Abstract
Background UK Biobank is a prospective study that is collecting biological samples and health and lifestyle data from 500 000 volunteer participants over a 4-year period. These data will be used to facilitate biological and medical research.
Methods Modern manufacturing principles were used to direct the development of the sample processing facility and automated systems.
Results A fit for purpose facility comprising technology, systems, dedicated process, infrastructure and an appropriate staff structure has been implemented that will deliver and maintain a resource that will support the long-term goals of the UK Biobank study.
Conclusions Modern manufacturing principles are appropriate for use in the development of a high throughput biological sample processing facility.
The UK Biobank project—understanding the challenge
The UK Biobank sample handling and storage protocol describes a process that will collect samples of blood and urine from 500 000 participants over a 3.5-year period.1 By the end of the recruitment phase, 15 million aliquots will have been produced and stored in two separate ultra-low temperature archives. Even in the early stages of protocol development, it was clear that a high-throughput process established in a centralized, automated facility would be required (Figure 1). The performance parameters required to process this volume of samples was the starting point for the procurement of the automation rather than designing the protocol around available robotics.
Sample processing and storage demands for the UK Biobank study (top half of figure) and on a daily basis (bottom half of figure) defined by the Sample Handling and Processing Protocol
During the design and implementation of the facility, we took two approaches; the first was to learn from comparable projects where the expected gains and successes were either not realized as quickly as had been anticipated or were not achieved at all. The second was to apply industrialization methods from manufacturing engineering where complex processes and facilities are implemented in parallel with product development despite uncertainty over their exact configuration until close to full operation.
A number of industries had varying degrees of success when they first tried to implement high-throughput automation to previously well-characterized processes in manual or semi-manual operations.2,3 At UK Biobank we have tried to learn from these previous efforts, especially those involving the life sciences and pharmaceutical industry.
Maintain control of inputs and outputs: avoiding the Forrester effect
In any process with more than four or five discrete steps with multiple inputs, the interactions between the various process elements become very hard to predict. If control is not maintained over the inputs and outputs of individual process steps, the overall operation becomes increasingly chaotic. Eventually it becomes entirely reactive and inefficient as it is impossible to schedule and control the overall operation.
This problem has been well characterized in other industries by Forrester.4 Instead, the focus should be on forward planning to ensure all inputs and outputs and processing capacity are available in sufficient quantity to enable the process to be run efficiently. This is particularly important for UK Biobank because of the continuous supply of blood and urine from the assessment centres that degrades if it is not processed and stored within 24 h.
Standardize process and quality
To ensure maximum future value of the UK Biobank sample resource, high-quality samples for comparative studies should be available. A clear separation between research (protocol piloting and process design) and operations (running the high-throughput process in the live study) is required so that protocols and technology are designed and tested with proper tolerances and performance characteristics. The sample handling and storage protocol was science led but each iteration was tested for process feasibility and highly standardized protocols and systems have been developed. In UK Biobank, this has delivered an affordable operation that runs at the required throughput and produces sample aliquots of high quality.
Achieve the correct balance of staff
Once the sample handling and storage protocol was approved, it was necessary to implement it in a high-throughput facility. This environment is more akin to a production facility and requires specialist technical staff with a focus on clearly defined targets of data and product quality and throughput. The staffing, culture and management of this part of the process should be quite different to the research environment.
Locate in fit for purpose facilities
The UK Biobank sample handling and storage protocol has been translated into a high-throughput operation involving industrial automation, heavy engineering and very large-scale sample cryopreservation facilities each requiring specialist plant and utilities. Therefore, a facility was designed and constructed in parallel with the late stages of protocol development when we were able to specify exact requirements. This has produced a low-cost fit for purpose building, specified exactly to our needs rather than a potential compromise that would result from trying to fit the operation into an existing building on a university campus or science park.
Thoroughly test and validate all new processes and technology off-line before implementation
It is likely that, over the course of the recruitment period, new processes, technology or systems will become available that may improve the sample processing operation. However, these modifications will not be integrated without careful off-line testing and validation of the impact on the entire process. Once this has been done, a careful integration plan (involving off-line technology integration, systems adaptation, training, testing, etc.) should be implemented to avoid affecting the overall process.
