Blood values: Landis, Garmin, and that 48.5

[UPDATE: oops! Everyplace you see "48.5", note that the actual value is 48.2. Sometimes we should look more carefully at the primary documents.]

One of the arguments offered by those who think "Landis doped" comes from blood values that were presented late in the process. These never really came up in the litigation, because they aren't relevant to the issue of testosterone doping which was the charge.

They are arguably relevant if one is suggesting that Stage 17 was enabled in some way by Oxygen vector doping, through either some EPO-like substance or blood doping.

The argument being made is based on a single datapoint in the collection of Landis hematology data, the reported hematocrit value on 11-jul-2006. This was after an easy stage following the first rest day of that tour.

Here's the data in raw form:

Figure 1: GDC 1372, located in the GDC Binder on PDF page 1801.
UCI collected hematology values for Landis.
(click for bigger)

Conveniently, a discussion that tries to hang Landis on this crops out the irrelevant stuff, and presents just the germane numbers, focusing on the 48.5 hct value:

Figure 2: Edited version of Figure 1, more legible.

The question that is asked is, "does a rising hematocrit 11 days into the tour indicate doping?" And the assumed answer is, "of course it does, fool!" Well, fools we may be, but that's not necessarily a supported conclusion.

As we know, the UCI limit is 50, and while 48.5 is high, it isn't pushing the 50 level hard, so there's mandatory cause for a "no start" by the UCI.

But this isn't the only answer. We well know that hematocrit is very sensitive to hydration, and we don't have any data to indicate Landis' condition on that day. The reported value is also sensitive to the particular machine being used, and its calibration.

Phonak, which was the first team to start trying internal controls of any time, checked rider's HCT on its own, and would not see the UCI values. The internal limit was 48, to be sure of clearing the UCI limit of 50. How then did Phonak let a 48.5 slip? Probably because their value wasn't 48.5, or because they knew something about Landis' hydration state that would explain an upward fluctuation.

If one goes back to Velonews TV interviews during the period, you can hear that Landis has the rattle of a cold. Did a cold cause underhydration with fluid loss, or did some treatment he got do so? We don't know, nor do we have Phonak's internal control values, which seem not to have been recorded.

So, what can we say about the raw 48.5 hematocrit? Unclear. Since the argument that it means something is about contextual understanding, let's look at some more context. We'll start with the other values reported in the UCI document, the Hb and Reti (reticulocytes) values.

Let's take a peak at what Michael Ashendon said recently to The Boulder Report:

Perhaps because of the UCI's old fixation on hematocrit, many of us similarly key on that value as an indication of clean racing. But according to Ashenden, it's one of the least important. Hematocrit, or the percentage of total blood volume that is red blood cells, fluctuates by large amounts in even normal human beings. It's susceptible to issues ranging from the subject's hydration to proper sample transport (improper refrigeration causes RBCs to swell, disproportionately increasing their volume).

He also notes the wildly varying values between testers, consistent with Phonak seeing different values than the UCI:

One of the first things that Ashenden noted in examining results was the variability between official UCI figures and ACE figures, even for values taken just days apart. The UCI numbers were vastly more uniform. "Almost like a different person," said Ashenden of some of the ACE results.

What does Ashendon look at instead of Hct?

"The first thing I look at in interpreting results is reticulocyte count and OFF-score, then hemoglobin," explained Ashenden. "I also look at anything about a sample that is strange and might explain off values due to transport and testing issues - mean cell volume and hematocrit chiefly."

The off-score is computed with the formula: (HB*10) - (60 * SQRT(reti))

Extracting values by eyeball from the graphs that are published in the article, we get the following values for Vande Velde and Millar. To see if our data and off-score calculation method is correct, we pulled out the reported off-scores with the ones we calculated, and they are all pretty close, so we think we have decent data and computations. Due to the vagueness of the graphs, the dates are approximate.

Vande Velde				computed	reported
Date	event	HB	Reti	Offscore	0ffscore
10/27/07	off season	16.75	0.60	121.02	120
01/23/08	alt training	16.00	1.00	100.00	100
01/25/08	alt training	16.50	0.80	111.33	111
02/15/08	camp	15.25	0.60	106.02	107
02/24/08	camp	15.00	0.90	93.08	92
04/01/08	early season	15.50	1.20	89.27	90
04/18/08	early season	15.00	0.90	93.08	95
04/24/08	early season	14.25	0.80	88.83	90
05/01/08	camp	16.00	0.90	103.08	102
05/28/08	giro	15.00	0.70	99.80	100
07/16/08	tdf	14.75	0.50	105.07	105
avg		15.45	0.81	100.96	101
stdev		0.77	0.20	9.84	9

Table 1: Vande Velde's data from Slipstream.

