Theranos blood test's Specificity and Accuracy?

Question

I am studying the limitations of Theranos device. They have 238 documents in Google Patents Search (Inassignee as Theranos), here. Which patents are relevant for their device? They have 70 tests which are not necessary to consider now. Just the basics of the device.

My intuition says that it uses PCR to multiply the microliter of blood for the analysis. What are limiations of multiplying so little blood? I think this is where their engineering work applies. What components do you need to multiply? I think all is not necessary. Otherwise, the specificity and accuracy can be difficult to hold.

How can Theranos blood test's be specific and accurate compared to conventional full blood test? I suspect if you can do the conventional full blood test with Theranos device with the same accuracy and specificity as the conventional one.

Answer 1

Which patents are relevant for their device?

• They currently only have about 27 issued patents (US)
• Of those patents about 10 of them are related to network/internet/health record/algorithm (so not really about their device)
• The remaining patents focus on a few areas: sample acquisition (getting the blood drop and handling of it), Patch based technology (before they pivoted), sample quality analysis (this is key to making sure the small sample is good quality), microfluidics/modular device (these are the patents relating to the actual testing and device)

What are the limitations of multiplying so little blood?

• From a molecular standpoint (PCR) the smaller sample you start with, the less of the original nucleic acid template you have to start with. Lower end PCR testing uses 10uL of sample. This is ok when you are trying to bring molecular diagnostics to the point of care or developing markets, but the best PCR tests use 50uL of sample (more than a drop of blood which is traditionally 25uL or so), regardless of how good your device is, at some point you start amplifying junk. The better the primers and other reagents the less junk you amplify, but at the end of the day you want more sample to be more sensitive.

• From a NON-molecular standpoint (chemistries, cellular analysis, coag, etc.), they aren't/can't multiply the analyte of interest. In these tests, they rely more on dilution or small volumes to perform detection. This is where their ability to reliably dilute and correct for the dilution comes into play. Although there is certain limits to this. Coagulation tests would not be representative if you diluted the sample so in this case they will have to rely on working with a small volume. Cellular analysis can only be diluted so much as you get to a point where you miss important but less common cell types. Take eosinophils (a type of WBC) while they can get to 15 or 20 percent of all WBC cell types, in general they only make up a couple percent. Assume that 2% of all your WBCs are eosinophils. If you only have 3500 WBCs per 1uL of blood (which is on the lower end of normal), you have approximately only 70 of these cells in that same volume. If you dilute the sample so that way there is more of it available for other tests like electrolytes, but you still analyze 1uL, The number of those cells is even less. If you had perfect detection you might be ok, but you generally want to analyze more blood so that statistics work in your favor and your performance goes up. Most of the other chemistries, proteins, and as mentioned earlier, electrolytes will probably be ok for dilution and detection and in fact this is why there are plenty of small devices already out there that use a drop of blood and perform the same tests.

• Their issue in diluting or amplifying analytes of interest is that biology still is governed by the laws of physics and even though technology may allow you to analyze samples down to 10nL, that just isn't use full in biology, and limits their their capability to perform the supposed menu of tests from one microtainer of blood. They don't publicize that they may need to draw a venous sample (needle in your arm) for some tests and depending on what tests you order, you may need a separate finger prick for each test.

How can Theranos blood test's be specific and accurate compared to conventional full blood test?

• If they reliably control dilution of the small sample, they maybe able to provide high quality results for some tests.

• Finger pricks actually can be easily skewed via technique. Theranos has specially trained people to draw these samples, but it is time consuming >5 mins and until it is fully automated this is one variable that no matter how good sensors are, you can't correct for variation pre-analytically when you have that small sample. Finger pricks also skew certain tests even when technique is good such as platelet and WBC values in the CBC.

• I discussed the sample volume issue above regarding PCR amplification and how good tests use up to 50uL. So if they compare to a 10uL test, they may be able to compete, but not full lab quality.

• A lot of the future techniques they have regarding small techniques (in their applications) like their coagulation tests work great in an academic setting, but they fail to be able to handle the full range of blood samples they will see, so it will be interesting to see how that plays out.

