Monitoring of continuous long run direct compression process

The accuracy of feed rates of API for both runs was monitored and recorded (Figure 27, Figure 28). As seen in Figure 27, the mass flow of API remained constant before overnight hold time in run 1, standard deviation (SD) being 0.214, and the feed rate achieved steady state conditions within a few minutes. The corresponding SD result was 1.017 after overnight hold time, before the concentration change. As seen in Figure 28, the mass flow of API in run 2 contained more variation (SD 12.269) before the overnight hold time, but achieved more stable condition between overnight hold time and beginning of concentration change (SD 1.260). The only difference in the processes between runs 1 and 2 was the locations of the API and MgSt feeders. In run 2, the vacuum conveyers introduced instability into the API feeding, as the API feeder was located at the second feeding and mixing stage, between the two vacuum conveyers. In run 1, this same instability, caused by the conveyors, affected the feed accuracies of the pre-mix, which was only missing of MgSt. Thus, this should not have any major effect on the API concentration in tablets as long as the disturbance from the conveyors does not cause any

76

de-mixing of the pre-mix. In both runs, there were times when the API feed rate was decreased or increased (Figure 27, Figure 28), in order to adjust the total feed rate. The API% is calculated from the individual mass flows of all materials and it can be seen that changes to the total feed rate did not affect the API% in the feed mixture. The main reason was the set-up, as both of the vacuum conveyers seemed to disturb the feeding accuracy of the API. The intentional change (doubling) of the API concentration (from 3.36 kg/h (=100 mg/tabl) to 6.72 kg/h (=200 mg/tabl)) occurred accurately in both runs (Figure 27, Figure 28). The transition to the higher concentration happened instantaneously and the feeders required no stabilization time to adjust to the new feed rate.

Figure 27. The adjusted theoretical API feed rate (kg/h) and the mass flow of API (calculated as API%) in the run 1.

77

Figure 28. The adjusted theoretical API feed rate (kg/h) and the mass flow of API (calculated as API%) in the run 2.

5.5.3 Compression force

The mean compression forces of runs 1 and 2 were monitored and analyzed. Before the overnight hold time, the mean compression forces were 21.1 kN (SD 1.047) in run 1 and 21.0 kN (SD 0.909) in run 2. After the overnight hold time, before the concentration change was initiated, the mean compression forces were 21.6 kN (SD 0.872) in run 1 and 21.5 kN (SD 0.574) in run 2. In conclusion, there was practically no influence of the overnight hold time on the compression force. As seen in Fig 29, the compression force did not show any time-related trends in either of the runs before or after the overnight hold time until the concentration change time point was reached. There were, however, deviations in the compression forces because the second vacuum conveyor induced an inappropriate suction effect on the mass in the tablet press hopper. This is an indication of the need to carefully design all interfaces between unit operations in order to prevent these kinds of disturbances. The variation in the compression force was smaller (as mentioned above) after the overnight hold time than that seen before the overnight hold time. Hence, it seems that the shut-down – overnight hold time – re start – operations exerted a stabilizing effect on the process. This could be due to better de-aeration of the powder during overnight hold time. During the run the airflow maybe cause a separation of the particles, and during hold time the de-aeration forms more stable conditions (Hertel et al., 2018).

78

Figure 29. The mean compression force (kN) of runs 1 and 2.

The time needed for the completion of the change in the API concentration was estimated to be too short i.e. it had not finished during the time process was being monitored. In run 2, the concentration change was achieved while the compression force remained constant after the completion of the concentration change. The difference between the runs, which determines how the concentration change will proceed, is dependent on whether the API feeder is in the first or second mixer, as can be clearly seen in Fig 29.

5.5.4 Tablet analyses 5.5.5 Tablet strength

The tablet strength (N) (n = 6), as determined every 5 min, is presented in Fig 30. The mean tablet strength before the concentration change in run 1 was 80.7 N (SD 0.279); in run 2, the corresponding value was 68.9 N (SD 0.217). These results indicate that the tablet strength was clearly lower in run 2 than in run 1. This should be due to the set-up, i.e. to the different locations of the API and MgSt feeders. In run 2, MgSt was introduced in the first feeding step which caused over-lubrication as the lubricant was mixed in both mixing stages and also in the pre-mix feeder. As the conveyors were witnessed to affect the feeders in second mixing step, both the variation (SD) and average deviation from set point of MgSt mass flow rate in run 1 and 2 were calculated. The variation (SD) before overnight hold time was 0.05 in run 1 and 0.01 in run 2. The average deviations from set point were −0.011 kg/h for run 1 and +0.001 kg/h for run 2. As a result, MgSt was slightly underfed in run 1 which may also contribute to the lower tablet strength in run 2.

