Memorandum

Insulet hosts webinars and investor presentations highlighting Omnipod 5 pivotal data and successes in broad populations – March 25, 2021 - Draft

Executive Highlights

  • With the readout of Omnipod 5 pivotal data at ENDO 2021 (see data from adolescents and adults here and pediatrics here) and anticipated launch of the system in 1H21, Insulet hosted three events this week around the hybrid closed loop system.

    • Omnipod 5 Pivotal Data Investor Webcast – Saturday, March 20, 2021 – webcast, presentation slides

    • Omnipod 5 Automated Insulin Delivery System: Pivotal Data Release & Product Debut Presentation – Tuesday, March 23, 2021 – recording available in ENDO 2021 conference platform (requires ENDO 2021 registration)

    • Omnipod 5 Product Debut & Clinical Study Results – Thursday, March 25, 2021 – recording and slides not yet available

  • Not surprisingly, each Insulet event began with a summary of results from Omnipod 5’s pivotal trial in adults/adolescents and children. Below, we provide some context and discussion of the Omnipod 5 results compared with Tandem’s Control-IQ and Medtronic’s MiniMed 780G pivotal trials. In general, all three automated insulin delivery systems demonstrate impressive outcomes compared to usual care and the new Omnipod 5 results continue to build on a growing base of evidence that AID is safe, effective, and easy-to-use. 

  • During Saturday’s investor call, we heard from Dr. Bruce Bode (Atlanta Diabetes Associates), one of the pivotal’s Principal Investigators. Overall, Dr. Bode summarized by saying, “This data is incredible” and the system is effective for supporting patients with diabetes and reducing burden because of its ease of use and algorithm adaptability, which makes it particularly well-suited to type 2s and patients coming from MDI therapy.

  • Saturday’s call also featured investor Q&A with CEO Shacey Petrovic, CMO Dr. Trang Ly, and SVP of Innovation and Strategy Eric Benjamin, who shared insights into the impact of Omnipod 5 on Insulet’s market position and discussed the advantages of Omnipod 5 compared to other AID systems and next steps toward FDA approval. In particular, their discussion of the features and advantages of the Omnipod 5 system were fascinating, not only in terms of how Insulet positions itself in the AID market, but also in how they see those features and advantages driving overall AID and pump market growth looking ahead, as well as the movement into the insulin-requiring type 2 and MDI spaces.

  • Insulet’s Thursday morning webinar included four detailed case-study presentations from Dr. Anders Carlson (International Diabetes Center) and Dr. Jennifer Sherr (Yale) demonstrating the applicability and effectiveness of Omnipod 5 across a wide variety of patient scenarios. Across all presentations, the most common theme was ease of use for patients and simplified diabetes management reducing the burden of diabetes for patients and caregivers alike.

1. Omnipod 5 pivotal trial shows excellent glycemic outcomes for children, adolescents, and adults; comparison with Tandem’s Control-IQ and Medtronic MiniMed 780G pivotal trials

Not surprisingly, each Insulet event began with a summary of results from Omnipod 5’s pivotal trial in adults/adolescents and children. We summarize the results from the trial in the table below – see our coverage from ENDO 2021 Day #1 for a full breakdown of these results. Here, we provide some context and discussion of the Omnipod 5 results compared with Tandem’s Control-IQ and Medtronic’s MiniMed 780G pivotal trials. In general, all three automated insulin delivery systems demonstrate impressive outcomes compared to usual care and the new Omnipod 5 results continue to build on a growing base of evidence that AID is safe, effective, and easy-to-use. 

 

Adolescents and adults pivotal

Pediatrics pivotal

Study sample

128 type 1s, ages 14-70

112 type 1s, ages 6-<14

 

Baseline

Study

Baseline

Study

 

Time in Range

65%

74%

52%

68%

 

A1c

7.2%

6.8%

7.7%

7%

 

Time >180 mg/dl

32%

25%

45%

30%

 

Time >250 mg/dl

10%

6%

19%

10%

 

Time <70 mg/dl

2%

1.1%

1.4%

1.5%

 

Time <54 mg/dl

0.2%

0.2%

0.1%

0.2%

 

