CHAPTER FOUR: GENERAL DISCUSSION
4.5 Relationships between different pellet properties
The sections 4.1 - 4.4 summarize the findings o f various studies undertaken and build up a clearer picture o f the characteristics o f the pellet batches. However, are these characteristics inter-linked and, in particular, what effect is the shape o f the pellets exerting upon their properties? In order to address these questions, the relationship between several o f the results was investigated.
Spearman’s ranking test was applied as described in section 2.6, to compare the rank orders o f two measured properties at a time.
For n = 8, since all the batches were included, the tabulated rs value is 0.738 at the 5% confidence level i.e. a calculated rs value exceeding this tabulated value indicates a positive correlation between the ranking o f the two sets o f results. Similarly, a negative correlation would be indicated by numbers more negative than -0.738. If the rank orders were exactly the same, the rank correlation coefficient would be 1.000.
Ranked properties Batches
Spearman’s correlation
coefficient
porosity + surface area all uncoated 0.000
porosity + mean dissolution time (MDT) all uncoated -0.881
porosity + surface roughness (Ra) all uncoated -0.238
surface area + surface roughness (Ra) all uncoated 0.619
surface area + mean dissolution time (MDT) all uncoated 0.167 coat thickness + mean dissolution time (MDT) all coat 5 0.417 3-D shape (Ccs) + mean dissolution time (MDT) all uncoated 0.667 3-D shape (Ccs) + mean dissolution time (MDT) all coat 5 -0.060 2-D shape (e^) + mean dissolution time (MDT) all uncoated 0.714 2-D shape (qr) + mean dissolution time (MDT) all coat 5 -0.012 3-D shape (ecs) + surface roughness (Ra) all uncoated 0.524 2-D shape (eR) + surface roughness (Ra) all uncoated 0.357
3-D shape (Ccs) + surface area all uncoated 0.048
3-D shape (ecs) + % underfilled capsules all uncoated 0.595 3-D shape (ecs) + % underfilled capsules all coat 5 0.714 2-D shape (eR) + % underfilled capsules all uncoated 0.881 2-D shape (ea) + % underfilled capsules all coat 5 0.714 surface roughness (Ra) + % underfilled capsules all uncoated 0.310
3-D shape (ecs) + bulk density all uncoated 0.024
3-D shape (Ccs) + bulk density all coat 5 0.452
surface roughness + bulk density all uncoated -0.548
3-D shape (ecs) + Heywood’s volume coefficient, k all uncoated 0.083 3-D shape (ecs) + Heywood’s volume coefficient, k all coat 5 0.476 2-D shape (eR) + Heywood’s volume coefficient, k all uncoated -0.250 2-D shape (eR) + Heywood’s volume coefficient, k all coat 5 0.286 3-D shape (ecs) + Heywood’s surface coefficient, f all uncoated 1.000 3-D shape (ecs) + Heywood’s surface coefficient, f all coat 5 0.976 2-D shape (eR) + Heywood’s surface coefficient, f all uncoated 0.833 2-D shape (eR) + Heywood’s surface coefficient, f all coat 5 0.786
3-D shape (Ccs) + aspect ratio all uncoated -0.905
3-D shape (ecs) + aspect ratio all coat 5 -0.702
2-D shape (eR) + aspect ratio all uncoated -0.976
2-D shape (eR) + aspect ratio all coat 5 -0.845
measured properties.
Table 4.1 shows the results for the Spearman’s rank coefficients comparing the rank orders o f certain pairs o f measured properties for uncoated or fully coated pellets. Looking at the result for two properties such as porosity and surface area, correlation coefficient = 0.000. This indicates no correlation at all which is not surprising since the porosity may be regarded as an internal feature o f the pellets and in this case the surface area was determined by the technique o f air permeametry which measures external surface area only. With other techniques o f measuring surface area such as gas adsorption, a correlation with porosity would be expected since the total surface area is measured. However, porosity and mean dissolution time for uncoated pellets
did show a negative correlation, i.e. the ranking order was inverted. Thus the highest porosity pellets would have the lowest mean dissolution time i.e. the fastest drug release, and this is not surprising since increased pores would lead to increased dissolution. The coated pellets however, were assumed to be non-porous and their release was controlled entirely by the coating layer. The relationship between mean dissolution time and other properties was also investigated but no definite correlations were found. Surface area o f the uncoated pellets did not appear to be related to the dissolution rate, whilst the same for coated pellets was not considered since there were no significant differences in surface area found between batches. The relationship between surface area and surface roughness (Ra) o f uncoated pellets was also investigated and in this case there was almost a positive correlation seen between the two properties.