79

Figure 30. Tablet strength (N) of the long runs, determined in every 5 min (n = 6).

This over-lubrication can also be seen in the dissolution profiles of the tablets obtained from the different runs (see Fig 33). Interestingly, the concentration change was not as clearly observed in the tablet strength in run 1 as in run 2. In run 2, the tablet strength was initially slightly increased, but then returned to the previous or an even lower level. This is probably due to the compression behavior of API, as compression pressure was not adjusted, and also to the over-lubrication effect. The amount of lubricant was constant in both formulations but when the concentration change was initiated, there was a transition period when the composition of the pre-mix was not satisfactory. This led to the lubricant amount being lower than desired, subsequently causing the tablet strength to increase at first to same level as in run 1. After the composition was gradually corrected, the strength was reduced below the strengths of the desired composition as there was more API which should have a poorer compressibility than the excipients. This same reduced tablet strength can also be seen at the end of run 1.

5.5.6 Average weight of tablets

The average tablet weight (mg) as determined every 5 min is presented in Figure 31. The target tablet weight was 300 mg ± 7.5%, and as seen in Figure 31, the average mass was within the target weight range during the whole process. Any variations were most likely due to inaccurate tablet press setting for the filling depth of the punches. Process

80

parameters for the tablet press were tested before the start of long run 1, and were not adjusted during the runs. This was a strategic decision since by maintaining constant conditions, the disturbances and deviations in the tablets could be better observed. In real-life scenarios, tablet press filling depth should have been adjusted. After the overnight hold time, the average tablet weight started to increase until the process was finished in both runs, due to the change in the concentration of the API (bulk density:

paracetamol 0.50–0.65 g/ml and MCC 0.28–0.33 g/ml, according to the manufacturer's specification). These results reveal that the deviations in the formulation (such as the concentration change) could be detected by analyzing the average tablet weight by working in open loop. Otherwise (not in this study) the tablet press controller will start manipulating the fill depth to compensate for the weight difference.

Figure 31. Average weight (mg) of the long runs, determined in every 5 min (n = 20).

Standard deviations are not shown for the sake of clarity.

5.5.7 Assay of tablets

The target value before concentration change was 95.0–105.0 mg/tabl and after concentration change 190.0–210.0 mg/tabl. The results of the paracetamol assay during the runs are presented in Fig 32. As seen in Fig 32a, before the overnight hold time, the values varied between 85.3 mg and 102.7 mg (mean 95.41 mg, SD 4.280) in run 1 and between 78.4 mg and 102.7 mg (mean 87.56 mg, SD 6.161) in run 2. In both runs, the acceptable level was not achieved in every sampling point (Fig 32), and furthermore the

81

level detected in run 2 was clearly lower than in run 1 before the concentration change (Table 10). The difference between runs could be due to the challenges for accurate API feeding caused by the process set-up (conveyors) in run 2 (Figure 28). One reason for generally low content was that the DC grade of API contains only 97%(w/w) API which was by mistake not taken into account in the initial plan, thus lowering the actual targets from 100 mg to 97 mg (before concentration change) and from 200 mg to 194 mg (after concentration change). When this is taken into account the content is much closer to target. The low content of run 1 after concentration change is at least partially explained by the slow transition to higher concentration. The API content seemed to be still rising when the process was stopped.

Figure 32. The assay results of paracetamol amounts from run 1 and 2. a) Assay every 15 min b) Assay every 5 min during the concentration change step.