Millar				computed	reported
Date	event	HB	Reti	Offscore	offscore
10/27/07	off season	14.50	1.80	64.50	63
11/29/07	off season	14.75	0.70	97.30	95
01/07/08	alt training	15.00	0.70	99.80	98
01/24/08	alt training	14.50	0.60	98.52	100
02/07/08	alt training	15.00	0.60	103.52	105
02/24/08	camp	14.25	0.80	88.83	90
03/25/08	early season	14.75	0.70	97.30	98
04/07/08	early season	14.50	1.20	79.27	78
04/15/08	early season	14.25	0.80	88.83	90
05/03/08	camp	14.50	0.70	94.80	95
05/28/08	giro	15.00	0.80	96.33	95
07/16/08	tdf	15.25	0.40	114.55	115
avg		14.69	0.82	93.63	94
stdev		0.32	0.36	12.55	13

Table 2: Millar's data from Slipstream.

Now, let's take Landis' data from the UCI and do the same analysis that Slipstream offered as proof that their two aces are clean:

Date	event	HB	Reti	Offscore
03/03/05	Paris-nice	15.60	0.98	96.60
06/07/05	tdf	14.80	1.07	85.94
07/18/05	tdf	15.30	0.45	112.75
07/23/05	tdf	14.50	0.74	93.39
08/25/05	vuelta	13.90	0.46	98.31
03/05/06	Paris-nice	15.80	0.99	98.30
06/29/06	tdf	15.50	1.30	86.59
07/11/06	tdf	16.10	0.92	103.45
avg		15.19	0.86	96.92
stdev		0.73	0.30	8.76

Table 3: Landis' data from the UCI

As we see, there is nothing in the Landis data that is out of the range of normal; if anything, it is more normal than Millar's.

We note also the absence of other hematology data from the UCI, even though Landis has blood taken on other occasions during the tour. In particular, there are no blood values offered for after the one in question, and we're working from two data points and trying to find enlightenment.

Well, what was going on in the tour from one data point to the other? Not very much. It was pretty much a week off, as we see from the stage descriptions and the data reports that Allen Lim made at the time, and which are also in the exhibits.

stages	what	place	kj	moving w	pedal w	pedal wkg
Prologue - July 1: Strasbourg ITT, 7 km	tt	9	no data
Stage 1 - July 2: Strasbourg - Strasbourg, 183 km	sprint	46	3075	205	250	3.56
Stage 2 - July 3: Obernai - Esch-sur-Alzette (Luxembourg), 223 km	sprint	30	3934	195	256	3.67
Stage 3 - July 4: Esch-sur-Alzette - Valkenburg (Netherlands), 216 km	break	44	3969	222	276	3.96
Stage 4 - July 5: Huy (Belgium) - Saint-Quentin, 215 km	sprint	22	3760	209	256	3.67
Stage 5 - July 6: Beauvais - Caen, 219 km	sprint	34	3749	196	242	3.48
Stage 6 - July 7: Lisieux - Vitré, 184 km	sprint	59	3349	223	275	3.96
Stage 7 - July 8: Saint-Grégoire - Rennes ITT, 52 km	itt	2	no data
Stage 8 - July 9: Saint-Méen-le-Grand - Lorient, 177 km	break	37	3481	227	274	3.94
Rest Day - July 10: Bordeaux
Stage 9 - July 11: Bordeaux - Dax, 170 km	sprint	20	2624	203	234	3.42
Stage 10 - July 12: Cambo-les-Bains - Pau, 193 km	mtn stage	50	4377	246	309	4.45
Stage 11 - July 13: Tarbes - Val d'Aran/Pla-de-Beret (Spain), 208 km	mtn stage	3	5870	267	314	4.52
Stage 12 - July 14: Luchon - Carcassonne, 211 km	mtn stage	20	4199	244	285	4.16
average 1-9		30.3	3492	210	257	3.7
stdev 1-9		17.47	464	12.53	15.9	0.22

Things indeed got much harder right following - 5870 kj vs. 3500 average in all preceding. It's not unreasonable to think Landis was recovering from hard training before the tour during the first week, which might also account for some HCT rise independant of hydration issues.

From this, we offer the following thoughts.

1. Raw HCT values are problematic, as testified by Ashendon.
2. Landis' values for Hb and reticulocytes are pedestrian, or at least comparable to those of "clean" riders provided by Slipstream.
3. If one wants to draw any conclusions against Landis from the hematological data available, then one ought to wonder about Millar's first off-season value.