Larry Smarr: the inflammation was in the numbers first

The situation

Larry Smarr is a computer scientist. He trained as an astrophysicist and ran a national supercomputing center in Illinois. He then became the founding director of Calit2, the California Institute for Telecommunications and Information Technology, at UC San Diego¹². He did not begin measuring his body because he felt ill. After moving to San Diego around 2000 he set out to get fit. He approached the project the way he would approach any complex system, by instrumenting it and watching the data over time³. He has described the process as pulling a loose thread on a sweater³. He lost weight, going from 205 to 184 pounds. He added step counting, sleep tracking, heart-rate monitoring, and consumer genotyping through 23andMe to a growing set of blood and stool panels¹.

Over the following decade the panels grew to roughly 100 biomarkers¹⁴. He paid for most of the testing himself. In 2012 he estimated that the out-of-pocket cost ran between $5,000 and $10,000 a year¹. That same year he published a paper in Biotechnology Journal describing the decade of blood and stool tracking. He interpreted it alongside his own genome and the genomes of the microbes in his gut. He framed the effort as one person’s attempt at the predictive, personalized medicine that Leroy Hood had been advocating⁵.

What he measured, and what it showed

The first marker to stand out was C-reactive protein, or CRP. The liver releases it into the blood in response to inflammation anywhere in the body. In November 2007 his CRP was 6.1 mg/L. Over the following seven months it climbed to 11.8¹. He has described those readings as five and then fifteen times the upper limit of healthy³. CRP tells you that something is inflamed. It does not tell you what or where.

The location came from a stool panel. Lactoferrin is a protein released by neutrophils, a type of white blood cell. A high lactoferrin count in stool therefore points to inflammation in the intestinal wall specifically. His lactoferrin rose several times to 200 against a normal count below 7.3. In the same period his stool tests showed low levels of beneficial bacteria and swings in lysozyme, another antibacterial protein. He read that pattern as his immune system fighting an episodic war with his gut microbes¹⁶. A 23andMe result added a variant associated with elevated Crohn’s disease risk⁶³.

Within a few months of the CRP rise, a sharp and persistent pain in the left side of his abdomen sent him to a doctor. The diagnosis was acute diverticulitis. His CRP hit 14.5 during the attack and settled at 4.9 afterward, still well above normal¹. He did not accept diverticulitis as the whole story. He overlaid his lactoferrin and CRP curves and went into the medical literature. There he found studies linking high lactoferrin to inflammatory bowel disease, the family of conditions that includes Crohn’s disease¹.

He then learned that UC San Diego had recently hired a new head of gastroenterology, William Sandborn. Sandborn had published on lactoferrin rising during flares of inflammatory bowel disease¹. By the time they met, Smarr’s lactoferrin had reached 900. A colonoscopy in December 2010 showed extensive diverticulitis. After reviewing the results Sandborn concluded that his new patient might have Crohn’s disease¹. By mid-2011 a stool analysis put his lactoferrin at 125 times the upper limit³⁷. He was eventually diagnosed with colonic Crohn’s disease⁷².

Sequencing the microbiome

Smarr did not stop at the panels. He had his gut microbiome sequenced at the J. Craig Venter Institute¹. He also began shotgun metagenomic sequencing of his stool samples. That method reads all the DNA in a sample rather than a single marker gene, so individual bacterial species and strains can be tracked over time⁶. He kept collecting. By early 2017 he had run about 90 stool tests at $375 each³. He has said that taking, freezing, and logging stool samples is the most time-consuming part of a routine that takes him about half an hour a day⁴.

A 2015 commentary in the journal Microbiome described his self-experimentation as the first published study to look at the relationship between the gut microbiome and Crohn’s disease⁸. In 2018 he was a corresponding author on a paper in Frontiers in Microbiology. It analyzed 27 stool samples collected over three years, from 2011 to 2014, from a single patient with Crohn’s disease confined to the sigmoid colon. Escherichia coli made up between 0.1 and 42.6 percent of that patient’s gut community, as much as 400 times the level in healthy controls. One strain recovered at peak inflammation resembled the adherent-invasive E. coli strains associated with Crohn’s disease. Over the same period the patient’s high-sensitivity CRP peaked at 27.1 mg/L. Fecal calprotectin is another neutrophil-derived inflammation marker. It peaked at 2,500, more than 50 times the healthy upper limit⁹. The value of that dataset comes from its density. A single microbiome sample says what is there. A series matched to inflammation markers says which organisms rise and fall with disease activity.

What he did with it

By late 2016 the diseased segment of his sigmoid colon had nearly closed². Its walls had thickened to seven times normal, and the internal channel had narrowed from 40 mm to 4 mm⁷. Surgery was scheduled with Sonia Ramamoorthy, chief of colon and rectal surgery at UC San Diego Health¹⁰.