A B

82

Table 10. The variation in the levels assayed at each process step in runs 1 and 2 (mean

Run1 95.41/4.28/4.49 97.23/2.91/3.00 181.81/7.07/3.89 Run2 87.56/6.16/7.04 88.58/3.91/4.41 186.96/6.69/3.58

The variation in values after the overnight hold time was lower than before in both runs, indicating that the hold time exerted a stabilizing effect on the process, maybe due to the de-aeration of materials in the process during overnight hold time. The amount of documented deviations (dev) occurred during the runs, calculated per hour were as follows; before the overnight hold time, 8.4 dev/h (run 1) and 7.0 dev/h (run 2); after the overnight hold time 7.6 dev/h (run 1) and 4.3 dev/h (run 2). More deviations per hour occurred in run 1, mainly due to tablet press switch off and turn on operations. Looking at the number of deviations, run 2 seemed better. Regardless of the amount of deviations, the results of assay still indicate that there was less variation when the API was mixed in two mixers (run 1), as compared to the situation with one mixer (run 2). Thus, it can be concluded, that not only the amount of individual deviations (or amount and type of deviation) affect the deviation in assay, as the process set-up had more effect on the results in this study. This is at least partially a result of vacuum conveyers, which disturbed the feeding accuracy of APl being fed into mixer 2. Furthermore, the values did not fulfill the assay requirements in run 2 after the overnight hold time. After the concentration change, the values remained near the minimum level (190.0 mg) and did not even reach that level in run 1.

5.5.8 The effect of location of MgSt feeders on dissolution of API

The dissolution profile was analyzed from both runs from the samples collected at time point of 205 min. As shown in Fig 33, the dissolution profile is faster in run 1 than in run 2. Clear differences are seen after 10 min, as the dissolved amount of API had reached 77.1% in run 1, but the corresponding result was only 47.7% in run 2. At the 15 minute time point, there was still a difference in the dissolved amount of API, i.e. 89.8% in run 1 and 64.4% in run 2. When combining these results with the results of tablet strength (Fig 30), it can be concluded that an over-lubrication phenomenon had been present when the lubricant was fed into the first feeding stage and went through two blending operations.

83

Figure 33. The effect of location of the MgSt feeders on dissolution profiles of API.

5.5.9 Residence time distribution calculations

Residence time distributions from API feeding to tablets in both runs were calculated to determine how long it takes the material from the feeders to be converted into tablets and thus to examine if any deviations in the feeding can be detected in the tablets. The results of the calculations are shown in Fig 34. The mean residence times for API in run 1 were 44.3 min compared to 30.6 min in run 2. As shown in Figure 34, the results were not unexpected, as API was fed at the beginning of the process (first feeders) in run 1 whereas in run 2, the API was fed closer to the end of the process (second feeding).

84

Figure 34. The mean residence time distributions for run 1 and run 2.

5.5.10 Process deviation documentation during the long runs

This set-up of continuous direct compression manufacturing process was monitored, and process parameters adjusted manually if disturbances/deviations were observed. The data was recorded using in-house software and all the disturbances and deviations were documented. These disturbances and deviations are listed in Table 11 and Table 12. This is the first time that this kind of list of process deviation document is published with respect to a continuous manufacturing process. As can be seen in Table 11 and Table 12, many intentional and unintentional deviations occurred. This was mainly due to the set-up and arrangement of the unit processes. In order to achieve a synchronized mass flow rate through all unit processes, the mass flow of API feeder and pre-mix feeder were adjusted and reduced to 5% (as the formulation was kept the same). This 5% adjustment was performed in both runs before and after the overnight hold time. The flow of all other feeders was increased to 10% only in run 2 before the overnight hold time (Table 11, Table 12). The mass stream was monitored by visually inspecting the fill level in tablet press hopper and corrected by adjusting the mass flow, since the tableting speed was kept constant. The mass flow was adjusted either by switching off the feeders (i.e. at time point of 2 h 47 min 5 s in run 1), by switching off the tablet press (several times) in the run 1 or by adjusting feed rates down by 5 or 10% in the run 2. All these corrective actions were applicable, however adjusting feed rates down was the best corrective action since this creates perhaps the smoothest effects on the mass flow and minimizes the interruptions 85

for tableting. Switching off the tablet press in run 1 had to be done since there were malfunctions with the second conveyor (not releasing product to the press) and the blend in the tablet press hopper went low fast which would have started to affect the filling of the dies. As can be seen from Table 11 and Table 12, the mass flow rates were more often decreased than increased. This is an indication that the amount of the blend in the tablet press hopper tended to be constantly increased during the runs. Thus, the mass flow rates were lowered in the feeding to the second mixer.

Table 11. The process deviation documentation during run 1.