  • From a study design standpoint, Insulet’s Omnipod 5 pivotal was a three-month, single-arm trial. Participants went through a 14-day baseline phase, wearing Dexcom G6 and performing their usual insulin regimen, followed by the three-month period with Omnipod 5. The single-arm pivotal design is similar to that of Medtronic’s MiniMed 780G. The three-month MiniMed 780G trial also included a cross-over halfway through the trial (45 days) with half of participants changing their target glucose set points from 100 mg/dl to 120 mg/dl, and the other half changing from 120 mg/dl to 100 mg/dl. In contrast, Insulet’s trial allowed users to choose their glucose set points throughout the trial; as we heard at ENDO 2021, the lowest 110 mg/dl set point was the most popular choice throughout the trial. The two single-arm pivotal contrast with Tandem’s Control-IQ pivotal study design, which was a longer, six-month randomized control trial. The Control-IQ trial randomized half of its participants to sensor-augmented pump (t:slim X2 pump and Dexcom G6 CGM in open loop) and the other half to Control-IQ.

  • For adults and adolescents, Insulet’s, Tandem’s, and Medtronic’s pivotal trial results show fairly similar glycemic improvements. All three trials finished with mean Time in Range in the intervention group above the consensus target for 70%, a remarkable achievement. Time in hyperglycemia remains a challenge, but all three systems made massive improvements in reducing time >180 mg/dl. Additionally, all three hybrid closed loop systems demonstrated that Time in Range and A1c improvements did not come at the cost of more time in hypoglycemia, with all hypoglycemia endpoints meeting consensus targets. As we’ve noted with most AID studies to date, much of the Time in Range improvement is driven by improved overnight control. All three systems delivered 10 percentage points or greater improvements in overnight Time in Range (with a particularly high baseline for the MiniMed 780G sample).

 

Insulet Omnipod 5

Tandem Control-IQ

Medtronic MiniMed 780G

Participants

129 type 1s, ages 14-70

168 type 1s, ages 14-71, 2:1 Control-IQ : SAP randomization

157 type 1s, ages 14-75

 

Baseline -> Study

Change

Control -> Intervention

Change*

Baseline -> Study

Change

Time in Range

65% -> 74%

+2.2 hours/day

59% -> 71%

+2.6 hours/day

69% -> 75%

+1.4 hours/day

A1c

7.2% -> 6.8%

-0.4%

7.4% -> 7.1%

-0.3%

7.5% -> 7%

-0.5%

Mean glucose

161 -> 154 mg/dl

-7 mg/dl

170 -> 156 mg/dl

-13 mg/dl

153 -> 148 mg/dl

-5 mg/dl

Time >180 mg/dl

32% -> 25%

-1.8 hours/day

38% -> 27%

-2.4 hours/day

28% -> 23%

-1.2 hours/day

Time >250 mg/dl

10% -> 6%

-1 hour/day

--

--

6.2% -> 4.6%

-23 min/day

Time <70 mg/dl

2% -> 1.1%

-13 min/day

1.9% -> 1.4%

-13 min/day

3.3% -> 2.3%

-14 min/day

Time <54 mg/dl

0.22% -> 0.17%

-1 min/day

0.24% -> 0.21%

-1 min/day

0.8% -> 0.5%

-4 min/day

Overnight Time in Range

64% -> 78%

+14%

59% -> 76%

+17%

71% -> 81%

+10%

Time in closed loop

95%

--

90%

--

95%

--

*Note that the results for Control-IQ are presented as the SAP control group’s outcome -> Control-IQ intervention group’s outcome. Changes are presented as adjusted differences.

  • With similarly positive three glycemic outcomes for all three AID systems, AID is poised to continue driving greater insulin pump penetration. Ultimately, choosing between the different systems will almost certainly come down to payer coverage, and then patient and provider preferences on perceived familiarity and ease-of-use and how the “systems” fit into larger “ecosystems.”

    • From the CGM standpoint, Tandem and Insulet’s compatibility with the no-calibration Dexcom G6 is an obvious advantage; both systems will also work with Abbott eventually.

    • Form factor and business model is also an area in which Insulet stands to benefit from having the only tubeless AID system.

    • In Tandem’s most recent earnings call, the company highlighted its smaller t:sport insulin pump as a starting point for simplifying the insulin pump experience and boosting MDI and type 2 adoption.

