What is ground truth: home meter or lab venipuncture?

Valve Replacement Forums

Help Support Valve Replacement Forums:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
I've been somewhat curious about the experience of others using the InRatio method -- it would be good to know if their INRs also differ by as much as yours and mine do. (Of course, the posted 30% variance is rather troubling and not inconsistent with our experiences) I've also wondered if you may be taking something that makes your meter report higher than accurate scores (as I suspected the Quercetin + C may have done).

I've stopped the Quercetin + C and will retest in a few days to see if the reported INR drops. (My lab-reported INR is fairly stable, and I don't expect much change between yesterday, when the effects of the Quercetin + C were still evident, and when I retest after the Quercetin + C effects probably have worn off).
 
30% is kind of troubling--especially if the values are HIGH for high INRs and LOW for low INRs. I'm especially concerned about the lower values.

I wonder if the CoaguChek XS has a similar range.

Your question about the CoaguChek XS helped to jog my memory a little. I vaguely remembered an accuracy discussion that came up about the CoaguChek last year, actually before I was even a member here. Now, don't ask how I happen to remember this, I usually forget things my wife tells me within 5 seconds!

Anyway, I would guess it was that thread, and similar posts in a few others, that was subconsciously framing my reference point for what type of accuracy I assumed these home monitors actually have. I was under the impression that the vast majority (90% or more) of readings were within +/- 10%, so around a 0.3 difference or so. This type of accuracy of course is pretty comfortable, enough to leave a little wiggle room for inaccuracy yet still be in range. So, for reference, I did a quick search and found that thread: http://www.valvereplacement.org/forums/showthread.php?30821-How-do-you-like-your-Coaguchek-XS.

You'll note two interesting posts in that thread. One mentions that a call to Roche indicated that 30% variance was in fact acceptable. The discussion here recently confirms this - it is actually in the ISO standard. The other interesting post was about Coaguchek XS monitors guaranteeing a maximum difference of 0.3 from lab. So, obviously, some disagreement between the two points.

Well, I just spent a little time researching this further, and so far have not found any mention of the 0.3 guarantee on the official product literature for the CoaguChek XS. I did find a few "expected results" data sets as well as accuracy publications, and they seem to me to indicate much closer to 30% variance than 10%. Anyway, here is what I found:

This CoaguChek page (http://www.coaguchek.com/com/index.php?target=/en/professionals/further_information/faq/6) mentions possible deviation as follows: 0.1-0.3 for INR < 2.5 | 0.5-1.0 for INR 2.5 to 4.5 | 1.0 to 2.0 for INR > 4.5. It further says that in general deviations up to +/- 20% are not unusual.

This outside analysis of CoaguChek XS performance (http://www.coaguchek.com/resource.php?id=Resourcefile-6576494b8e3296faf&f=UkZJTEU2NTc2NDk0YjhlMzI5NjAxNi5wZGY=) that is referenced on the CoaguChek site (http://www.coaguchek.com/com/index.php?target=/en/professionals/further_information/correlation_studies_and_evaluations) indicates that 97% of values were within the ISO +/- 30% standard. More meaningfully, though, it breaks it down further into the following approximate (there were multiple sets and test strip lots, as well as both blood methods, so I'm just doing a quick eyeball average - see the actual publication for full details) capillary based percentages for INR values less than 4.5: 98% of values were within +/- 30% deviation | 95% of values were within +/- 20% deviation | 67% of values were within +/- 10% deviation.

So, it seems to me that maybe 20% deviation should be considered a reasonably reliable "working range of accuracy". But 10% certainly not. And based on the evidence Roche presents, a 0.3 maximum difference guarantee seems unlikely. I may be wrong, of course, so it would be interesting if any users here could confirm if that's true.

Here's another visual aid on accuracy results, this time for the CoaguChek XS. It's slightly different in format than the INRatio 2 plot earlier, so be careful with direct comparison:

Roche.png
 
It almost looks as if neither meter would be without worry if the reported 'acceptable' errors were reached. For example, if an INR of 2.5 on the meter was actually 1.0 HIGHER than a lab test, a person could believe he or she is IN RANGE when, in actuality, an INR of 1.5 puts him or her into a zone where dosing adjustment may be prudent.

