20 Dec 2010, 7:51 PM
Blood test flawed?
What makes this case study so important is the meticulous collection of the facts by the farmers themselves- they recorded the skin test measurements in mm of every animal and and compared this to the tables for standard and severe readings of the skin test. This together with the figures obtained for the Bovigam test for each animal under the Freedom of Information Act make their case so very striking. Dr Ruth Watkins, the virologist who provided the laboratory report below wonders (email 31/10/10) what Dr Vordermeier would say now after there have been more herd tests, some of which she has heard about and that have involved large numbers of animals being culled on the Bovigam test result after a negative TST. Has he changed his mind on latency? Would he rather do peptide based testing to eliminate natural killer cell false positive results for infection with M bovis?
Back in 2007 a large organic farm (dairy and beef) was involved in their first incident regarding bovine TB. It was not as a result of the usual routine skin test of the cattle but was from the tracing back of an animal that had been sent for slaughter and found to have TB lesions. The farmers involved, an established family farm of many generations that had farmed for years, had never experienced a bTB breakdown in the last fifty years. The animal concerned was alleged to have come from their farm and was one that had been purchased from a neighbour as a young calf to suckle a cow that had lost her own calf. The purchased calf was reared on the farm for two years and then went for slaughter as she was not their pedigree and they wanted to keep the sock pure. In fact, subsequently, they have always been skeptical that the animal involved was actually one that was from their farm and wonder if it was case of mistaken identity. This was because the strain of bTB identified was one found in Cornwall and different to the strain found in neighbouring farms that had suffered TB breakdowns. Certainly, as heads are removed as part of the BSE precautions, one could assume some errors may result. DEFRA insisted that all animals (over 700) on the farm had to be tested for bTB and the blood test was used. It is understood that this was one of the first of six farms used to trial this method on such a large scale. The test took four days and was undertaken by a Spanish vet.
The results of the blood test revealed that 107 animals had bTB. They were all young heifers and ones that had been in an area of the farm away from the dairy and kept together. All these animals had to be slaughtered and went into the human food chain. They were all of pedigree stock and some were in calf. The family and stockmen were devastated as these were pedigree animals and all were in good condition. Even the farm workers were seriously emotionally upset as they had put a lot of time and care into rearing the animals. The family were concerned and asked to see the results of the blood tests. Initially DEFRA refused, saying they would not understand the results. However, only after a third request for this information was made under the Freedom of Information legislation were the results made available to them. It was clear from their results that there was an obvious error with one of the animals. It had been marked down as failing the blood test but the result indicated clearly it had passed. When the family pointed this out to DEFRA it was accepted that a mistake had been made and she had been wrongly classified as having bTB. They agreed compensation would be paid. Two independent valuers (paid for by the family) from Devon and Wales valued the animal, taking into account the loss of a pedigree animal of excellent breeding stock, and the loss of future milk and calves from the animal. The organic animal would have had a likely milking span of ten years, and would have produced ten calves. This was taken into account in the valuations. She was a pedigree animal called 'Nancy', named after the person who had started the herd back in 1922. Extensive calculations from both valuers, working independently of each other, revealed a value of £10K. DEFRA insisted on sending their own valuer. This valuer refused to take into account the additional premium organic milk gave and valued the animal at £7K. However, DEFRA even refused to accept their valuer's valuation. One of its vets said as the cow had calved, a replacement animal could easily be found and so DEFRA need only compensate for the six weeks loss of milk. The family were offered just £271! When they sought the advice of a solicitor it was revealed that the cost of legal action would be around £10K so even if the case was found in their favour they would be no better off. If they lost they would lose a further £10K, so it was not worth pursuing the matter. The family were powerless to do anything. They undertook a lot of research into the subject and it became clear there were a lot of problems with the system.
The family insisted on visiting the laboratories where the blood tests are processed. Accompanied by a DEFRA representative, and their own expert, who made notes, it was revealed there was much scope for human error (see notes taken set out below).
None of the 107 animals slaughtered as a result of the blood test revealed any signs of TB lesions on postmortem and tissue cultures for all were negative too. Bovine TB was therefore never confirmed in this herd. In the routine skin test that followed six to nine months later, 29 animals were identified as inconclusive. On the subsequent test they all passed and no reactors or inconclusives have been revealed since then, despite TB breakdowns continuing on neighbouring farms.