    • Still, Insulet’s no-upfront cost, pay-as-you-go model for Omnipod 5 should be a very compelling adoption for those new to pump therapy – coverage also shouldn’t change significantly from Omnipod DASH, as Insulet shared last month that Omnipod 5 will have pricing parity with DASH, at least initially.

    • Moving ahead – which CGM system will have the ability to measure ketones? This is hugely important for those with type 1 who want to take SGLT-2s but can’t due to risk that is due high for developing euglycemic DKA. This is a small but growing number and all bets are off if kidney protection or heart failure reduction is proven.

  • Finally, the digital ecosystem supporting these AID systems remain an area ripe for potential and even ongoing differentiation.

    • Insulet and Tandem have both made significant investments in smartphone control for their AID systems, with Insulet planning to launch with full Android smartphone control compatibility and Tandem expecting to launch smartphone bolusing for both Android and iOS users in 1H21. I

    • n recent updates,  both companies have also hinted at meaningful software improvements to support patients and providers, with Tandem recently announcing its second-generation provider data management platform, Tandem Source, and Insulet hinting at new data-analytics and review software at JPM 2021.  

    • Medtronic, of course, has advantages within it is own ecosystem, particularly its own Carelink – it also historically has had many arrangements with various payer groups and HCPs. That’s certainly becoming the case now for Tandem and we see lots of these opportunities for Insulet, particularly on the type 2 front. Where Abbott chooses to play, when it can, will be incredibly important, as it will enable much lower cost.

2. “This data is incredible”: Dr. Bruce Bode on Omnipod 5’s ease of use and adaptivity and its potential for MDI and type 2 patients

During Saturday’s investor call, we heard from Dr. Bruce Bode (Atlanta Diabetes Associates), one of the pivotal’s Principal Investigators. Overall, Dr. Bode summarized by saying, “This data is incredible” and the system is effective for supporting patients with diabetes and reducing burden because of its ease of use and algorithm adaptability, which makes it particularly well-suited to type 2s and patients coming from MDI therapy. As a reminder, Dr. Bode presented partial – but very exciting – data from a type 2 feasibility study on Saturday at ENDO 2021.

  • Reflecting on the results, Dr. Bode focused on the “unheard-of” pediatric results and the reduction in hypoglycemia. Dr. Bode focused in on Omnipod 5’s reduction of hypoglycemia (-13 minutes/day <70 mg/dl from 2% to 1.1%), which was the greatest hypoglycemia improve “so far I’ve seen.” Dr. Bode highlighted Omnipod 5’s combination of reducing hypoglycemia and achieving 70% Time in Range, demonstrating that Omnipod 5, and AID systems overall, will “be the standard of care in the future for type 1s and eventually, type 2s who need insulin.”

  • Dr. Bode also emphasized Omnipod 5’s ease of use. He argued that for patients, the reduced number of alerts and the patch pump model (i.e., no tubes, just fill with insulin, attach, and go) reduces technology fatigue and burden and in particular, making AID system use easier for pediatric patients and older patients. Dr. Bode also noted that the flexibility of pod location (e.g., back of arm, stomach, back) allows different patients to choose the position best suited to their needs. He argued that together, these features make the Omnipod 5 system “so simple” compared to Medtronic’s MiniMed 780G, which requires calibration and cannula attachment, and Tandem’s Control-IQ, which “tears up” insulin (this is a complaint that we have heard when t:slim X2 is used with Novo’s Fiasp insulin, off-label, though we need to investigate this more thoroughly or call dQ&A in for a study on this as it is at this point anecdotal), takes time to fill, must be recharged, and requires cannula attachment and tubes. For providers, the system is easier because are fewer settings and components to support patients in setting. In his experience in the trial, he didn’t have to help people other than adjusting carb ratio and encouraging them to be more aggressive, although we’d note that these were generally patients with tight glycemic control and some level of technology experience at baseline (e.g., only 20% MDI), which means that in the real world, providers may have to provide additional support to patients.

  • The algorithm’s adaptivity is another advantage of the Omnipod 5 system, per Dr. Bode. He argued that the algorithm’s constant micro-bolus adaptation is one of the most important and meaningful features of the system, of which he is in awe: “I have no idea how you did that. No one has been able to do that.” As a reminder, the system adapts its micro-bolus dosing in real-time, but also adjusts the factors driving its algorithm to more accurately predict the micro-dose needed to keep a user at their target. Based on our understanding of the system, the algorithm uses retrospective insulin dosing data, Time in Range, and hypoglycemia data from the last ~week (two pod wears/6-7 days) to modulate its algorithm factors moving forward, improving the system’s accuracy and ability to keep patients in range.