Smarr decided to treat the operation as a pilot for what he called quantified surgery⁷. He already had abdominal MRI scans. Jurgen Schulze, a virtual-reality researcher at Calit2, stacked 96 MRI slices into a 3D volume. He segmented each organ by hand and color-coded the result, so the colon, spine, vessels, bladder, and spleen could be told apart³⁷. The model was shown in Calit2’s StarCAVE, an immersive virtual-reality room. A section of the diseased colon was 3D printed at life size⁷².

A few days before surgery Ramamoorthy came to look at it². In the model she could see that the sigmoid colon kinked at both ends and lay against the bladder. She could also see that the colon turned near the spleen at a point where it might be adhered to it⁷. She has described the inflamed segment pressing on the bladder as a no-fly zone, and said the intestine sat higher than she expected³. That let her place the ports for the surgical robot where they needed to be from the start. Without that knowledge, she said, she would have had to abandon the robot or cut new access holes mid-procedure. Either choice would have added 30 to 40 minutes². During the operation on November 29, 2016, Schulze piped the model into the da Vinci console. She could switch between the virtual colon and the real one⁷¹⁰. She estimated the planning saved at least 30 minutes of a five-hour surgery, with another 45 minutes saved by positioning the robot correctly. Blood loss was about 15 milliliters³. The eight-inch section was removed⁷.

The measurements continued afterward. He set himself a goal of 10,000 steps a day within two weeks of surgery and hit it⁴. For the first time in 12 years his CRP dropped below 1, and his lactoferrin returned to the normal range⁷.

What others can take from it

Smarr himself has said that he does not recommend other people do what he does, and that he considers his regimen fairly extreme⁴. The treatment decisions were his and his doctors’. What transfers is how he handled the data.

A marker that is out of range for years is a finding, even without symptoms. His CRP was elevated for a long stretch before the abdominal pain arrived. The trend line carried more information than any single reading¹.

Pairing a general marker with a specific one narrowed the question. CRP said inflammation existed. Lactoferrin said it was in the gut. Neither alone would have pointed him toward inflammatory bowel disease. The overlay of the two curves is what sent him into the literature¹.

He also found the clinician who worked on his marker. He did not bring a self-diagnosis to a random doctor. He found the gastroenterologist who had published on lactoferrin in inflammatory bowel disease and brought him a time series that spoke his language¹.

Sampling through the whole course of the illness is what made the data useful to others. He collected stool samples before, during, and after diagnosis and treatment. That let his data support a strain-level analysis of how one organism tracked inflammation over three years. After surgery he could confirm that the inflammation markers had normalized⁹⁷.

Imaging counts as data as well. An MRI that gets read once by a radiologist can be reconstructed into a model that a surgeon studies before making the first cut. Ramamoorthy’s account is that what she learned changed the plan, not just her confidence³².

Footnotes

1. Jon Cohen. The Patient of the Future. MIT Technology Review, 2012. https://www.technologyreview.com/2012/02/21/187511/the-patient-of-the-future/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴

2. Clare Scott. Crohn’s Disease Patient Invites the Public to Look Inside His Virtual Colon, Shows 3D Model to Surgeon Before Procedure. 3DPrint.com, 2017. https://3dprint.com/167568/crohns-disease-3d-colon-model/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

3. Usha Lee McFarling. Welcome to my colon: A tech pioneer turns to virtual reality to guide his own surgery. STAT, 2017. https://www.statnews.com/2017/03/09/colon-virtual-reality/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰

4. S. Campbell. The Quantified Patient Checks In: Larry Smarr’s Experiments in Self-Tracking for Health. IEEE Pulse, 2017. https://www.embs.org/pulse/articles/quantified-patient-checks/ ↩ ↩² ↩³ ↩⁴

5. Larry Smarr. Quantifying your body: a how-to guide from a systems biology perspective. Biotechnology Journal, 2012. https://doi.org/10.1002/biot.201100495 ↩

6. Ernesto Ramirez. Larry Smarr on Crohn’s Disease and Quantified Self. Quantified Self, 2013. https://quantifiedself.com/blog/larry_smarr_croneshope_in_data/ ↩ ↩² ↩³

7. UC San Diego Computer Science and Engineering. CSE Professor, Lecturer Team with UC San Diego Health to Bring 3D Visualization to Abdominal Surgery. UC San Diego CSE News, 2017. https://cse.ucsd.edu/about/news/cse-professor-lecturer-team-uc-san-diego-health-bring-3d-visualization-abdomenal-surgery ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹

8. Carine Gimbert and François-Joseph Lapointe. Self-tracking the microbiome: where do we go from here? Microbiome, 2015. https://pmc.ncbi.nlm.nih.gov/articles/PMC4676868/ ↩

9. Xin Fang, Jonathan M. Monk, Sergey Nurk, et al. Metagenomics-Based, Strain-Level Analysis of Escherichia coli From a Time-Series of Microbiome Samples From a Crohn’s Disease Patient. Frontiers in Microbiology, 2018. https://www.frontiersin.org/articles/10.3389/fmicb.2018.02559/full ↩ ↩²

10. Scott LaFee. Visualizing the Future of Surgery. UC San Diego Today, 2017. https://today.ucsd.edu/story/visualizing_the_future_of_surgery ↩ ↩²