Why Industry Should Model and De-Risk its Pharma Supply Chains

As new biotech drugs come on the market, they will require supply chains that are more agile but less efficient. AstraZeneca’s MedImmune biologics unit looks to boost agility by modeling its supply chains and using a neural network to de-risk them.

The 14-year-old girl competed on her high school track team until she developed rheumatoid arthritis so painful she couldn’t get out of bed in the morning.

Then MedImmune launched Enbrel, a first-in-class tumor necrosis factor inhibitor that gave her her life back – but only while supplies lasted.

She would receive her subcutaneous injection on Thursday nights, Andy Skibo recalled. “By Friday she can walk downstairs and have breakfast. By Monday she can walk to school. Enbrel supply runs out. Within a week, she's back in her bed.”

MedImmune divided the job of updating patients about its efforts to end the shortage. Skibo was assigned to the bedridden athlete. “I had to talk to her every month for the 14 to 18 months that it took us to double Enbrel capacity. I will never, ever do that again if I can avoid it.”

That searing experience may have been over a decade ago, but without heightened vigilance, it is increasingly likely to recur in the next few years.

It was a combination of difficulties that drove the Enbrel shortage: difficulties anticipating demand, boosting yield and adding contract manufacturers (Also see "Enbrel Supply Interruption Will Be Resolved By June, Immunex Says" - Pink Sheet, 1 Apr, 2002.).

A similar combination of factors will drive the next shortages. Skibo, who is head of biologics operations and engineering, MedImmune/AstraZeneca, told the audience at Washington’s Mayflower Hotel what those factors are and what his company is doing about them.

Skibo was speaking in June at a quality manufacturing conference that FDA, the International Society for Pharmaceutical Engineering and the Product Quality Research Institute had organized.

He noted that two or three years ago, he had raised concerns that rigorous supply chain cost control was leading to quality issues and potential shortages (Also see "Pharmaceutical Cost Cutting Seen Hurting Drug Quality" - Pink Sheet, 26 Jul, 2012.) (Also see "Skibo: Careful Assessment of Supply Chain Risks Could Prevent Shortages" - Pink Sheet, 25 Jul, 2013.). “What’s happened in the meantime,” he said, “is a profound change in the entire industry.”

He shared a McKinsey & Co. chart showing that over the past 20 years, the pharmaceutical industry has doubled earnings even as it doubled its spending on research & development. It did this by taking half the cost out of the supply chain.

Today’s highly efficient global pharmaceutical supply chain can absorb much of the risk from its low inventories, high network utilization and poor visibility – but only because there are multiple innovator products for the same indication and even more generic options in the mature, declining markets it serves.

However, with R&D finally bearing fruit, a different challenge is emerging. Tomorrow’s industry will be replete with niche biotechnology products that like Enbrel will require supply chains that are more agile than efficient.

“This tidal wave is crashing down on a network that’s not the network of 15 years ago,” Skibo added. “Five years ago, you couldn’t get a board of directors to approve a bio bulk facility to save your soul because the industry was awash in unused capacity.”

Now there are 17 biotech facilities under construction, and they’ll meet the new demand when it hits in 2017 through 2020 – but only if everything goes perfectly.

De-risking the supply chain

To the extent that it’s focused on individual facilities, FDA’s quality metrics initiative won’t achieve one of its goals, which is to prevent shortages, in Skibo’s view (Also see "FDA’s Quality Metrics Proposal Poses Challenges, Including Gaming, Cost" - Pink Sheet, 13 Aug, 2015.).

“Understanding the risk associated with a single production site is … of zero value in understanding the risk to overall product supply,” he said.

What’s important in getting that overall perspective is to assess four key factors for each node in the supply chain:

• Inventory policy
• Network utilization
• Redundancy
• Visibility

Rather than just focus on preventing quality issues, the key is to ensure that if a quality issue arises at one node, there is sufficient inventory, network capacity, redundancy and visibility for other nodes to take up the slack.

Quality metrics analysis can provide additional benefit, he added, in terms of characterizing the risk at each node as well as the cost of reducing that risk.

Sensing danger with a neural network

It is possible to model supply chain risks, Skibo said, citing recent University of Iowa research.

A colleague of Skibo’s went further, explaining how Medimmune is modeling its supply chain.

R. Lance Minor, VP, Global Biologics Operations Network Strategy and Performance, said that the company has built a stochastic model simulating high-level risks throughout the supply chain “so that we could understand … all the inherent risks – operational, manufacturing, process, distribution – so that we could identify how to best ensure highest customer service.”

Ultimately, the company aims to create a neural network, he said. “We want a feedback mechanism throughout the whole process, throughout every element, to understand what our risks are and how to feed that back into each of these basic elements.”

The vision, he explained, is to have “this self-healing, self-correcting tool where we have real-time data feeding directly from SAP [MedImmune’s enterprise application software vendor], directly from our manufacturing floor into the model, so that we have a clear picture of the risks that we are seeing on a daily basis."

Doing the math on shortages

That kind of a self-correcting supply chain could be critical as manufacturers introduce more products like MedImmune’s Synagis into the market.

Synagis is the only product available for preventing respiratory syncytial virus in premature infants. “I can do the math of how many premature infants will have a severe outcome if I don’t supply Synagis,” Skibo said.

If it works as intended, the neural network MedImmune plans to build will ensure he never has to make that calculation – or those phone calls.