The farm is only just getting over the dire consequences of the breakdown, which put the family under enormous strain. The stress affected the whole family. It has taken considerable time to get stocking levels up and a considerable amount of money, time and effort has had to be spent in dealing with the whole sorry affair.
Ruth Watkins is a microbiologist. She attended the meeting with the family and her notes are below.
Summary of visit to Luddington VLA on 10 July 2007, written by Dr Ruth Watkins (email dated 31 October 2010 refers).
Very nicely situated laboratory, plenty of lab space, the staff welcomed us. The only person impatient to see us go was the manager who wished to get on with his meeting to discuss the question of the bovigam test cut-off with Dr Martin Vordemeier and side-kick, a post doc.! We had 2 hours. We enjoyed our visit and felt that our hosts went out of their way to be helpful. We were appreciative of their hospitality and helpful responses. They are not responsible for the policy and use of their test results, that falls to DEFRA. They simply get the contract for testing and relay the results to DEFRA.
Summary of bovigam test as carried out at Luddington
First part of the test
Receive blood samples next day, each blood sample having been placed in either one of two different types of transport box when taken; the blood must remain within 10 - 26 0C. The blood is processed the day of receipt.
Each specimen has a unique barcode so that it can be matched to the holding and the individual cow it was taken from. (the origin of one of the mistakes was not discussed, a cow who was recorded as two different animals with the same eartag number each with a different result)
(I remain unclear on the origin of the labels- preprinted especially for that farm from the BCMS records? Therefore human typing error booking in specimens is eliminated?)
From each specimen 200 microlitres aliquoted into each of 4 wells on a multi-well plate in duplicate (1600 microlitres in all). 25 microlitres of antigen added to each; well1 staphylococcal superantigen (titrated so that its effect is maximal, on the plateau of its non-specific stimulatory effect), this is the positive control, well 2 bovine TB antigen, purified protein derivative M bovis culture, referred to later as B, well 3 an avian mycobacterial antigen similarly prepared to that in well 2, referred to later as A, and finally in well 4 as the negative control the growth medium RPMI is added.
These are incubated overnight at 37 oC. They are spun down next day in a multiwell plate centrifuge and 110 microlitres of supernatant is aliquoted off and stored in matched (?) barcoded multiwelled ELISA plates at –70 oC until they are assayed as a batch for gamma interferon by ELISA at a later date. Haemolysis of some samples apparently does not interfere.
All the pipetting is by hand. The initial steps on receipt of the specimen is in one lab, using individual pipettes with plugged tips to handle each specimen. The antigens might be added by multichannelled pipettes using a trough for each antigen as is done in the next lab (where all the next steps are carried out). There is opportunity for human error in pipetting individual specimens and aligning specimens for the work and identifying plates. However the environment and work conditions, the intelligence and attitude of the staff makes this unlikely, at least maintained at the lowest level possible. It is essential that there should be no splash contamination of the superantigen into other wells -but carefully done using the plugged pipette tip held onto the side of the well this should not happen. Incidentally it would be safer to add the cells to the antigen, 200 to 25 rather than 25 to 200 to avoid error as it would be difficult to tell if both were added to each well in the latter instance. However this is probably not done as I expect the same tip is used to aliquot each specimen into the 8 empty test wells. I would like to see the SOP over this and other issues referred to later. (As the test is not automated except for an ELISA plate washing facility used during the gamma interferon ELISA, procedural details to eliminate human error are all important)
The second part of the testing, for gamma interferon.
The gamma interferon released is assayed by a standard ELISA test format. This is a sandwich test and the details of the several steps were not described to us but would be in the SOP. Suffice to say that first each specimen would be added to the coated wells with a pipette. After incubation this would be washed out and the next layer of the sandwich added in solution by a multichannel pipette, incubated, washed etc. The final result is obtained by scanning a colour developed in the final solution in each well and this is recorded in optical densitiy units, OD, so that the presence of gamma interferon is objectively measured.