I suspect that we've got to learn out meters -- that, through repeated validation with lab tests, we can get a sense of how well the meter actually correlates with the lab. (This is not to say that labs are ALWAYS right, either). Here's where a clinic may actually be of more value - if it has a long history of relating meter values to lab values - there may be a sense of what to expect from the meters.

I've lost confidence in my meter actually coming very close to the lab WHEN I TAKE QUERCETIN+C. I'll have a better sense of its accuracy when I test tomorrow, four days after discontinuing that supplement.

My anticoagulation clinic (which I've gone to exactly twice and really don't think I need) uses a Hemochron -- this meter is used in operating rooms to test different blood factors. It's also apparently pretty close to lab accuracy. It may be a bit more of a process to do a test on this meter than it does on an InRatio or CoaguChek XS, but it may deliver values that are closer to those that a lab determines.

It would be useful if 'problem' materials (like, perhaps, Quercetin+C) cause a meter to incorrectly report a much higher INR than the lab can be put into a list, so that self-testers and clinics can adjust for these difference or use lab values (or, perhaps, a Hemochron or one of the other meters that actually times clot formation), as a control to occasionally confirm results
 
Hello ElectLive,

Thanks for the posts about the accuracy.
I had also been investigating the Coagucheck-XS as a possible alternative based on what I discovered about the Inratio-2, but it seems that the Coagucheck is pretty much the same.

There is a paper available on the Roche website page:
https://www.poc.roche.com/poc/rewrite/generalContent/en_US/article/POC_general_article_119.htm#3

if you click on the "CoaguChek XS Precision White Paper" link in the "Case Studies/White Papers" section. It has the same graph format in that report as the one you posted for the Inratio-2 meter, which is also in the report I referenced earlier myself. (update - I figured out how to add the image to the post)

CoaguchekXS_Accuracy.gif




The thing I'm still struggling with on the meter accuracy issue is whether the number of relatively high error readings (up to 30% different from lab values) that are allowable are characteristic of different people in the general population, or if a given individual's readings will vary by that amount over time.

If it is different people who have relatively consistent offsets/bias, as PEM's results would seem to indicate, then it is a fairly straightforward process to collect enough Lab-Meter comparison tests to create our own individual correction formulas.

Of course, changing medications and supplements can create a new condition that may necessitate revising one's correction formula, but that's another problem that can be handled assuming each individual has consistent offsets/biases in the meter versus the lab test results.

If 90% of the general population consistently have correlation to lab tests better than 10% (i.e. +/- 0.3 in INR for a typical INR of 3.0) and it is only 10% of the population that has to screw around with finding their individual adjustment factors, then I can't really fault the meter manufacturers.

I don't have enough personal data points to make any conclusion regarding my own meter tests versus lab tests, but I do plan to gather as many more comparison test points as I can over the coming months.
 
The question of accuracy -- and whether some people may just have factors that result in larger variations from lab tests is an interesting one. InRatio (and probably CoaguChek XS) are sensitive to the number of red blood cells in the blood. Too many or too few could throw off the accuracy. It's possible that these meters would not fail quality control, even if some of the patient's blood characteristics (aside from INR) may be marginal. A meter that tests for actual clotting time - not relying on some secondary indicator (like changes in impedance, perhaps) might be less sensitive to these factors. (This makes my old ProTime 3 and a new meter that does a mechanical test for clotting more attractive). It's too bad that there is no meter that can check for confounding factors when taking the tests. It may be interesting for people who are consistently in the 30% difference from lab results group to have other blood tests taken. Who knows - perhaps that much variation between meter and lab may be indicative of something going on with their blood?
 
The thing I'm still struggling with on the meter accuracy issue is whether the number of relatively high error readings (up to 30% different from lab values) that are allowable are characteristic of different people in the general population, or if a given individual's readings will vary by that amount over time.

Excellent question! I have been wondering whether or not relevant historical data could be obtained from a clinic. For example, I've seen Swedish studies that base their findings on millions of data points. It wouldn't surprise me if they make those data publicly available, though I suspect that lab vs. meter comparisons would most often result from out-of-range values - so there could be a preponderance of comparative data for extreme INR values but perhaps not as much for in-range values. This is all speculative, but I intend to look into it a bit, time permitting.

Best,
pem
 
InRatio (and probably CoaguChek XS) are sensitive to the number of red blood cells in the blood. Too many or too few could throw off the accuracy.

