In this second part of his article,
1 Nigel Fletcher, principal consultant at Foster Wheeler Energy’s pharmaceutical division, looks at the plant design process when implementing CIP into API plants
In the first part of the article,1 the two elements of the protocol – the cleaning agents and the cleaning target – were discussed, together with a set of cleaning test points to be achieved during the procedure. In this part we will look at route selection: the order in which cleaning is carried out.
Typically, starting cleaning at the top and working downwards is reasonable – after all, fluids flow down under the action of gravity! But this is where some care is needed and knowledge of the plant’s physical design is required. If we are dealing with a simple train or set of vessels then this may dictate passing the cleaning fluids from one to the next and so on until the last, lowest one is reached.
This may work if the cleaning fluid is suitable for all the materials dirtying the plant on the way down. But problems occur when one of the materials is not soluble in the fluid flowing into the equipment. So it is necessary to identify exactly at which point to stop this flow and divert it out of the process train and introduce a new ‘solvent’ to clean the equipment downstream.
There are a couple of points to make here relating to this route selection. First, the solvent does not have to simply flow once through the equipment. If the cleaning agents table has suggested that the solubility is low or dissolution is slow then this is indicative that the solvent needs to be recycled or re-contacted with the ‘dirt’ until the solvent reaches the chosen exhaustion point.
The second point is that with slowly dissolving materials, there can be some merit in using a pulsed or burst washing technique. Here, the solvent is sprayed into the equipment in a burst lasting maybe 15 to 20 seconds and then this is allowed to drain for a short period, say 30 seconds, before the ‘burst’ is repeated. This technique can help reduce the amount of solvent used and can make the washing more effective as ‘saturated’ solution drains away from the residual dirt allowing better access to the next fresh burst wash.
If possible, the chosen routes should be washed in parallel to save as much time as possible in the overall turnaround time; i.e. while the reactor train is being progressively washed then the centrifuge or isolation train is also being washed.
Part of the route selection is also to identify the correct wash techniques to go with the route. Cascade washing, as described above, is a very simple and effective method but is not the only technique that must be considered. If we have a vessel with a problem crust or ring of material to be removed, then the main washing cascade or route may have to bypass this vessel. This is because simple flushing may not be appropriate and another technique such as refluxing or high-pressure washing may be needed. In fact, flushing may be counter-productive and cause glazing of the crust, which could give rise to a cleaning failure from this vessel.
In the case of a ‘crust’, there has to be specific information from the research data to warn of this phenomenon. Usually the crust consists of the reaction product and some of the unreacted raw materials but usually in a ‘glazed’ form. This is more difficult to remove than more common or ordinary deposits and the selection of the correct technique is critical. Typically the main techniques are:
• Prolonged flushing – often not very successful due to glazing;
• Immersion (filling the equipment with solvent until the crust is covered and then stirring, hot or cold, until the crust is dissolved) – reasonably successful when the correct solvent is chosen and the crust is a simple ring around the body of the vessel but often fails due to poor solvent selection;
• Refluxing (better known as boiling out) – gives similar results to immersion, and is more appropriate where the crusting is distributed across the whole of the internal surface of the equipment;
• High-pressure washing – a less common technique as it requires the equipment to be opened for the HP washer to be inserted, but an extremely successful technique;
• Focused washing – rarely used. This technique requires the equipment to be opened and a low pressure spray head is inserted into the equipment which directs jets at the crusted areas. Solvent is recycled. The washing uses both dissolution and dislodgement techniques and can be successful ;
• Manual cleaning – a last resort method nowadays but requires a special cleaning sequence following the manual intervention to bring the equipment back to a GMP state.
In respect of this last point, it is important to know the GMP boundary of the plant where cleaning validation can stop. This is not where cleaning stops but where the validation of the results will stop. It is also useful to know so that equipment outside the boundary can be used to receive spent cleaning solutions. It is important to know this because no samples for analysis should come from this area of the plant as they are not validated.
What this tells us is that sample points should be inside the GMP boundary of the plant and that these need cleaning just as much as the main process equipment. In other words, the sample points are part of the route selection and these routes only end up in the sample bottle. The case study below illustrates what can happen when this point is not understood.
Up until this point, only passing reference has been made to methods by which equipment is cleaned. Two techniques can be adopted here: pattern analysis and draining calculations.
Pattern analysis uses the spray device or cleaning solvent injection point as a point source. Most sprayball manufacturers make balls that have a defined spray pattern usually described as 360°, 180° upwards or downwards, 270° upwards or downwards, conical, elliptical, hollow cone etc. As well as this, the ‘throw’, or distance the jets travel needs to be known to carry out the pattern analysis.
This information is combined and used by the CIP consultant to either diagnose where a problem lies or for the CIP designer to attempt to get the CIP fluid to the right point in the equipment. The methods of pattern analysis are simple and can be used to identify exactly where the jets/sprays will contact the surface, provided the analysis is carried out in plan and elevation.