The cells in the blood sample from the cattle that release gamma interferon do not amplify overnight so the test measures the potential gamma interferon release of cells already present in the blood at the time the sample is taken. In cattle up to 6 months old there can be a significant release from natural killer cells even giving strong positive results, this is non-specific. It reflects the ratio of gamma/delta T-cells in young cattle, which take about 6 months to fall to a low level (similar to that in humans) when they make up only a small proportion of the total T-cell population in blood. Thus the gamma interferon test cannot be done to assess infection with M bovis until after 6 months of age in cattle.
The release of gamma interferon as a specific response to TB is taken to prove prior exposure of the animal’s immune system to TB because they have been exposed and infected thus they have developed a specific immune response. The gamma interferon releasing cells are certain subpopulations of T-cells and they need the intervention of antigen presenting cells in the blood which present the antigen added to the well to them and which stimulate them to release gamma interferon. Thus the viability of the cells in the blood sample is vital to the test.
Procedures in the lab and giving results.
50 microlitres are aliquoted from each thawed sample for gamma interferon testing. A further 50 microlitres remain in case the ELISA test needs to be repeated. Thus each blood specimen has all results in duplicate as the test is set up in duplicate from initial receipt of the specimen. Obviously this is done because in order to repeat the test from the beginning another sample must be taken from the bovine, so it is reasonable to set up the test in duplicate to avoid having to rebleed the animal.
The ELISA is carried out as outlined above.
A result can only be valid if the positive control and negative control for each specimen perform as expected. We did not see the rules for this, these would be detailed in the SOP. High background might be evident in too high a result for the negative control and if the cells were not viable in the sample the positive control would give too low a result. The results are presented as OD reading for the bovine antigen, B, minus the OD reading for the avian antigen, A. B – A must be 0.1 or greater to be considered a positive result. There is no indeterminate range; the result is negative or positive and the cut-off taken as 0.1. Theoretically in a perfect world the cut off of this test would be 0.01 as the result is a measure of difference between the response to two antigens and the response to M bovis should be greater than that to M avium spp. if an M bovis infection has occurred.
The laboratory requests a repeat blood sample in the case of an invalid result on the positive or negative controls, but it gives out the invalid result. They aim to keep requests for repeat samples below 5% if possible of all specimens received. I do not believe these invalid results should be issued except to request reepeat samples.
One error from the testing has been addressed, the case in which an animal with a negative result, B – A = less than 0.1, was called positive and culled by DEFRA. When the results are released on the computer report to DEFRA a negative result is automatically interpreted as such now and cannot be over-ridden by an operator wrongly pressing the positive key instead of the negative key; human error in reporting has been ruled out.
The volume of testing on bovigam and perhaps other tests may not justify fully automated robotic testing so the testing procedure is manual apart from washing ELISA plates. This could lead to a grater number of human errors. However I would say it seems the test is carried out correctly and as well as it can be in the circumstances- full automation is complex to run and could be problematic in the short time span after receipt available so manual back up would always be necessary as an option.
The strong impression the scientists, particularly Dr Martin Vordermeier left us was that the bovigam test was robust and the vast majority if not all of the bovigam results on the herd in question were true positives. The cross reactivity of other organisms was he felt ruled out by prior validation and experience of using the test, and environmental mycobacteria not he thought likely to account for the results. He offered to check any further tests by the bovigam test using CFP 10 and ESAT 6, which were available as an extension of the bovigam routine test that uses purified protein derivative antigens from culture. Unfortunately all the hundred bovigam positive animals of this herd were culled on a slaughter order so no checks could be made, either repeating low test results or for non-specificity due to environmental mycobacteria infection.
It would be intriguing to know if the discussions held did change the cut-off to be used, so that the lowest positives only just above 0.1 were not after all culled- a rational policy would be to rebleed those with indeterminate or low positive results and check them by the CFP 10 and ESAT 6 test which is generally more specific.
Discussion of the results with the scientists referring to the doubts I had on the validitiy of the interpretation of the Bovigam test
In this case 258 animals gave a result between 0 and 0.09, deemed negative below the cut off of 0.1. To put the test in a fuller context, 211 animals gave a result in the negative range, between –1 and 0 and there were 47 animals with results between 0 and 0.09. 36 animals gave a result between 0.1 and 0.19, interpreted as positive just above the cut off of 0.1, and 19 animals gave a result between 0.2 and 0.29 showing the tailing off of the test results. How repeatable are these results just above the cut off and so close to it?