Protimenow,

Do you have a reference for this from which I could get more detail?

Thanks,
pem
 
Pem:

Page 46 of the InRatio2 manual says 'A hematocrit (percentage of your blood that is red blood cells) that is higher or lower than the validated operating range of the monitoring system can cause an inaccurate result.'

I'm not sure if CoaguChekXS has the same limitation. Lab tests - and at least one meter - don't test the red blood cells - they test the fibrins in the plasma.

Also -- an update on my suspicions that Quercetin can do something that makes the InRatio (and InRatio 2) report artificially high INRs:

When I was taking Quercetin+C, both meters reported 3.8. An hour later, the lab came up with a 3.09. On May 8, I tested my blood on the InRatio2 and got a 3.8. The clinic's Hemochron got a 3.2.

I haven't taken Quercetin+C since 5/8. It's cleared from my system by now. My INR on the InRatio was 2.9, and on the InRatio2, it was 2.8.

As far as I can tell, the only thing that has changed in the interim is not taking Quercetin+C. I was actually a bit surprised that my INR seems to have actually DROPPED since my test 3 days ago.

I'm going to try to somehow either get cuvettes for my ProTime 3, which uses a different method for testing INR than do the InRatio or CoaguChek meters, or possibly a new meter that uses a mechanical device built into the strip to detect clotting, and use these as 'controls' for further testing. I really have STRONG doubts about the accuracy of my meter when certain substances are taken. If so, this information should be made available (and had I known a few weeks ago, I could have upped my dosage and perhaps prevented my stroke). (I'm okay now, but it should be made clear if there IS a risk of inaccurate results as large as the ones I seem to have encountered, in order to protect public health).
 
'A hematocrit (percentage of your blood that is red blood cells) that is higher or lower than the validated operating range of the monitoring system can cause an inaccurate result.'

I'm not sure if CoaguChekXS has the same limitation.

The CoaguChek XS does have a similar validated operating range warning for hematocrit level. According to the package insert: "Hematocrit ranges between 25-55 % do not significantly affect test results." This is almost identical to the INRatio2 which is 30% to 55%. For reference, the ISO standard actually only requires a validated range of 35% to 50%. Also, the analysis of CoaguChek XS performance I posted above included a brief description of the patient group and noted that the hemotocrit values of test samples covered a range of 26% to 54%.
 
Page 46 of the InRatio2 manual says 'A hematocrit (percentage of your blood that is red blood cells) that is higher or lower than the validated operating range of the monitoring system can cause an inaccurate result.'

Oh - right. Thanks, Protimenow. I forgot that I'd covered this ground and already verified that my own hematocrit is within this range (see earlier postings in this thread). For some reason I momentarily confused hematocrit with MCV. I wonder if either company (Alere or Roche) has identified a validated operating range for Mean Corpuscular Volume (MCV) - or perhaps it is irrelevant. Given that the meters seem to rely upon capillary action to draw the blood into the strips, I also wonder if the viscosity of the blood is a factor.

Thanks again,
pem
 
Latest meter vs lab comparison

Latest meter vs lab comparison

Last Thursday I made another meter vs lab vs lab comparison:
InRatio2: 3.1
Lab1 (ISO=1.8): 2.5
Lab2 (ISO=1.0): 2.3
Predicted Lab value (model2): 2.6

This is the first time that the lab results have been different from each other. But even in this case, not by much (0.2).

The average of the two lab values (2.4) is a bit lower than the value (2.6) predicted by model2 (please see prev posts).

So in this case, if I were to use my meter to generated a predicted lab value, I would believe myself to be in the therapeutic range. If I trusted Lab1, I would also be in range. If I trusted Lab2, I would be out of range.

I may collect a few more side-by-side samples. Each one builds (or erodes) statistical confidence in the model.