For uncoated pellets at the 5 % confidence level, the shape factor Cc3 almost showed a positive correlation with mean dissolution time, whereas the coated pellets showed almost none. This means that for the uncoated pellets, a more spherical shape produced a higher mean dissolution time. This is partly due to the disintegration o f the elongated pellets DU and LD leading to shortened dissolution times, and not due to surface area effects.
There were no other correlations found between properties such as bulk density, shape and surface roughness. Porosity and surface roughness o f pellets was seen to be unrelated which is in contrast to work done by other authors [Ozkan and Briscoe 1996; Rowe 1978a, 1979], who claimed that there is a relationship between surface roughness and porosity o f tablets or compacts.
Capsule filling performance was also ranked against other properties to look for correlations. The percentage o f underfilled capsules was nearly positively correlated to the ec3 shape factors for both uncoated and coated pellets which was in agreement with the observations made during the filling o f capsules i.e. an elongated shape did not fill as successfully as a more spherical shape. This correlation was demonstrated more definitely by comparing the 2-dimensional shape factor c r with the percentage o f underfilled capsules, where the uncoated pellets showed a positive correlation and the coated batches almost reached the significant value at the 5 % confidence level. In
this respect the 2-dimensional shape factor appears to give better results than the 3- dimensional ec3 shape factor. Surface roughness (Ra) did not exhibit a correlation with the capsule filling performance o f uncoated pellets and thus the shape on its own appeared to influence their capsule filling ability. Looking at the values o f shape obtained for all the batches studied, the approximate limits for adequate capsule filling performance are as follows for uncoated pellets:
2-dimensional shape factor cr > 0.36
3-dimensional shape factor Cc3 >0.18 or aspect ratio AR< 1.2.
These are the values o f shape measured for the OVO pellets which were visually oval; any further elongation or deviation from the spherical shape led to an obvious deterioration in capsule filling ability as seen with the DUO, LDO and CYO pellet batches. The capsule filling results for the coated pellet batches were very poor compared to the uncoated pellets and probably due to the presence o f the filmcoat. However, if the process were improved e.g. by the addition o f a lubricant or by adapting the equipment to obtain better flow o f product, the point where the process becomes unacceptable would again coincide with the shape measured for the 0V 5 pellet batch. Thus for coated pellets the limits would be similar to the values for uncoated pellets with the exception o f the 3-dimensional shape factor Cc3 which should be > 0.29. This increase in the limiting value is a reflection o f the ec3 shape factor finding that coat 5 pellets o f the elongated batches (OV, DU, LD and CY) had higher values i.e. better shape, than their uncoated counterparts. The actual limiting values may be considered to be fairly low since the theoretical value for a perfect sphere is 1.000 (for or, ec3 and aspect ratio), and the values for the best capsule filling pellet batches (SP and GR) were only -0 .5 for Or, -1.1 for AR and -0.35 for the ec3 shape factor. However, the pellet batches studied were selected to give a fair representation o f shapes which could be produced by commonly used processes such as extrusion-spheronisation and granulation, and thus the range o f shape measurements and limits can be considered to be realistic.
Further, the relationships between different shape factors was investigated. The two three-dimensional shape factors, Cc3 and Heywood’s coefficients were compared. The
rank orders for Cc3 and Heywood’s surface coefficient, f, showed very good correlations for both the uncoated and coated pellets. However the volume factor, k, did not correlate with the ec3 shape in either case. Looking at the aspect ratio and the shape factor Cc3, negative correlations were found, and the same applied to aspect ratio and the 2-dimensional shape factor cr. This shows that overall all the shape factors ranked the shape o f the pellet batches in a similar fashion. However this does not overcome the fact that aspect ratio was not as efficient as the other shape factors in distinguishing between batches, and that Heywood’s shape coefficients produced misleading results in cases where pellets deviated highly from the spherical shape.
Rank correlations such as these only demonstrate the presence o f a relationship between the rank orders o f two different properties, they do not provide a numerical value or indicate the extent o f the relationship. In this study though, the correlations highlight the properties that are influenced by shape, and also aid in comparing the different methods available for characterizing shape. Based on this information, further studies could be carried out to study the possible relationships in depth.