Addressing Segregation of a Low-Dosage Direct Blend

Directly compressing a dry blend of pharmaceutical ingredients into tablets has the benefits of requiring minimal powder handling and reducing processing costs. Maintaining content uniformity of the dry blend, especially for low-dose therapeutics, throughout the process poses significant challenges. But personnel can modify the manufacturing equipment and process to help ensure content uniformity of low-dose tablets.

The authors produced a low-dosage tablet (i.e., one containing < 2% active ingredient) using a direct-compression process. The process included screening the drug along with the excipients, blending the ingredients in a V-blender, transferring the blend from the V-blender to an intermediate bin, and compressing the tablets using a 51-station D-Hata tablet press (Elizabeth-Hata International, North Huntingdon, PA).

Based on the process, an initial demonstration batch was manufactured at 787.5-kg scale. The proprietary composition comprised a low-dose drug, a carbonate buffer system, and coprocessed sugars as filler. Additional common excipients were included to aid in the tableting process. Given the low dose of the active ingredient, samples were collected and analyzed for blend uniformity (BU) and content uniformity (CU) per FDA's draft guidance for stratified in-process dosage-unit sampling (1, 2). The uniformity data were collected during various transfer steps and during compression. For the first demonstration batch, the BU data met the specification, but the CU results did not meet the specification: a location average exceeded 110% of the label-claim limit at the end of compression

To understand the lack of CU during the tableting process, the authors conducted several tests at the bench scale to elucidate the segregation mechanism and flow properties of the formulation blend. The authors hypothesized that fluidization during the transfer of the blend from the blender to the intermediate bin, and subsequently from the bin to the tablet press, could result in segregation of the active in the formulation.

Based upon the test results described in this article, corrective process and equipment modifications were implemented for a second demonstration batch. These modifications successfully reduced the segregation during the blender-to-press transfer steps, so that the CU data passed FDA's draft guidance for the second demonstration batch (1). These modifications were incorporated into the commercial process and successfully validated. The details of the root-cause analysis and process and equipment modifications are discussed in the following sections.

Root-cause assessment and confirmation

A root-cause analysis of the CU variation for the first demonstration batch was conducted using established troubleshooting methods (3). The BU data collected from the blender and bin had minimal variation and was within specification as described in FDA's guidance. The BU samples were obtained using a sampling thief from 10 locations within the V-blender and 12 locations within the bin. These samples weighed 1–2 times as much as the tablet. On the other hand, the CU data had higher variation (RSD = 3.2%, n = 60 samples) due to a distinct upward trend at the end of compression (average = 112% label claim, n = 3 samples).

During an analysis of variance of the CU data, the authors observed that more than 90% of the variation occurred between locations, as opposed to a variation of individuals at a single location. Samples collected from the intermediate bin also exhibited higher variation (RSD = 2%, n = 12 samples) than the blender samples. Based on these results, the authors hypothesized that segregation during the postblending steps (i.e., the blender-to-bin or bin-to-press transfer steps), in combination with the flow pattern from the intermediate bin, resulted in the upward CU trending at the end of compression. In particular, fluidization with or without dusting segregation during the transfer steps can result in a concentration of active-rich fines at the periphery or top of the bin. When the segregated blend discharged in funnel flow (i.e., a first-in-last-out flow pattern), the drug-rich blend would be present in the tablets obtained toward the end of compression. The authors discussed changes in material flow patterns (i.e., mass flow versus funnel flow) and the effect they can have on CU trending in a previous article (4).

To confirm the root-cause hypothesis, bench-scale segregation tests were conducted to confirm the dominant segregation mechanism for the blend. A test was conducted on the blend to assess its potential to experience fluidization segregation. The test method, described in detail in the literature, is conducted by fluidizing a column of material by injecting air at the bottom of the test column, and allowing it to deaerate in a controlled manner (5). When the test is concluded, the column is split into 16 equal sections, and selected sections are assayed for drug content. If the blend were prone to fluidization segregation, samples from the bottom (i.e., Layer 1) would have a coarser particle size and contain a lower concentration of the drug than the samples from the top (i.e., Layer 14). The concentration of drug from the various layers from the fluidization segregation test data are graphically depicted in Figure 1. A trend of increasing drug concentration toward the top layers of the blend confirmed that the blend had a high tendency to segregate, predominantly because of fluidization.

Figure 1: Fluidization-segregation test results showing drug content across individual layers. (ALL FIGURES ARE COURTESY OF THE AUTHORS)

Although fluidization was the primary factor, the authors performed a sifting segregation test by filling a narrow test cell to one side, thus allowing a pile of the dry blend to form within the cell. If a material segregates by sifting, the coarse particles flow to the far side, away from the fill point, while fine particles accumulate under the fill point. After the cell was filled, the samples were extracted from the bottom of the tester in layers, with a given layer subdivided into five slots. The first layer extracted was designated Layer 1, and as many as 14 layers were collected, depending on the surcharge angle of the material. Slot locations were designated A through E, with A being away from the fill point, and E under the fill point. If the blend is prone to sifting segregation, samples collected below the fill point (i.e., slot E) are drug-rich and finer than those collected away from the fill point (i.e., slot A). However, an increase in drug-rich fines away from the fill point (i.e., at slot A) is consistent with a fluidization or dusting segregation mechanism. The sifting segregation test data, as shown in Figure 2, demonstrated that the blend had a low tendency to segregate by sifting.