Time Disturbance/deviation 0 Start-up the process I 5 s Normal delay in start-up I 13 min 43 s Tablet press start-up I

17 min 36 s Tablet press shut-down (mass level not acceptable) U 18 min 13 s Tablet press re-start U

1 h 4 min 30 s Feeder 5 re-fill U 1 h 7 min 25 s Feeder 5 re-fill stopped U

1 h 24 min 28 s Tablet press shut-down (software failed) U 1 h 25 min 36 s Tablet press re-start U

1 h 27 min 13 s Tablet press shut-down (mass level not acceptable) U 1 h 29 min 53 s Tablet press re-start U

1 h 34 min 5 s Feeders 4, 5 and vacuum 2. Shut-down U 1 h 36 min7 s Feeder 1 re-fill U

1 h 39 min 5 s Feeder 1 re-fill stopped U 1 h 41 min 10 s Feeders 4,5 and vacuum 2. Re-start U

1 h 48 min 23 s Tablet press shut-down (mass level not acceptable) U 1 h 50 min 23 s Tablet press re-start U

2 h 27 min 13 s Tablet press shut-down (mass level not acceptable) U 2 h 27 min 30 s Tablet press re-start U

2 h 32 min 21 s Tablet press shut-down (mass level not acceptable) U 2 h 32 min 43 s Tablet press re-start U

2 h 38 min 8 s Feeder 5 re-fill U 2 h 40 min Feeder 5 re-fill stopped U 2 h 45 min Feeder 2 re-fill U

2 h 47 min 5 s Feeder 4 and 5 shut-down (too much mass) U 2 h 49 min Feeder 2 re-fill stopped U

2 h 52 min Feeder 3 re-fill U 2 h 54 min 10 s Feeder 4 and 5 start-up U 2 h 55 min Feeder 3 re-fill stopped U

2 h 55 min 43 s Tablet press shut-down (mass level not acceptable) U 2 h 57 min 43 s Tablet press start-up U

3 h 0 min 28 s Tablet press shut-down (mass level not acceptable) U 3 h 3 min 28 s Tablet press re-start U

86

Time Disturbance/deviation

3 h 9 min 3 s Tablet press shut-down (mass level not acceptable) U 3 h 9 min 54 s Tablet press re-start U

3 h 50 min 33 s Tablet press shut-down (mass level not acceptable) U 3 h 51 min 58 s Tablet press start-up U

4 h 2 min Feeder 5 re-fill U 4 h 3 min Feeder 5 re-fill stopped U

4 h 15 min 43 s Tablet press shut-down (tablets were jammed) U 4 h 17 min 3 s Tablet press re-start U

4 h 24 min Feeder 4 and 5 re-start U 4 h 55 min Feeder 1 re-fill U 4 h 57 min 47 s Feeder 1 re-fill stopped U 5 h 11 min Feeder 5 re-fill U 5 h 13 min Feeder 5 re-fill stopped U 6 h 17 min Feeder 1 re-fill U 6 h 17 min 54 s Feeder 1 re-fill stopped U 6 h 27 min Feeder 5 re-fill U 6 h 30 min Feeder 5 re-fill stopped U 6 h 35 min Feeder 2 re-fil U 6 h 35 min 37 s Feeder 2 re-fill stopped U 6 h 39 min Feeder 3 re-fill U 6 h 40 min Feeder 3 re-fill stopped U 7 h 18 min 16 s Labview software failed 7 h 28 min 54 s Process shut-down I

Overnight hold time I 0 Start-up the process I 8 s Normal delay in start-up I 0 h 11 min 10 s Feeder 4, 5 and mixer 2. start-up U 0 h 11 min 48 s Tablet press start-up I 0 h 21 min 20 s Vacuum 2 shut-down U 0 h 27 min 20 s All feeders shut-down U 0 h 30 min 16 s All feeders re-start U 0 h 30 min 45 s Vacuum 2 re-start U

1 h 1 min 18 s Tablet press shut-down (mass level not acceptable) U 1 h 4 min 3 s Tablet press re-start U

1 h 19 min Feeder 5 re-fill U 1 h 20 min 55 s Feeder 5 re-fill stopped U 1 h 22 min 35 s Feeder 1 re-fill U 1 h 23 min 30 s Feeder 1re-fill stopped U

1 h 25 min 23 s Tablet press shut-down (software failed) U 1 h 26 min 23 s Tablet press re-start U

1 h 30 min Formulation 2 (feeders 1,2,3) I 1 h 47 min 50 s Feeder 2 re-fill U

1 h 48 min 38 s Feeder 2 re-fill stopped U

87

Time Disturbance/deviation

1 h 54 min Feed rates 5% decreased of all feeders U 2 h 45 min Feeders 1,2,3 shutdown I

2 h 51 min 31 s Vacuum 1 shut-down I 2 h 54 min 37 s Feeders 4 and 5 shut-down I 2 h 55 min 32s Mixer 2. shut-down I 3 h Shut-down the process I

3 h10 min 48 s The last tablet sampling was obtained I

(U = unintentional, I = intentional).