  • As Dr. Ly put it during the session, Omnipod 5 is intended to be “a solution for the masses,” including MDI patients and type 2s. During his commentary, Dr. Bode argued that MDI patients did “just as well” as technology-experienced patients and that they often found it easier to use to manage their diabetes than MDI, even though there was no run-in period in the trial. In particular, he drew attention to the ability to deliver boluses via a smartphone app or locked Android device without using a pen or syringe (as is the case with MDI) or having to go to the pump itself (as is the case for other AID systems). He also noted that unlike with MDI, patients using Omnipod 5 are only using a rapid-acting insulin, which should reduce the insulin-related costs compared to MDI therapy, although, of course, the Omnipod 5 system itself will add additional costs. While the anecdotal data for MDI patients that we saw on Saturday and later during Thursday’s presentation (see below) seemed highly positive, we would have loved to see the results for MDI participants in particular, as well as to have seen the distribution of their A1c and CGM metric improvements.

  • Dr. Bode also discussed the potential of Omnipod 5 in type 2s and made similar arguments as he did for patients coming from MDI therapy. Drawing upon the early feasibility data he had presented earlier in the day at ENDO 2021, Dr. Bode argued that the early results are highly promising and that the system could be an incredibly meaningful improvement for patients both on MDI and on basal therapy. To illustrate what he hopes the system can do for a wide range of patients with diabetes, he discussed the case of one of the four type 2 patients who has completed the feasibility study, a middle-aged woman who worked night shifts. At baseline, she had an A1c of 10.1%, and by the end of the eight-week trial, she had seen a 2.8% reduction to 7.3% and was able to receive FDA approval to stay on Omnipod 5.

3. Case studies highlight Omnipod 5’s flexibility and adaptation for a wide variety of patients, especially children and adolescents

Insulet’s third Omnipod 5 webinar recapped results from the system’s pivotal trials in adults, adolescents, and children with presentations from Drs. Anders Carlson (International Diabetes Center) and Greg Forlenza (Barbara Davis Diabetes Center). The webinar also included four detailed case-study presentations from Dr. Carlson and Dr. Jennifer Sherr (Yale) demonstrating the applicability and effectiveness of Omnipod 5 across a wide variety of patient scenarios. Across all presentations, the most common theme was ease of use for patients and simplified diabetes management reducing the burden of diabetes for patients and caregivers alike.

  • Dr. Sherr presented data from a 22-year-old male medical student enrolled in the Ominpod 5 pivotal trial. Prior to enrolling in the study, Dr. Sherr had worked with this patient and though he was relatively well controlled at baseline, his diabetes management required significant effort. When this patient learned about the Omnipod 5 pivotal trial, their main goal was finding a way to reduce the burden of diabetes without compromising outcomes. At baseline, this patient had a diabetes duration of 18 years and an A1c of 6.5% with 76% TIR. After three months on Omnipod 5, Dr. Sherr’s patient had decreased their A1c by 0.4% to 6.1%, increased their TIR 8% (+1.9 hours/day) to 84% and decreased their time <70 mg/dl from 6.3% to 3.4% (-42 minutes/day). Additionally, these results were achieved and maintained over the winter holidays (Thanksgiving and Christmas) as well as during the patient’s medical school rotations. Wow! Given that this patient began with good glycemic control at baseline, it is impressive that they were still able to improve their outcomes while on the Omnipod 5 system, particularly during a difficult time. According to comments from Dr. Sherr, this patient was also able to achieve their goal of reduced diabetes burden by “handing over the burden of diabetes care to the [Omnipod 5] system.”

