Machine learning-based diagnostics startup Dascena Inc. has won the U.S. FDA’s breakthrough device designation for its Previse algorithm, which is designed to predict acute kidney injury (AKI) before clinical symptoms. In early validation tests, Previse detected AKI more than a day before patients exhibited kidney damage or impaired function.
The breakthrough designation, which provides companies with more FDA input and priority review, is the first for a cloud-based machine learning algorithm for the early detection of AKI, the company said.
AKI occurs when the kidneys can’t filter waste from the blood, causing dangerous amounts of waste to build up in a person’s circulatory system. Diabetes, high blood pressure and advanced age are major risk factors for AKI, but it also is increasingly seen in hospitalized patients as a complication of other illnesses or interventions. According to CDC data, the number of hospitalizations with AKI increased from 953,926 in 2000 to 1.8 million in 2006 and roughly 4 million in 2014. In each of those years, diabetes was a comorbidity in more than a third of AKI cases.
As with any serious chronic condition, early detection and treatment of AKI is key to improving patient outcomes. However, the clinical criteria for diagnosing the problem – serum creatinine and urine output – often lead to delays in treatment. Currently, urine flow rate and volume are monitored intermittently and manually by ICU nurses, and acute changes in urine flow are hard to detect.
High level of accuracy two days before onset
Dascena’s machine learning algorithm continuously monitors inpatients to predict AKI more than a full day before patients meet the clinical standard for diagnosis. In a 2018 study published in the Canadian Journal of Kidney Health and Disease, Previse predicted AKI 48 hours prior to onset with 84% accuracy and a diagnostic odds ratio of 5.8.
“Acute kidney injury commonly afflicts hospitalized individuals, and if not caught early, can result in dangerous outcomes for patients, said Ritankar Das, CEO of the Oakland, Calif.-based company. “Our machine learning algorithm is able to analyze patient vital sign data and determine whether a patient is at risk of developing acute kidney injury. With this technology, we believe we’ll be able to provide physicians with ample time to intervene and prevent long-term kidney injury in their patients.”
Dascena leverages machine learning techniques and its vast patient database to build its algorithms. To predict AKI, Previse draws upon values of heart rate, respiratory rate, temperature, serum creatine, the Glasgow Coma Scale and age.
“AKI is evidenced by increases in serum creatinine (SCr) level, and studies have shown that even modest increases in SCr are significantly associated with poor outcomes,” Das told BioWorld. “One study reports that ‘an increase in SCr ≥0.5 mg/dl was associated with a 6.5-fold (95% confidence interval 5.0 to 8.5) increase in the odds of death, a 3.5-d increase in LOS, and nearly $7500 in excess hospitals costs.’ This suggests that earlier interventions (e.g. 24-48 hours in advance) may be critical and perhaps life-saving in AKI cases.”
The company will be pursuing the 510(k) regulatory pathway, targeting clearance of Previse in 2021, Das said. He added that the company will work with U.S. FDA staff to determine what validation data are needed to support the submission.
Algorithms for sepsis and decompensation
Previse is one of three algorithms Dascena is currently developing, with an eye toward informing patient care strategies and improving outcomes. The company also has a diagnostic algorithm for sepsis, called Insight. In a randomized, controlled trial of intensive care unit patients, Insight reduced deaths and hospital stays by 58% and 21%, respectively. The algorithm’s performance was validated in a larger study in ICU, other hospitalized and emergency room patients.
The third algorithm, Autotriage, is intended to predict decompensation, or the unplanned readmission of a patient to the ICU. According to Dascena, Autotriage has a 93% accuracy rate and specificity of 94.5%. Overall, the algorithms have been validated via 18 peer-reviewed publications of studies funded by the National Institutes of Health and National Science Foundation.
Founded in 2014 with the aim of creating smart diagnostics to catch life-threatening conditions before they become serious, Dascena recently raised $50 million in a series B round to advance its machine learning algorithms. The financing, which closed in April, was led by Frazier Healthcare Partners, with participation from Longitude Capital, existing investor Euclidean Capital and an unnamed investor.
Das said the money will be used “to grow the company, including expanding the company’s teams, contributing to more algorithm development and supporting FDA submission processes, including conducting validation studies.”
Other companies are targeting the AKI market. Earlier this year, Serenno Medical Ltd., of Yokneam, Israel, debuted Sentinel, a device that automatically and continuously monitors urine output to detect kidney damage in patients. The company hopes to get FDA clearance and begin selling the device in the U.S. before the end of the year, with regulatory approvals in other regions to follow.
And in May, Carlsbad, Calif.-based Hemaflo Therapeutics Inc. saw Venture.co facilitate the $7.5 million capital raise for a phase I trial of its drag-reducing polymers (DRPs) technology, which can be given intravenously to improve blood flow by reducing friction and stimulating blood vessels to dilate. The company is developing the DRG platform as a solution for AKI.