I will never forget that day. Freshly graduated and discovering clinical supplies 10 years ago, I found a forgotten piece of paper on a client's desk in a meeting room. It listed rules of thumb for calculating study drug overage (excess of drug not dispensed to patients but necessary to cover for uncertainty, network effects, damaged and lost...). Something like:
- Minimal overage: 10%
- If study has titrations: +20%
- If study has more than 2 treatment arms: +15%
My illusions of mathematically sound industry practices had just shattered! There are so many examples of how these rules could go wrong, but this is not the purpose of this post. Just promise me not to use them!
Over the past decade, clinical supplies chain management has significantly evolved with the ever-increasing trials complexity and pressure on costs. Quite a few vendors now propose professional, dedicated software solutions. In this post, I would like to propose one possible classification of these solutions and highlight the differences, oriented towards the depth and accuracy of calculations.
Back of the envelope
Header says it all. Quick calculations based on experience, intuition. Each supply planner has his/her own approach.
Generally not a safe way to take decisions during the planning or execution of the trial, but it could sometimes make sense in early planning, for example to improve the design of the supply chain, packaging... when accurate and comprehensive data is missing anyway.
A powerful, flexible tool. Users understand perfectly the input and output of their custom formulas... until they don't. Spreadsheets have a tendency to grow complicated, the custom changes get forgotten, and progressively errors and miscalculations creep in. Before you know it, you'll have created your own Frankensheet that will guarantee at best a few headaches and hours of maintenance at each trial.
I have seen however some companies where template clinical supplies spreadsheets are managed and controlled by IT, and are partially locked for user edition. A good way to alleviate some of the drawbacks of the method.
Forecasting software bring a clean cut between clinical supply chain generic modelling which should not be modified by users, and the model and data relative to one particular study.
More standardised than the spreadsheets, more maintainable, and leaving the programming to specialised vendors, this is to me the first "professional" level of clinical supplies management systems.
Within the related solutions, some cover only enrolment or kit demand, while some others include supply planning and shipment scheduling, upstream (DP, API) planning and possibly integrations with other key systems (ERP...).
A weakness of this approach (and of all of the above) however, is that only the average requirements can be forecasted, and any overage needs to be defined manually.
When average projections are insufficient, it becomes necessary to resort to more advanced calculation techniques such as Monte-Carlo simulations.
Models in this case include statistical distributions of possible variables in the system (e.g. enrolment rates, site opening, randomisation, titrations...), and attempts to derive experimental distributions of site/depot demand to then calculate release and distribution plans.
These solutions bring an answer to the calculation of overage, but they depend a lot on the accuracy and completeness of the data you can feed them with. Also, with the complexity, those systems tend to behave as "black boxes", implying a lack of control and visibility on how the calculations are performed.
Analytical / statistical models
An alternative to simulation is to analytically compute statistical distribution models. This is more powerful than simulation, but quickly limited by the complexity of the resulting formulas.
Automated scenario testing
Once a particular scenario can be simulated, and a supply strategy can be calculated, it is possible to have a system automatically test variants of a reference supply plan.
If KPIs are defined, such as the service level and total supply chain cost, the tested scenarios can be compared to one another and a better strategy can be selected.
The notion of optimisation is very often misused. Rigorously, it means the selection of a best element (with regard to some criteria) from some set of available alternatives (thanks Wikipedia). Optimisation is achieved through the use of mathematical techniques such as (Non-)Linear Programming, Mixed Integer Programming, Constraints Programming, Heuristics etc.
In the context of clinical trial supply chain, it would mean finding the best supply chain parameters (IRT resupply parameters, depot management, packaging and labelling schedule...), by simultaneously optimising several competing objectives such as total cost of supply and risk of missed doses. This is called multi-objective optimisation.
No solution on the market is offering this as of today and to the best of my knowledge. Maybe in the future?
So which one is best?
There is no simple answer. The actual question is to find out which one is best for you. Here are some key questions to consider.
What are your objectives? Which decisions are you looking to take? Is it more early planning, or rather execution? Is it more cost reduction, control of trials during execution, simplification of the planning tasks ...?
What is the quality of the data you have? You know the idiom, garbage in, garbage out. If your enrolment estimates are typically poor, a simpler solution may be more appropriate than feeding abracadata to an advanced one.
How complex or costly are your trials? For large trial portfolios, how do trials classify against complexity and cost? Small, simple but expensive trials are sometimes even better managed "manually", with detailed tracking and a level of out-of-the-box thinking that no generic system could cover.
The more advanced the solution, the more costly and/or time consuming it becomes, so assessing the costs/benefits of the different options is key. Not all trials are worth the same effort, and perhaps several solutions should be used in conjunction, or based on a segmentation of the clinical trials.
What are the skills and backgrounds of your supply chain team members? Again, the more advanced technologies require a deeper understanding of supply chain principles, statistics etc. A choice may need to be made between training, hiring, or outsourcing supply chain management for the most complex/costly trials.