We never discussed how repeatable the bovigam testing was- how many low positives would be negative next time they were tested? How many just negative results would be just positive when repeated? I do not know if the repeatiablility of the test has been examined.
We did discuss the plot I had made of B – A, the results, on one axis against the number of animals for each value on the other axis. For a herd test this is quite legitimate. The plot I had made showed that almost all the cattle tested fitted one bell shaped curve, all the "negative cattle" and the low positives which were in the tail of the "negative cattle" bell shaped curve. There was no clear separation of these low positives from the negatives. This is not satisfactory for a diagnostic test especially one that is not validated in any way ie checked out on that animal by a different test using a different strategy.
There were a small number of positive results lying outside the bell shaped curve, the 17 animals giving results between +1 and +5 being clear examples of this. It is disturbing to learn that the false positives due to natural killer cell activity, non-specific immunity, can give a strongly positive bovigam test result (in tests used in human diagnosis false 'positives' usually give weak results which can be picked out and checked, falling between the negative and positive results of validated negative and positive samples). The use of peptides, CFP-10 and ESAT 6, instead of purified protein from cultures in the Bovigam test for example, would prevent the false positives due to non-specific immunity of natural killer cells because these cells respond to large or native proteins as presented in the Bovigam test.
The cattle owners felt that organic cattle had not been assessed in the bovigam test before and it was possible that their state of non-specific immunity, their natural killer cell activity for example, might be different, and perhaps this might be so in organic pregnant heifers for example. The scientists had no answer to this question. Environmental mycobacteria infection was also discussed and was not thought to be the cause of the results on the bovigam test on the this herd by the scientists.
Disappointingly no other complete herd test results were available. These if available are not published on the mistaken reason of confidentiality, after all the names of the farms are not necessary. We would not get very far in human medicine if no epidemic information of outbreaks was available to enable understanding of infection. This can be done without naming names.
Another reason why there is little comparable information to the family is that the TST (tuberculin skin test) positive animals are slaughtered before testing the rest for gamma interferon in most herd gamma interferon tests. However as DEFRA have some animals positive on both tests to use in their vaccine studies (see later) this must occasionally be done simultaneously and not always serially. In fact it so happened that no animal tested positive on the TST on the standard interpretation in this herd (see later).
How should the large number of bovigam positive animals be interpreted in the this case? There are no records for an organic herd tested similarly it seems.
In fact other than the index case diagnosed on lesions and culture post mortem when the heifer was sent as a beef animal to slaughter, so that her TST or bovigam testing status is unknown as these were never done of course, no other of 250 beef and TB slaughtered animals has had lesions seen- the TB culled animals went into the food chain so presumably were inspected after slaughter- no other animal has yet been seen to have had TB on post mortem.
Of the TST results, the 4 and subsequent 3 positives, these were only positive on the severe interpretation of the TST and not positive on the normal interpretation of the TST, on which they were inconclusive reactors. Of these seven TST severe only interpretation positive, four were negative on the bovigam test done in February, two were positive and one was just above the cut-off.
If the herd had been TST tested in the first place as part of a schedule rahter than in response a report from the abattoir, it is possible that only inconclusive reactors would have been detected on the TST- if the heifer with lesions had been anergic, that is TST negative (it is possible that she was never postive it being a case of mistaken identity at the abattoir). Yet an additional one hundred or so animals were revealed on the bovigam test and culled.
The index case, the heifer with lesions, was placed in a group of 25 other beef animals in a shed separate from other sheds that winter until she went with the other beef animals to slaughter on the 22nd of December. The herd TST and bovigam test was done 9 weeks later on the 27th of February. Time for the skin test to become positive in cases recently infected possibly by her before she went to slaughter. The vast majority of the bovigam positive cattle were not in her shed. They were young animals, for beef and pregnant heifers for the milking herd, and some dry cows- I presume mostly at a separate farm of the several farmed by the family to the dairy herd.
How can this be explained if the majoritiy, if not all, of the bovigam tets were true positives? The scientists had no answer to this question as they believed all the results were true positives.
Assertions were made on the fact that animals without lesions could be infectious (I have heard this before from Glyn Hewinson and it applied to animals in the initial period of infection before they have made any immune response, the first few weeks after infection). Other than a small experiment done in laboratory conditions on cattle at Weybridge there is no wider evidence for this in cattle. Also this does not generally apply in human TB.