UK Biobank identified early that is should adopt a fully industrialized approach to the implementation of the overall processing and archiving capacity. There have been a number of papers published on transferring lessons from manufacturing engineering to the implementation of large-scale automation projects in the pharmaceutical and life science area. Archer5 was the first to describe adopting an industrialized approach building on his engineering background and work implementing automated processes in a number of industries.
As well as trying to learn from the experience of others, we have also followed the approach used in manufacturing engineering in the planning, design and implementation of complex processes and supporting infrastructure. Briefly, this approach relies on careful design, analysis and prototyping to minimize process risk and to develop and test the main technological components of the planned systems. Only then are the large capital systems constructed and commissioned. The approach can be split into three phases.
Design phase: develop concept and identify major risks to success—feasibility study and Failure Modes Effect Analysis (FMEA)
This approach was successfully applied to the requirement for automated fractionation of blood samples into their constituent elements; plasma, buffy coat (white cells) and red cells as specified in the sample handling protocol. The daily production requirement of 19 800 sample aliquots makes processing the tubes manually impractical. A feasibility study was therefore commissioned with a specialist supplier of integrated robotic systems, RTS Life Sciences PLC (http://www.rtslifescience.com/). The challenge from an automation perspective is the inherent variability in blood (volume, turbidity, viscosity) and consequently the fractions that are created when it is centrifuged (Figure 2).
Samples of whole blood in vacutainers centrifuged at low speed. The figure shows the variability in total volume (tubes 1 and 2), fraction volume (the plasma in tubes 2 and 3) and in turbidity (tubes 3 and 6)
The difficulty for an automated system is the detection of the fraction boundaries and the subsequent separation into aliquot storage tubes. The technology selected during the feasibility study was an imaging system that digitally photographed the centrifuged tubes and used a software application to determine the exact positions of the fraction boundaries. These data are transmitted to a pipetting robot enabling to it aspirate the fractions correctly.
In parallel to the design study a ‘Failure modes effects analysis’ (FMEA) was run.6 This is a method developed for systems engineering that examines potential failures in products or processes. It is used to understand what failures may occur, how detectable these failures would be and what the impact would be on the overall process. Each potential failure mode is scored and a table of highest priority failures states created. This serves two purposes: (i) influencing the design (ii) implementing procedural changes to accommodate these failures as they occur. At this stage of the project there is still scope to modify the proposed design in light of the FMEA results; this enabled some relatively subtle but extremely important design changes which eliminated those failure modes which could greatest impact on the overall operation and consequently the laboratory's ability to produce the 19 800 aliquots for storage each day. An example of a design change resulting from the FMEA was the inclusion of a second aliquot tube input store (the piece of equipment that holds the empty tubes to be filled with the blood fractions). Should a single unit fail, there would be no way of introducing the empty tubes, so the entire process would stop. The solution to include a second unit thereby increased capacity and process redundancy.
By the end of the feasibility and FMEA study, the major risk areas associated with the high-throughput automated fractionation of whole blood had been addressed and demonstrated to be technologically feasible. The final design was technically better, more robust and cheaper than that first proposed in the procurement process (Table 1). A similar approach was taken with the providers of the –80°C sample archive The Automation Partnership (Cambridge) described in the sample handling protocol.1 As with the technology for the automated fractionation of blood, the design for this archive was greatly improved (including a fundamental modification to a redundant, low-risk cooling system), the overall technology risk identified and addressed and the cost reduced significantly (Table 2).