Thanks,
pem
 
A few questions seem to be in your last two posts:

First, I really have to wonder if you're doing something (you mentioned drinking Green Tea in a message) that may be 'fooling' the InRatio into reporting an artificially elevated INR. You shouldn't HAVE TO use a predictive model to correlate your meter's reported INR to the lab values. It makes me wonder if this amount of bias is true across ALL InRatio meters (and, if so, how could they have gotten FDA approval), or if you and I are taking something that causes the meter to report inaccurately high results. (I think I reported on my result last Friday, when Quercetin was out of my system --- my meters reported 2.8 and 2.9. (Four days earlier, while on Quercetin, the clinic's value was 3.2 and my meter was 3.7. I'd be inclined to think that it's POSSIBLE for my actual INR to have dropped in the three days since my clinic's test, and still remain confident that my assumption that Quercetin, and possibly other things (like, perhaps, your tea) may skew InRatio results up by .4-.6 or so). Labs aren't always perfect - but the difference you report is still somewhat troubling.

Second - I don't think that I've seen anything about MCV. I hope that you're not suggesting that Warfarin actually changes the viscosity of the blood. I suspect that, for the InRatio and CoaguChek XS meters that use capillary action to draw the blood into the strip, the only thing that matters is that enough blood actually gets drawn into the testing area of the strip. Since the meter isn't actually MOVING blood once it's in the strip, I doubt that 'viscosity' is an issue.

(FWIW -- the Protime actually DOES draw blood into the cuvette -- you put the blood into a small area on the strip, and it's 'sucked' into the strip. There may be another meter that operates the same way, actually moving the blood into the strip, mixing it with reagent, and monitoring for coagulation)
 
If it is different people who have relatively consistent offsets/bias, as PEM's results would seem to indicate, then it is a fairly straightforward process to collect enough Lab-Meter comparison tests to create our own individual correction formulas.

If some of the variance in the reported results is indeed due to individual differences, it would seem useful to isolate the variables that result in these differences. I would hope that Roche and Alere had access to (or recorded) other data about the samples they used. It would seem that they tracked hematocrit. I wonder what else they might know about the samples (other aspects of the blood profile) or donors (gender, age, etc) that could be used to ferret out sources of covariance. Too bad the data aren't in the public domain.

Best,
pem
 
There are a few questions that I keep returning to:

If your model applied to ALL tests, then wouldn't Hemosense (which developed the InRatio) have known about this and already programmed the adjustment into the meter? (This should have been a simple thing to do during manufacture). If so, they could have produced a meter with a very close match to Lab results.

I'm assuming that the InRatio results. over a large population, closely approximate lab results (and, for most, do much better than the +/- 30% that the paper you noted reports). I assumed, when I first started using my InRatio, and the results closely matched my ProTime meter - and later, that the results closely matched my CoaguSense S, that the results of all three meters were near matches to lab results. As with most patients (and I assume, clinics using the InRatio to assess INR), that the results of InRatio results closely matched the Lab results from, say, 1.0 - 4.0.

I also assumed that, as long as there were no quality control errors, the result was very close to lab results.

Here's where I have some questions:

Why did my INR climb when taking Quercetin - or did it, actually rise. The lab results, when compared to my InRatio results, were about .4-.8 DIFFERENT (the InRatio reporting higher results). When I discontinued Quercetin for a few days, my INR reported by the InRatio meter actually dropped slightly below the lab results of a few days earlier. This strongly suggests that there's something about Quercetin + C that I was taking that may have caused the InRatio meter to produce a value that was .4-.8 higher than actual lab value. (In my case, when I had an ischemic event, the meter showed a 2.6 -- what was my ACTUAL INR and how long was I walking around with 2.0 or substantially lower?)

Is your INR being reported higher than lab value because of something that you consume? (Green Tea?)

If it's true that some medications/foods can FOOL the InRatio (and perhaps other meters) into reporting INRs that are considerably higher than actual lab values, SHOULDN'T THE FDA, MANUFACTURERS, SELF-TESTERS, AND MEDICAL FACILITIES THAT USE THESE METERS AND RELY ON THEIR ACCURACY know about it. Shouldn't there perhaps be research into substances that can have this effect on meter accuracy and a registry be developed? Shouldn't some effort (aside from my own empirical testing) be put into determining if this is an actual effect; if it happens to others taking the same substances; if it's reproducible; and which meters are effected?

If I had the resources, I would repeat my Quercetin + C 'experiment' using other meters in addition to the InRatio and the lab to document the results.

ITC, ROCHE, Others - if you'd like to help by providing meters and strips, PM me and we can discuss how to get your device included in my testing.
 
There are a few questions that I keep returning to:

If your model applied to ALL tests, then wouldn't Hemosense (which developed the InRatio) have known about this and already programmed the adjustment into the meter? (This should have been a simple thing to do during manufacture). If so, they could have produced a meter with a very close match to Lab results.

