectroscopy,'' Anal. Chem., 68, 509 Ž
Hailey, P., P. Doherty, et al., ‘‘Automated System for the On-Line Unger, D. R., and F. J. Muzzio, ‘‘Laser-Induced Fluorescence Tech- Monitoring of Powder Blending Processes Using Near-Infrared nique for the Quantification of Mixing in Impinging Jets,'' AIChE Spectroscopy. 1. System Development and Control,'' J. Pharm. J., 45, 2477 Ž
Biomed. Anal., 14, 551 Ž
Wargo, D., and J. Drennen, ‘‘Near-Infrared Spectroscopic Charac- Harwood, C. F., and K. A. Walanski, ‘‘Monitoring the Mixing of terization of Pharmaceutical Powder Blends, J. Pharm. Biomed. Powders,'' Org. Coat. Plast. Chem., 33, T305 Ž
Anal., 14, 1415 Ž
Harwood, C. F., and T. Ripley, ‘‘Errors Associated with the Thief Probe for Bulk Powder Sampling,'' Powder Bulk Solids Technol., 1,
Muzzio, F. J., P. Robinson, C. Wightman, and D. Brone, ‘‘Sampling Practices in Powder Blending,'' Int. J. of Pharm., 155, 153 Ž
Manuscript recei®ed No®. 22, 2000, and re®ision recei®ed May 11, 2001. Vol. 47, No. 11
Impaired muscle metaboreflex-induced increases inventricular function in heart failure Donal S. O'Leary, Javier A. Sala-Mercado, Robert A. Augustyniak, Robert L.Hammond, Noreen F. Rossi and Eric J. AnsorgeAm J Physiol Heart Circ Physiol 287:H2612-H2618, 2004. First published 15 July 2004;doi:10.1152/ajpheart.00604.2004
Asthma is the most common chronic disease in childhood in the united states representing about 5% to 10% of ed visits annually
The Acute Management of Asthma in Children – Revision Draft Updated May 2014 The Members of the Maryland Chapter of the American Academy of Pediatrics MDAAP Committee on Pediatric Emergency Medicine COPEM, have developed a clinical pathway for the management of children presenting to the emergency department with asthma exacerbations. This clinical pathway can serve as a clinical practice guideline and was developed using the latest in evidence-