  • Dr. Carlson presented a case study from an 18-year-old male in his senior year of high school enrolled in the Omnipod 5 pivotal trial. This patient was on MDI prior to enrolling in the trial and only treated himself with insulin at mealtimes. At baseline this patient, who was diagnosed with type 1 at 13, had an A1c of 7.6% but was spending only 55% TIR. Prior to enrolling in the study, this patient didn’t engage with his diabetes outside of mealtimes and had a goal of improved TIR without having to substantially alter his life. After three months on the Omnipod 5 system, this patient decreased their A1c by 0.6% to 7%, meeting ADA guidelines and increased his TIR 8% to 63% giving him an additional 1.8 hours/day in range. According to Dr. Carlson this patient characterized the Omnipod 5 system as giving him a “new understanding of how to manage diabetes” that he didn’t have while on MDI therapy. Dr. Carlson also expressed that this participant was able to continue with his life much the same as prior to enrolling in the study and even worked as a pizza delivery person eating pizza for dinner every night over the summer while still maintaining his improved glycemic control.

  • Dr. Carlson also presented data from a 24-year-old nursing student with a fear of hypoglycemia enrolled in the Omnipod 5 pivotal study. This patient was extremely concerned about hypoglycemia when bolusing and exercise, leading her to spend a significant portion of her time above range. At baseline, this patient, who had a diabetes duration of 13 years, had an A1c of 6.6%, but spent only 52% TIR. After three months on the Omnipod 5 system, this patient experienced a 0.2% drop in A1c to 6.4% and a remarkable 26% increase in Time in Range to 78%, the equivalent of an extra 6.3 hours/day in range. This patient, who experienced fear of hypoglycemia, was also able to reduce her time <70 mg/dl on the Omnipod 5 system to only 1.6%, compared to 3% at baseline – a reduction equal to 22 minutes/day.

  • Dr. Sherr presented Insulet’s final case study of a four-year old boy enrolled in the Omnipod 5 pivotal trial. This patient’s care was complicated by Trisomy 21 as well as developmental delays, hypothyroidism, and celiac disease. Prior to enrolling in the trial, this patient used a tubed pump and sensor with predictive low-glucose suspend, but the young patient had recently become aware of his infusion sets and was “ripping them out.” The patient and his family were also facing challenges as his activities such as speech, occupational, and physical therapy were often interrupted by his diabetes management. At baseline, this patient, who had a diabetes duration of 3.5 years, had an A1c of 9.2% and baseline TIR of only 28% with only 35% TIR overnight. After three months on the Omnipod 5 system, this patient and his family saw massive improvements with an A1c decrease of 2.4% to 6.8% and a 30% increase in TIR to 64%. Notably, the patient also spent 78% TIR overnight and was able to decrease the amount of insulin he took every day by 25%. According to Dr. Sherr, the patients mother called the Omnipod 5 system a “gamechanger” expressing she never thought A1c and TIR values like what her son had achieved on Omnipod 5 would be possible for him. Based on the pivotal read-out of Omnipod 5 as well as the existing base of Omnipod users, there is certainly great potential for improved outcomes among children with diabetes.

4. Investor Q&A: Insulet leadership argue that Omnipod 5’s features, form factor, business model, and clinical results will drive market share growth and overall market growth; system to fully launch “at the end of this year, early next year”

Saturday’s investor call featured thirty minutes of investor Q&A with CEO Shacey Petrovic, CMO Dr. Trang Ly, and SVP of Innovation and Strategy Eric Benjamin, who shared insights into the impact of Omnipod 5 on Insulet’s market position and discussed the advantages of Omnipod 5 compared to other AID systems and the next steps toward FDA approval. In particular, their discussion of the features and advantages of the Omnipod 5 system were fascinating, not only in terms of how Insulet positions itself in the AID market, but also in how they see those features and advantages driving overall AID and pump market growth looking ahead, as well as the movement into the insulin-requiring type 2 and MDI spaces.