Table 12. The process deviation documentation during run 2.

Time Disturbance/deviation 0 Start-up the process I 7 s Norma delay in the process I 1 h 11 min Feeder 2,4 and mixer 2. start-up I 1 h 18 min 14 s Tablet press start-up I 1 h 31 min 35 s Feeder 5 re-fill U 1 h 34 min 35 s Feeder 5 re-fill stopped U 1 h 48 min Feeder 1 re-fill U 1 h 50 min Feeder 1 re-fill stopped U

1 h 58 min 49 s Table press shut-down (mass level not acceptable) U 2 h 1 min 4 s Tablet press start-up U

2 h 51 min 25 s Feeder 2,4 and mixer 2 shut-down U 2 h 55 min 25 s Feeder 2,4 and mixer 2 start-up U 3 h 13 min 20 s Feeder 5 re-fill U

3 h 15 min 30 s Feeder 5re-fill stopped U 3 h 18 min 14 s Tablet press clock failed, re-start U 3 h 24 min Feeder 1 re-fill U

3 h 25 min 46 s Feeder 1 re-fill stopped U

3 h 32 min 19 s Tablet press shut-down (mass level not acceptable) U 3 h 33 min 34 s Tablet press re-start U

3 h 41 min Feeder 2 and 4 feed rate 5% decreased U 3 h 45 min 11 s Tablet press shut-down (tablets jammed)U 3 h 46 min 14 s Tablet press re-start U

3 h 56 min Feeder 1,3,5 feed rate 10% increased U 4 h 15 min 45 s Feeder 2 re-fill U

4 h 16 min 37 s Feeder 2 re-fill stopped U 4 h 20 min 40 s Feeder 3 re-fill U 4 h 21 min 37 s Feeder 3 re-fill stopped U 4 h 22 min 45 s Feeder 5 re-fill U 4 h 24 min 15 s Feeder 5 re-fill stopped U

4 h 26 min Feeder 2 and 4 feed rate 5% decreased U

88

Time Disturbance/deviation 4 h 29 min 10 s Feeder 1 re-fill U 4 h 30 min 20 s Feeder 1 re-fill stopped U

4 h 58 min Adjusting the cleaning filter for vacuum1 U 5 h 40 min 35 Feeder 5 re-fill U

5 h 42 min 50 Feeder 5 re-fill stopped U 5 h 46 min 20 s Feeder 1 re-fill U 5 h 47 min 30 s Feeder 1 re-fill stopped U 5 h 57 min Feeder 2 and 4 shut-down U 6 h 6 min 13 s Feeder 2,4 and mixer 2 re-start U 6 h 6 min Tablet press shut-down U 6 h 12 min 44 s Tablet press re-start U

6 h 28 min Feeder 1,3,5 feed rate 10% decreased U 6 h 32 min 52 s Tablet press shut-down (mass finished) U 6 h 34 min Feeder 2,4 feed rate 10% increased U 6 h 33 min 49 s Tablet press re-start U

7 h 3 min 49 s Feeder 2 re-fill U 7 h 4 min 35 s Feeder 2 re-fill stopped U 7 h 8 min 30 s Feeder 3 re-fill U 7 h 9 min 28 s Feeder 3 re-fill stopped U 7 h 13 min 59 s Feeder 1 re-fill U 7 h 15 min 4 s Feeder 1 re-fill stopped U 7 h 16 min 17 s Feeder 5 re-fill U 7 h 18 min 12 s Feeder 5 re-fill stopped U

7 h 34 min 19 s Tablet press shut-down (mass finished) U 7 h 37 min 45 s Tablet press re-start U

8 h 26 min 38 s Labv software failed U Overnight holdtime I

0 Process start-up I

8 s Norma delay in the process I

8 s Norma delay in the process I

In document Impact of formulation and process design on tablet quality in continuous manufacturing (sivua 78-0)