Summary of the development of the automated blood fractionation technology following a design study and failure modes effect analysis (FMEA)
| Initial design proposal | Final design and prototype |
|---|---|
| Five dedicated processing stations | Three generic, redundant processing platforms |
| Serial, coupled processes | De-coupled, cellular processes |
| Localised cooling | Single, cooled environment |
| Variety of tube formats | Single tube format |
| Risks around key technology elements | Robust technology solutions |
| 42% reduction in cost |
| Initial design proposal | Final design and prototype |
|---|---|
| Five dedicated processing stations | Three generic, redundant processing platforms |
| Serial, coupled processes | De-coupled, cellular processes |
| Localised cooling | Single, cooled environment |
| Variety of tube formats | Single tube format |
| Risks around key technology elements | Robust technology solutions |
| 42% reduction in cost |
The initial design proposal was based around five dedicated robotic platforms each capable of processing one of the vacutainer tube types (but none of the others). The processes were serial and highly coupled and were therefore prone to breakdown and quality issues. Other risks were introduced by localised cooling, a variety of tube formats (that greatly increased process complexity and time) and uncertainty over the feasibility of various technology elements. After the design and FMEA, the system consisted of three fully redundant platforms (reducing risk and cost) that had the constituent process parts de-coupled (reducing the risk of quality issues). Process complexity was further reduced by using a single tube format and tested technology.
Summary of the development of the automated blood fractionation technology following a design study and failure modes effect analysis (FMEA)
| Initial design proposal | Final design and prototype |
|---|---|
| Five dedicated processing stations | Three generic, redundant processing platforms |
| Serial, coupled processes | De-coupled, cellular processes |
| Localised cooling | Single, cooled environment |
| Variety of tube formats | Single tube format |
| Risks around key technology elements | Robust technology solutions |
| 42% reduction in cost |
| Initial design proposal | Final design and prototype |
|---|---|
| Five dedicated processing stations | Three generic, redundant processing platforms |
| Serial, coupled processes | De-coupled, cellular processes |
| Localised cooling | Single, cooled environment |
| Variety of tube formats | Single tube format |
| Risks around key technology elements | Robust technology solutions |
| 42% reduction in cost |
The initial design proposal was based around five dedicated robotic platforms each capable of processing one of the vacutainer tube types (but none of the others). The processes were serial and highly coupled and were therefore prone to breakdown and quality issues. Other risks were introduced by localised cooling, a variety of tube formats (that greatly increased process complexity and time) and uncertainty over the feasibility of various technology elements. After the design and FMEA, the system consisted of three fully redundant platforms (reducing risk and cost) that had the constituent process parts de-coupled (reducing the risk of quality issues). Process complexity was further reduced by using a single tube format and tested technology.
Summary of the development of the −80°C automated sample archive following a design study and failure modes effect analysis (FMEA)
| Initial design proposal | Final design and prototype |
|---|---|
| Two separate stores | One integrated store |
| Electrical/mechanical cooling of −80°C environment | Low risk cooling system |
| Variety of tube formats | Single tube format |
| Significant engineering and acoustic issues | Elimination of engineering and acoustic issues |
| Significant facilities demands | Standard facility requirements |
| Very high running costs | 27% reduction in build cost, significant reduction in operation costs |
| Initial design proposal | Final design and prototype |
|---|---|
| Two separate stores | One integrated store |
| Electrical/mechanical cooling of −80°C environment | Low risk cooling system |
| Variety of tube formats | Single tube format |
| Significant engineering and acoustic issues | Elimination of engineering and acoustic issues |
| Significant facilities demands | Standard facility requirements |
| Very high running costs | 27% reduction in build cost, significant reduction in operation costs |
The initial design proposal was based around two separate stores, cooled by conventional electrical compressors operating at the limit of their capability with concomitant acoustic and facilities issues. The design study and FMEA resulted in a single store with a very low risk cooling system using liquid nitrogen. This eliminated the acoustic and facilities and reduced cost significantly.