Takeaway

• Theranos is working on modular micofluidic devices/cartridges that will use small sample
• There are certain biological limitations that stand in the way of them performing all the tests in their catalog with only a drop of blood
• The FDA will eventually be requiring them to submit their tests for approval as they are currently doing tests that are available on commercially and FDA approved devices already (this makes sense because LDTs were never ment to be used in place of standard tests and equipment and since there is NO published data on their performance it can be worrisome to providers and payer)
• Their true business plan has yet to be elucidated, but if they really did have a box that did all this, they would be selling them to physicians and clincs where those results could actually be used in diagnosis rather than 4 hours later.
• I hope they succeed, but there is just way too much hype to buy in on it yet.

I am sorry I don't have sources, most of this is from intimate knowledge of the space and from the little info available on them so I could definitely be wrong on some of this. I hope this helps!

Answer 2

I haven't seen their device and I won't want to see it, I want to first understand the concept better independently. I am not an expert in biology but I have feeling how such device could be done.

Noname covers sensitivity and new concept "amplification" raising a new puzzle: is amplification done by analog methods (op-apms), digital methods, dilution or/and some other methods?

---

Summarising the Noname to sensitivity, amplification and background

---

(I added bolding)

1. "Theranos is working on modular micofluidic devices/cartridges"

2. "regardless of how good your device is, at some point you start amplifying junk. The better the primers and other reagents the less junk you amplify, but at the end of the day you want more sample to be more sensitive."

3. "they aren't/can't multiply the analyte of interest. In these tests, they rely more on dilution or small volumes to perform detection. This is where their ability to reliably dilute and correct for the dilution comes into play. -- if you had perfect detection you might be ok, but you generally want to analyze more blood so that statistics work in your favor and your performance goes up. Most of the other chemistries, proteins, and as mentioned earlier, electrolytes will probably be ok for dilution and detection and in fact this is why there are plenty of small devices already out there that use a drop of blood and perform the same tests."

4. "They don't publicize that they may need to draw a venous sample (needle in your arm) for some tests and depending on what tests you order, you may need a separate finger prick for each test."

where by (2) more sensitivity can actually be a good thing; by (3) same tests already done by other devices in small size: what is so special with Theranos device? and by (4) one drop of blood not enough for all tests.

"What are limiations of multiplying so little blood?"

---

1. Decision Analysis (DA) and Compositional Blood Analysis (CBA): I see such a device as a Decision Analysis (DA) device where you need to do compositional blood analysis (CBA). You get a lot of vague information on which you do sensitivity analysis. Conventional full blood tests can be seen as devices where the sensitivity is very very low to other factors. It is not surprising that development has lasted about 10 years, sensitivity analysis is very low research project. For example, Chevron's DA model started 1990 and continuing, implementation takes time. For example, scale hydration level to some level in all samples. When you have low-quality information, you need to use ad hoc models (often based on high-quality samples) such as supposed correlations between things to improve prediction accuracy.

---

1.1. What are Decision factors?

---

1.1.1. Rheology of blood? Dry or liquid phase samples?

1.1.2. Correlation of red/white shells: I don't use the word "approximation" because the number of red shells correlated with the number of white shells

1.1.3. Density (~1.05 kg/l)

1.1.4. Colour of blood

1.1.5. Hydration level

1.1.6. other factors?

---

"How can Theranos blood test's be specific and accurate compared to conventional full blood test?"

---

I think this question mis-framed. It is better to ask "How much extra value does such a device can offer?"

Smaller size and less blood mean convenience. It may be useful with athletes, new markets. More convenient to patients/users even though suffering from some abnormalities due to its intrinsic limitations related to its small-volume measurement method.

---

What components do you need to multiply?

---

I think this is an important insight: is it possible to multiply blood somehow ensuring its quality for testing? This would make blood tests less painful for patients. Perhaps this is related to stem-cell research? Perhaps not, the Noname covers this in terms of dilution more closely.