The scientists present seemed reluctant to admit that the bovigam test could be detecting latently infected animals as happens in human testing which limits the usefulness in general of the gamma interferon test in humans (one third of the entire human population is said to have latent TB). There is no saying when latent infection becomes active and progressive though this can happen with old age, again analagous to humans.
I have heard also from Glyn Hewinson that bovines do not have a truely latent state- however Dr Vordermeier is going to look at latency in cattle. Though 90% of TST positive animals are gamma interferon positive, they know that the same is not true of the converse when a much larger number of gamma interferon positive animals, more than 10%, are negative on TST. Unfortunatel in my opinion understanding is hampered by the peremptory slaughter of TST positive animals and subsequently bovigam positive animals before retesting and evolution of test results is documented.
Discussion of some test series more evidence is required on these examples given by Dr Vordermeier
A On the infectivity of infected animals
20 reactor cattle were purchased as infection donors in vaccination studies, they were TST and gamma interferon positive. They were divided into four groups of 5 each in a control study and housed with 5 calves. In a period of 19 months 18 out of 20 calves converted on gamma interferon testing, and all were TST negative excepting one who was an inconclusive reactor. On culling they had no lesions visible, NVL, and yet M bovis could be cultured from the lymph nodes of 5 of these 18 gamma interferon positive calves.
My own observation on this is two fold, first no information was given on the shedding of M bovis from the reactor cattle, still alive and in vaccine studies, and secondly that latent TB in humans is hard to find histologically and often cannot be cultured even from the primary lesion, the Gohn’s focus in the lungs when this is found. The bovigam testing if truly positive could be picking up latently infected animals as it does in humans.
B On the correlation of culture, TST and gamma interferon positivity.
As an example of a whole herd test he gave that of 200 cattle observed for 2 months out of which at culling 20 were found to be culture positive for TB. He could not relate whether lesions were seen at PM but assumed they were in most. Only 3 of the 20 culture positive animals were TST positive, and 14 were gamma interferon positive including the 3 TST positive animals. Of the 20 culture positive animals 6 were not detected by either the TST or gamma interferon testing in life at all.
My own observations on this are as follows: It would be good to know how many culture positive animals had visible lesions at post mortem. Were any shedding the bacterium from the lungs? Bronchial sputum sampling done at PM could help to answer this question. If the animals were kept together for the preceding 2 months the 6 culture positive infections not detected on testing at all in life could have been very recent infection, acquired within the week prior to culling. It would be helpful to know the PM on those 6 animals: if lesions were visible then they could not have been very recent infection but were advanced anergic infections. (Perhaps the two types of infection were present).
Some background on sensitivity and specificity
The calculation of the specificity of the skin test (standard reading) and its sensitivity is based on very careful postmortem and culture of the herds under observation and published during the 1990s. The number of animals in these herds that were gamma interferon positive, truely so, and therefore had latent, culture negative, infection was not assessed. Nor were the culture positive but TST negative animals assessed for gamma interferon. Thus though the specificity of the positive standard interpretation of the skin test is high the sensitivity in detecting all infected cattle in a herd if this is to include latently infected cattle as disclosed by truely positive gamma interferon testing, must be lower than the published figure of about 80% sensitivity.
In this herd the sensitivity is 0% on the standard interpretation of the skin test, or if the index heifer would have tested positive on a skin test then about 1%. As 9 weeks had elapsed between her slaughter and the testing of the herd simultaneously for both delayed hypersensitivity, skin testing, and gamma interferon release in response to M bovis antigens, by the Bovigam test, recent infections from her, the index heifer were ruled out as these should have progressed from gamma interferon positivity alone to skin test positivity as well. Do the results on this herd Bovigam test represent a large number of latently infected cattle? Or do they represent a large number of recent infections not yet skin test positive from an undetected source? Or, has the gamma interferon testing been almost completely wrong, ie most positives were false positives? Even if 50 were false positives then the sensitivity of skin testing is at best only 2%! This should be very worrying to the policy makers.
Dr Ruth Watkins 18/7/07 from notes.
Spelling mistakes corrected and clarification made on 31.10.2010.