Figure 2: Sifting segregation test results showing drug content across slot locations. (ALL FIGURES ARE COURTESY OF THE AUTHORS)

The segregation test results confirmed that both the blender-to-bin and bin-to-press transfer steps could result in fluidization or dusting segregation, thereby leading to CU problems. Subsequently, flow-properties tests, including cohesive-strength and wall-friction tests, were conducted at the bench scale to confirm the flow pattern of the blend during the blender-to-press discharge (6). Based on the wall-friction test results, the authors concluded that the rectangular-to-round hopper angles for the bin used for the first demonstration batch were not smooth or steep enough to provide mass-flow discharge, and that the bin discharged the blend in funnel flow, thereby resulting in higher variation of the drug content in the tablets. Based on the flow-properties test results, a specially designed bin (i.e., a cone-in-cone) was used for the second demonstration batch. This bin design provided mass-flow discharge, thus minimizing the effects of dusting segregation during transfer.

Process and equipment modifications

Modifications to the blender-to-bin transfer step. The blender-to-bin transfer step for the first demonstration batch consisted of an uncontrolled free fall from the blender into the bin, likely resulting in fluidization and dusting segregation. To reduce the likelihood of segregation during the blender-to-bin transfer step for the second demonstration batch, a flexible sleeve, or sock, was designed to control and reduce the transfer rate from the blender and minimize the free fall of the material during transfer. The customized sock had a conical section at the top that converged from the larger V-blender outlet (8-in. diameter) to a smaller diameter (i.e., 4 in.) to provide greater control of the material and reduce the free fall of material into the bin. In addition, the blender valve was throttled during discharge (as high as 10% open) so that the material could deaerate within the sock before the sock was lifted from the top surface of the material to transfer it into the bin. Based on the stratified blend samples collected within the bin after filling (RSD = 1.0%, n = 10 samples), these process modifications were successful in minimizing segregation during this transfer step.

In addition to modifying the process equipment used for the blender-to-bin transfer step, the cone-in-cone bin was expected to provide mass flow and was used for the second demonstration batch. Visual observations of the material discharge from the bin during the second demonstration batch confirmed mass-flow discharge, as predicted by the bench-scale flow-properties tests conducted beforehand.

Modifications during bin-to-press transfer. Segregation also could occur during the bin-to-press transfer step and contribute to the CU trending observed in the first demonstration batch. The transfer chute used for the first demonstration batch consisted of large-diameter (i.e., 8-in.) tubing without any valves to reduce the free fall of material or venting to reduce the air counterflow up through the powder as the chute is filled. Since air counterflow during free fall as the chute is filled can result in fluidization and dusting segregation, thus carrying drug-rich fines back up into the bin above, a new transfer chute design was used for the second demonstration batch. The new transfer chute consisted of the following parts:

• A mass-flow conical reducer at the top of the chute and small-diameter (i.e., 4-in.) tubing to reduce the displaced air and counterflow during filling

• Two butterfly valves to reduce the free fall height during filling

• A passive filter vent to allow displaced air during filling to exit the chute rather than conveying back up through the blend and causing segregation.

The modifications to the bin and transfer chute design from the first demonstration batch to the second demonstration batch are shown in Figure 3.

Figure 3: Tote-bin and transfer-chute modifications from (a) demonstration batch #1 in a bin-to-press feed system and (b) demonstration batch #2 in a modified bin-to-press feed system after process and equipment modifications. (ALL FIGURES ARE COURTESY OF THE AUTHORS)

Results and conclusion

The process and equipment modifications implemented for the second demonstration batch were successful in reducing the CU variation and alleviating the upward CU trend that was observed at the end of compression in the first demonstration batch, as shown in Figure 4. The CU data for the second demonstration batch showed minimal variation (RSD = 0.9%, n = 60 samples) and met the FDA draft guidance specifications. The successful validation of the commercial process further demonstrated the robustness of these modifications.

Figure 4: Stratified tablet content-uniformity data for (a) demonstration batch #1 with segregation and (b) demonstration batch #2 after process and equipment modifications. (ALL FIGURES ARE COURTESY OF THE AUTHORS)

Acknowledgment

The authors thank their colleagues Prasad Challapalli, senior scientist at Patheon; Jacques Mowrer, associate director of analytical chemistry, and Alicia Ng, analytical chemist, at Transcept; and Mary Thomas, former manufacturing-process engineer at Patheon for their collaboration on this project.

Nipun Davar is a vice-president of pharmaceutical sciences at Transcept Pharmaceuticals. Thomas Baxter is a senior consultant at Jenike & Johanson. Pauly Kavalakatt is a senior scientist of formulation development, and Sangita Ghosh* is an associate director of product development, both at Transcept Pharmaceuticals, 1003 W. Cutting Blvd., Pt. Richmond, CA, 94804, tel. 510.215.3500, sghosh@transcept.com. Herbert Schock is a technical manager of commercial operations at Patheon Pharmaceuticals

*To whom all correspondence should be addressed.

Submitted: Aug. 19, 2010. Accepted: Nov. 8, 2010.

References

1. FDA, Powder Blends and Finished Dosage Units—Stratified In-Process Dosage Unit Sampling and Assessment (Rockville, MD, Oct. 2003).

2. FDA, Powder Blends and Finished Dosage Units—Stratified In-Process Dosage Unit Sampling and Assessment, Revised Attachments (Rockville, MD, Nov. 2003).

3. J.K. Prescott and T.P. Garcia, Pharm. Technol. 25 (3), 68–79 (2001).

4. J.K. Prescott, and R.J. Hossfeld, Pharm. Technol. 18 (6), 99–114 (1994).

5. J.K. Prescott et al., Pharm. Technol. 30 (12), 54–64 (2006).

6. J.K. Prescott and R.A. Barnum, Pharm. Technol. 24 (10), 60–84 (2000).