  • Mr. Benjamin argued that Omnipod 5’s algorithm and user interface make it easy to use and offer advantages compared to other systems (i.e., Tandem’s Control-IQ, Medtronic’s MiniMed 770G and 780G), arguing that “the entire package is one of tremendous simplicity.” On the algorithm, Mr. Benjamin argued that Omnipod 5 offers the three differentiators compared to other AID systems: (i) it offers customizable “on target” technology where the system is constantly adjusting its delivery to hit the user’s personalized target between 110 mg/dl and 150 mg/dl; (ii) very little of the automated insulin delivery process requires the traditional pump settings that a user must enter; and (iii) the clinical outcomes (i.e., Time in Range, A1c, and hypoglycemia) are “fabulous.” We’d argue that it is particularly #1 (customizable targets) and #2 (the limited user engagement) that sounds particularly unique for some users compared to currently available systems, although MiniMed 780G, which is also under FDA review also has customizable targets (120 mg/dl and 100 mg/dl), albeit fewer and without the ability to set different targets for different times of day, as is the case with Omnipod 5. Mr. Benjamin also argued that the pod and the user interface offers additional advantages to users, specifically in terms of its: (i) fully wearable experience, (ii) SmartBolus calculator that takes into account both CGM value and trend, and (iii) HypoProtect feature. We’d add smartphone control to this list, especially when iOS phone users will also have the opportunity to control their system from their smartphone (more on this below). Moreover, in general, Omnipod 5 seems easy to train on, particularly via telehealth – and all the systems are getting easier to train on, which is great to see. The absence of required sets, of course, will be a big difference for some, while some won’t care; others don’t want to wear something “all the time” bigger than the set adhesive long associated with traditional pumps like Medtronic and Tandem (and Beta Bionics in the future). We look forward to learning more about costs and what can be manufactured at scale; payers, of course, will appreciate, as will many patients, less up-front cost, although by the end of a three or four year cycle, some Omnipod users or payers may pay more over time; contracts are presumably always changing, as are co-pays. It will be interesting, as well, of course, to see the “adherence” to each system over time; Insulet’s has the additional advantage for patients that they can change it at any time.

  • Management argued that Omnipod 5 will further strengthen Insulet’s position in the insulin delivery space and will grow the overall pump market, contributing to doubling market penetration in the “next few years.” As a reminder, Insulet’s market share has remained relatively consistent in the last few quarters around 30% (29% in 4Q20). Per Ms. Petrovic, Omnipod 5 should help drive market share gains for Insulet, particularly because Insulet “designed the system for the masses.” Ms. Petrovic explained her reasoning further, sharing, “Everything about the system is going to be really appealing to all users. When you combine that with our business model (that we are offering Omnipod 5 through the pharmacy with no up-front cost and no lock-in period), anybody can try the system” regardless of if a patient is on MDI or a tubed pump already. Ms. Petrovic argued that Omnipod 5 will also help expand the overall AID and pump market because of the system’s efficacy and simplicity (design choices + clinical results), as well as its business model in terms of affordability (price parity with Omnipod DASH, no up-front costs) and accessibility (pharmacy channel), all of which makes her “even more confident” that the US type 1 pump market penetration will double in the “next few years.” Additionally, management say these features as important in Insulet’s ability to penetrate the type 2 market, which the company has set their targets on after sharing some very impressive, albeit limited (n=4), early type 2 feasibility data at ENDO 2021.

  • Insulet aims to launch Omnipod 5 in a limited launch in 1H21, followed by a full launch “at the end of the year, beginning of next year.” Based on the call’s language, it seems that management is gearing more toward a full launch in the beginning of 2022 (in fact, Mr. Benjamin said, that the full market release would be “next year” specifically). In the six+ months between limited and full launches, we’d anticipate the company will focus on ramping up manufacturing capacity and ensuring widespread coverage, based on management commentary. During Saturday’s call, Mr. Benjamin mentioned that Insulet is “getting ready to produce commercial supply,” suggesting that the manufacturing capacity and/or available supply is not yet high enough to support a full market launch. He also noted that Insulet intends to spend the 2H21 expanding Omnipod 5 coverage. As a reminder, Insulet has already been submitted to the FDA, and management has expressed confidence that pandemic-related factors will not delay its clearance timeline.

  • Several investor questions asked why Insulet is launching with smartphone control for Android phones but not iOS phones and when iOS smartphone control can be expected. Mr. Benjamin shared that their decision to launch with smartphone control for Android phones but not iOS phones was solely due to the timeline of their engineering process. Per Mr. Benjamin, Insulet elected to write separate software for iOS and Android phones because the company wanted smartphone control to be simple, wanted to ensure cybersecurity, and wanted to ensure long-term maintainability. Because Omnipod DASH uses a locked Android phone as a controller, their development of the software for Android app control was further along when they began developing the programming for iOS phones. Regardless, Mr. Benjamin shared that iOS smartphone control is “well underway” and should be available “shortly after [the Omnipod 5] launch” although he did not specify if he meant the limited or full launch, which would set iOS smartphone control on two very different timelines. 

 

--by Hanna Gutow, Katie Mahoney, Albert Cai, and Kelly Close