Summary of the development of the −80°C automated sample archive following a design study and failure modes effect analysis (FMEA)
| Initial design proposal | Final design and prototype |
|---|---|
| Two separate stores | One integrated store |
| Electrical/mechanical cooling of −80°C environment | Low risk cooling system |
| Variety of tube formats | Single tube format |
| Significant engineering and acoustic issues | Elimination of engineering and acoustic issues |
| Significant facilities demands | Standard facility requirements |
| Very high running costs | 27% reduction in build cost, significant reduction in operation costs |
| Initial design proposal | Final design and prototype |
|---|---|
| Two separate stores | One integrated store |
| Electrical/mechanical cooling of −80°C environment | Low risk cooling system |
| Variety of tube formats | Single tube format |
| Significant engineering and acoustic issues | Elimination of engineering and acoustic issues |
| Significant facilities demands | Standard facility requirements |
| Very high running costs | 27% reduction in build cost, significant reduction in operation costs |
The initial design proposal was based around two separate stores, cooled by conventional electrical compressors operating at the limit of their capability with concomitant acoustic and facilities issues. The design study and FMEA resulted in a single store with a very low risk cooling system using liquid nitrogen. This eliminated the acoustic and facilities and reduced cost significantly.
Prototyping phase: test and refine technology elements
Before committing to the purchase and construction of the final systems, the highest risk elements were built and tested to ensure they were capable of functioning to the required throughput and quality standards. This method adopts the approach developed by Genichi Taguchi who recognized that 80% of quality problems result from the design of systems and processes and only 20% during actual production (for a good discussion of these approaches see ref.7). An example was the sample storage towers contained within the −80°C compartments of the automated sample archive. A single tower was constructed and rigorously tested for its thermal properties and sample storage and retrieval at ultra low temperatures. It was not necessary to test the central aisle robot (involved in the transfer of samples from the loading buffer to the storage towers) because this was an off-the-shelf product with proven performance data.
By this stage, all of the processes and supporting systems had been fully described and were being implemented and validated in parallel with the technology development.
Implement and commission
Once the discrete elements of the systems have been constructed they must be implemented. At UK Biobank this involved the delivery and integration of all of the sample processing automation and software at the central processing laboratories in Cheadle. Each component was thoroughly tested against agreed test schedules that reflect actual operation. During this period all Standard Operating Procedures (SOP), training plans and health and safety documentation were written and implemented. The final process, that of facility commissioning, involves the integration of all of the various elements (process, people, technology, systems, facilities) into a validated operation capable of meeting the demands of the sample handling and storage protocol. The importance of, and the time required for, the commissioning stage is often overlooked but failure to address this properly will result in many unanticipated faults and problems that will affect the overall performance of the facility.
Build in quality: prevent quality problems (quality assurance), detect quality problems (quality control)
The three steps described above represent a sequential process. However, it is a fundamental part of a manufacturing engineering approach to build in quality considerations throughout the whole of design to commissioning the project. In UK Biobank we have implemented ISO:9001:2000 throughout the organization. This quality management system sets standards for any part of the organization that has the potential to impact the sample of data resource. In the context of the laboratory systems, it has been used to implement quality assurance and quality control procedures for the prevention and detection of quality problems. For example, UK Biobank has worked with manufacturers to develop blood collection tubes labelled with unique barcodes (Figure 3). These barcodes link the participants in the assessment centres to their blood and urines samples in a secure and anonymous way that prevents data entry errors. They also prevent operators in the laboratory incorrectly processing the samples because the automation is programmed to recognize specific barcodes for specific protocols.
A UK Biobank blood collection tube showing the unique barcode that links the tube to a specific participant and identifies the tube type to the automation ensuring correct processing. The first three characters identify the tube type e.g. 007 = Acid citrate dextrose tube. The code also contains a Modulus10 checksum, an algorithm that is used to confirm that the label or manually typed human readable number has been entered correctly
Discussion
The process of design through to commissioning of the UK Biobank sample handling and storage facilities has taken about three years and has followed principles used in manufacturing engineering. Rather than focussing purely on technology, it has addressed all aspects of process, IT, facilities, people and technology in parallel with the development of the protocol. We now have a robust facility that is capable of operating at high-throughput and to high-quality standards that will process and store the samples securely for many years. In the preface to this supplement the authors note that they hope that the information will be useful to other groups planning similar studies. The principles described here are an important component of building a large Biobank and should not be overlooked.
Conflict of interest: None declared.