Indeed, I am often surprised that when there is an opportunity to do something that makes sense from a given perspective, it is not always realized in the way one would expect.

I'm assuming that the InRatio results. over a large population, closely approximate lab results (and, for most, do much better than the +/- 30% that the paper you noted reports). I assumed, when I first started using my InRatio, and the results closely matched my ProTime meter - and later, that the results closely matched my CoaguSense S, that the results of all three meters were near matches to lab results. As with most patients (and I assume, clinics using the InRatio to assess INR), that the results of InRatio results closely matched the Lab results from, say, 1.0 - 4.0.

I also assumed that, as long as there were no quality control errors, the result was very close to lab results.

Given our recent findings, I would hesitate to make those assumptions.

Here's where I have some questions:

Why did my INR climb when taking Quercetin - or did it, actually rise. The lab results, when compared to my InRatio results, were about .4-.8 DIFFERENT (the InRatio reporting higher results). When I discontinued Quercetin for a few days, my INR reported by the InRatio meter actually dropped slightly below the lab results of a few days earlier.
This strongly suggests that there's something about Quercetin + C that I was taking that may have caused the InRatio meter to produce a value that was .4-.8 higher than actual lab value. (In my case, when I had an ischemic event, the meter showed a 2.6 -- what was my ACTUAL INR and how long was I walking around with 2.0 or substantially lower?)

If the lab result preceded or coincided with the offset of Quercetin, then your lab result would seem to reflect your INR in the presence of Quercetin, whereas your meter INR (taken a few days later) would seem to reflect your INR with diminished Quercetin. However, since you do not seem to have a parallel (same day) lab result to accompany your latest meter reading, it is difficult to draw any conclusions about the differential effect of Quercetin on meter vs. lab results. It may support your belief that Quercetin elevates your INR, but I'm not sure it says anything about the effect of Quercetin on lab results.


Is your INR being reported higher than lab value because of something that you consume? (Green Tea?)

If it's true that some medications/foods can FOOL the InRatio (and perhaps other meters) into reporting INRs that are considerably higher than actual lab values, SHOULDN'T THE FDA, MANUFACTURERS, SELF-TESTERS, AND MEDICAL FACILITIES THAT USE THESE METERS AND RELY ON THEIR ACCURACY know about it. Shouldn't there perhaps be research into substances that can have this effect on meter accuracy and a registry be developed? Shouldn't some effort (aside from my own empirical testing) be put into determining if this is an actual effect; if it happens to others taking the same substances; if it's reproducible; and which meters are effected?

If I had the resources, I would repeat my Quercetin + C 'experiment' using other meters in addition to the InRatio and the lab to document the results.

ITC, ROCHE, Others - if you'd like to help by providing meters and strips, PM me and we can discuss how to get your device included in my testing.

If we are interested in doing the work of answering these questions (which I think are good questions), it may behoove us to "crowdsource" the effort - that is, to recruit the participation of large numbers of people who are on anticoagulation. I will give some thought about how me might do that (and certainly welcome ideas). Among steering us toward answers to your above questions, it would also likely help answer the question of individual differences.

Best,
pem
 
Thanks for taking my comments apart. I think we're in overall agreement on all points.

I don't think that we, as a group, can do much of a 'scientific' paper on this, although it would be helpful to get the experiences of others. What would be good to learn is how closely lab values match meter values - and if any of those on this forum HAVE this data - whether the values are close or widely divergent - it would be good to know. However, if we tried to do this as real research, we'd have problems because we had no control about training the users to run their tests and no control over methodologies, etc.

I remember years ago that a cardiologist (or cardiac surgeon, I'm not sure which) wrote a letter to a medical journal about an arrhythmia (or other problem) that he experienced when he put the 'massage' from a shower massage against his neck. Sure, this was anecdotal and experiential, but was valuable because it pointed out a possible medical risk for some people. I don't know if there was any research follow-up on this issue.

My (our?) report on possible bias in meter reported INR values when we take certain OTCs or drink Green Tea (or something) would similarly be experiential and anecdotal, but may also point at areas where research hadn't been done. Plus, for people like me who keep their INRs low (or even for clinics looking to shoot low), it may point out a situation where 'low' INRs that are actually .6 - .8 lower than the meter reports could be dangerous for those tested.