This particular method is often not carried out and the operator is disappointed by the apparent poor performance of the spray device. Figure 1 shows a simple analysis where spray devices with throws between 1.0m and 1.5m will contact the vessel surfaces. The red areas show poor contact, yellow areas may have contact but should be investigated and the white areas are those where spray contact will be satisfactory. The one small area of blue is an area where it is unlikely there will be any satisfactory contact and where uncleaned deposits may be found.
Repeating this analysis in elevation will show which nozzles will not be cleaned properly and which inserts, such as dip-pipes, will interfere with the cleaning. This information is critical to determine if any other techniques should be adopted or where to look for “dirt” so decisions can be made as to whether they need cleaning or can be accepted. This may be important if the total amount of a product held up in the process train is critical – an area that causes many disputes in the pharmaceutical industry.
Once the pattern analysis has been carried out then there has to be a consideration of how the cleaning fluid sprayed into the equipment will drain out. There are two distinct considerations with respect to fluid draining. The first is how it drains out of the vessel, and the second is how the fluid is introduced into the equipment.
A common approach to cleaning is to use the spray devices as a means of getting as much cleaning fluid into the equipment as possible. This is wrong. It is vitally important to ensure that the cleaning fluid drains out of the equipment as fast as it is introduced into it. If this does not happen then there will be ‘puddling’ or the collection of fluid in the bottom of the equipment.
This gives rise to two possibilities: first, the possibility of re-deposition of the “dirt” just removed or the settling out of suspended material in the base; second, there is a possibility of starving the CIP system if the fluid is in recycle because the fluid is being held up the system. This, in turn, means slowing down the cleaning process, which is clearly not desirable. If the calculations show that the fluid cannot leave fast enough then adopting the burst spraying technique, described above, may be useful to limit or eliminate the hold-up of fluid in the equipment.2,3
Burst spraying can also be useful where glazing appears to be a problem and here the second aspect of determining draining can become important. There is a calculation that may be undertaken that determines the time taken for the film formed on top of the “dirt” or any surface to drain away. This can be used to determine the burst wash duration and subsequent draining time prior to the next burst wash.4
The calculated draining time can also be used to decide when to step forward in the cleaning protocol as once this time has elapsed then the equipment can be declared fluid free. These calculations are not perfect but should be considered indicative and a first estimate for commissioning purposes.
The information is now available to create the first draft of the CIP protocol. This information will include the best cleaning agent with acceptable cost and environmental properties, GMP considerations, cleaning objectives, route selection and pattern analysis (with drainage calculations if necessary).
Once the draft protocol is prepared it must be tested in a full plant trial using riboflavin on a new plant or, possibly, the real process on an existing plant. The results of the trials then must be fed back into the draft protocol and, if necessary, these results taken through the roadmap process and a second draft and trial undertaken. At the end of this second trial it should then be possible to ‘finalise’ the protocol and submit it for validation and approval by QA.
It may be that there are difficulties in one or two areas not anticipated during the “design” process. These will have to be addressed and specific solutions designed for them. This is almost inevitable and should not be seen as a failing of the CIP designer, providing the number of the problem areas is small. It is important to analyse these problems scientifically, methodically and then spend adequate time working out solutions.
So far, the left hand side of the Foster Wheeler CIP Roadmap given in part 1 has been considered but not the physical design of the plant. The design of piping that drains, correct valve selection, equipment design details, layout considerations and many other aspects of the plant design have to be taken into account when looking at the cleaning protocol. Ideally, plant design and cleaning protocol design happen simultaneously with the designer of the cleaning protocol being directly involved with the design of the plant itself. Sadly, this rarely happens and the cleaning protocol designer or cleaning consultant is left to make the best of the situation.
Fortunately, more attention is now being paid to cleaning and more effort is being put into the “cleanability” of the plant at the design stage. The above techniques can be used to assist either with the analysis of cleaning problems or to design cleaning protocols to avoid CIP failures.
Once the problems have been identified and analysed then the information is available to start the process of identifying a solution. Here the old adage of “keep it simple” should never be forgotten.
A case study on cleaning failure
At one plant visited by the author, a large number of cleaning failures were reported. On inspection, the sample points were valves on branches off the main process lines with short, small bore tail-pipes about 300–500mm long terminating in an extracted box. These boxes were covered in product residues and splashes of unidentifiable materials. The operators had to take samples from these tail-pipes straight into sample bottles for analysis.
Several points became clear after seeing samples being taken.
First, the extracted sample boxes were never cleaned so there was a clear risk that when the operator opened the box he risked contaminating the open sample bottle.
Second, the sample valve had been mounted with a relatively large dead-leg from the main process line (many times more than 6 diameters) so the branch was not being cleaned during the main cleaning sequence.
Third, the branch, valve and tail-pipe were not part of the cleaning sequence, so there was no obvious way in which the sample point could reach the same level of cleanliness as the main process piping. Finally, there was no strict timed period for flushing the sample point when taking a sample. So although the operators did flush the sample point, the flushing time was variable and considerably shorter than required to ensure that the sample would not be contaminated from the previous one.
Once proper cleaning and flushing had been instigated on the sample points the rinse analysis cleaning failures declined significantly.