I agree that it would be best to do all tests with lab and meter done in parallel so that the effects of Quercetin on meter results can be verified. I hope soon to have a meter that more accurately matches lab results, and I'll use these results to get a better feel for any actual differences from meter to meter to lab, and from times with and times without the medication in my system.

If I'm comfortable that the issue really exists, I'll be comfortable reporting this to the FDA and possibly as an experiential letter to an appropriate journal.

From my own recent experience -- even if we're not researchers at research labs or medical facilities -- reporting any suspicious medication-related errors in meter functioning can be life saving.
 
From my own recent experience -- even if we're not researchers at research labs or medical facilities -- reporting any suspicious medication-related errors in meter functioning can be life saving.

Certainly if you've identified a supplement (Quercetin) that affects one kind of measurement device (InRatio2) and not another (Lab), as you suspect, then it seems worth reporting.

I think a broader study could be conducted in cooperation with a site such as "patientslikeme.com" or "curetogether.org" (both excellent sites). I'm sensitive to the sources of variability that would bias a data set derived from a smaller population, as you suggest. But with a larger population, those sources of variability tend to wash out. If we could obtain, for example, parallel results from one hundred people, even if each person is contributing noise to the measurements due to Quercetin or Green Tea or alcohol, etc., I think overall trends could be identified via measures of central tendency.

Best,
pem
 
Good ideas. I wasn't aware of patientslikeme.com or curetogether.org, but it makes sense that such sites must exist, and may actually help to provide enough data for getting a (preliminary) answer to this question.
 
Last Thursday I made another meter vs lab vs lab comparison:
InRatio2: 3.1
Lab1 (ISO=1.8): 2.5
Lab2 (ISO=1.0): 2.3
Predicted Lab value (model2): 2.6

This is the first time that the lab results have been different from each other. But even in this case, not by much (0.2).

The average of the two lab values (2.4) is a bit lower than the value (2.6) predicted by model2 (please see prev posts).

So in this case, if I were to use my meter to generated a predicted lab value, I would believe myself to be in the therapeutic range. If I trusted Lab1, I would also be in range. If I trusted Lab2, I would be out of range.

I may collect a few more side-by-side samples. Each one builds (or erodes) statistical confidence in the model.

Thanks,
pem

Faithfully, I'm continuing to report my INR data :)

Due to a low INR on 5/10/2012, I increased my Coumadin from 6.5mg/day to 7mg/day, which I've taken for the past week.

Yesterday (5/17/2012), I had the following results:
InRatio2: 2.4
Lab1 (ISO=1.8): 2.4
Lab2 (ISO=1.0): 2.0
Average Lab: 2.2
Predicted Lab value (model2): 2.2

Here is an updated plot using average lab value when more than one lab value was available:

model2-20120517.jpg

Since my INR continues to be out of range on the low side, I will be increasing my daily dose of Coumadin to 7.5mg.

Best,
pem
 
Throwing a little more fuel on the fire:
I just tested with BOTH InRatio machines (InRatio and InRatio2) and got the same INR and nearly identical prothrombin times -- INR of 2.3. This puts me slightly below a range of 2.5-3.5.

The anticoagulation 'clinic' feels threatened by home testers, so I'm not about to tell them about this. They'd rather be ignorant of any patient's INRs EXCEPT for the times when a monthly blood draw is taken.

I haven't changed diet (well, yes, I'm probably eating even less than usual, if this is possible), and my dosage has been consistent. I'm not taking Quercetin (which seemed to raise my InRatio results by about .6-.8), and really don't have a reason I can pinpoint for the .8 drop from a week ago, when I last tested.

I don't plan to make ANY changes to my dosing -- but I DO plan to check again next Friday. (I think I've got a prescription for a blood draw the first week of June). I still have a hard time understanding why my clinic can be so comfortable with monthly blood draws. (They want to know if there are symptoms of an INR that's too high (excessive bruising, bloody nose, etc.), but a low INR is pretty asymptomatic. Of course, a thrown clot is a really GOOD symptom of low INR for a week or more).

As far as the 'clinic' is concerned, my INR of 3.1 when I last went there is what it will ALWAYS be between tests.

I just don't get it.
 
Back
Top