Are they really sick?

Are they really sick?

Predicted values for spirometry for elderly subjects, and the consequences for diagnosing COPD

Tjard Schermer & Philip H. Quanjer

November 29, 2006

Presentation made at meeting of general practioners in the Netherlands

December 2, 2006 T Schermer & Ph Quanjer

Are they really sick?

When dealing with spirometric data there are certain essential requirements that need to be met: 1 Use of appropriate equipment, regular calibration 2 Manoeuvres performed correctly 3 Quality assessment of data 4 Appropriate predicted values 5 Adequate interpretation (including bronchodilator effects) 6 If appropriate: classification of severity of any abnormality. Standardising a number of aspects of lung function testing was first undertaken in Europe in 1960 by the European Community for Coal and Steel, a branch of industry with a high prevalence of disease due to inhaled particles. At regular intervals international bodies in Europe and the USA improved and refined their recommendations and included a wider spectrum of lung function indices. Below is a chronological account of recommendations: 1960 Jouasset D. Normalisation des preuves fonctionnelles respiratoires dans les pays de la Communaut Europenne du Charbon et de lAcier. Poumon et Coeur 33; 1145-1159. 1971 ECCS Update. 1979 - ATS statement - Snowbird workshop on standardization of spirometry. Am Rev Respir Dis 119: 831-838. 1983 - Standardized Lung Function Testing, ECCS. Bull Europ Physiopath Resp 19, suppl. 5, 5-95. 1993 - ECCS-ERS update. Eur Respir J 6, suppl. 16, 5-100. 1995 - ATS update. Am J Respir Crit Care Med 152: 1107-1136. 2005 - ATS/ERS recommendations. Eur Respir J, series 1-5 ATS/ERS Task Force: Standardization of Lung Function Testing.

One is inclined to conclude from the above list that all has been attended to by now. However, that is not the case: whereas standards are issued in many fields, most people just pay lipservice to them. Sadly, supervising spirometric tests in clinical surveys shows that the quality and adherence to international guidelines often leaves something to be desired. Among the errors most frequently encountered are: No blast at start of the manoeuvre Premature end of manoeuvre Cough Only one or two manoeuvres Inspiration not to TLC Unsatisfactory reproducibility Selected FEV1 and FVC not the largest of 3 manoeuvres No calibration or linearity check of equipment

Premature end of manoeuvre

December 2, 2006

T Schermer & Ph Quanjer

Are they really sick?

No blast at start manoeuvre and no maximal effort.

Poor cooperation, results.

no

reproducible

No blast at start manoeuvre. It is more or less obvious how this condition can be improved; we will deal with that later. At the start of this century a number of people, the GOLD group, with the support of most if not all large pharmaceutical companies, addressed the problem of underdiagnosis of COPD. Their objective was to improve the diagnosis and increase awareness of underdiagnosis. They felt that establishing

December 2, 2006

T Schermer & Ph Quanjer

the true lower limits of normal for spirometric indices was too complicated for practicing physicians; they therefore introduced a very simple rule: if FEV1%VC after bronchodilatation is less than 70% this signifies airway obstruction, in fact COPD. They also provided guidance how to classify the severity of airway obstruction (table on previous page, GOLD recommendations 2004, see www.goldcopd.com). COPD is therefore a laboratory finding, similar to hypertension or anemia. The GOLD committee does not claim that any of its recommendations is evidence based. The GOLD guidelines quickly gave rise to a flourishing literature on COPD which led to the conclusion that the prevalence of COPD was even larger than hitherto suspected. This led to a threatening scenario of excessive rise in the burden of disease due to the ever increasing proportion of elderly and aged people in the population, so that the cost of healthcare would explode. Medical organisations worldwide were alarmed, adopted the GOLD guidelines, tried to take measures so as to identify and treat subjects with COPD at an earlier stage, and thus potentially stem the tide. Obviously Dutch general practitioners form part of this worldwide community and are prepared to make their contribution. In this document we will try to answer the question to what extent support for the GOLD guidelines can be retrieved from the literature and large databases. In doing so we focus on: 1 2 3 Is FEV1%VC < 70% a valid index of the presence of airway obstruction? Is the use of FEV1%predicted a valid index to assess the severity of airway obstruction? How large are differences in published predicted values for various spirometric indices, and to what extent do they affect the evaluation of spirometric data in diagnosing COPD, as well as the severity of airway obstruction?

Is FEV1%VC < 70% a valid index of the presence of airway obstruction? The Falling Ratio Working Group, an international group of investigators, retrieved from the international literature about 45 publica tions on FEV1%FVC and its lower limit of normal (LLN). In most of these publications the prediction equation for FEV1%VC is of the form: Y = a + bAge + cHeight and provides guidance how to compute the LLN. The two graphs (this and next page) show FEV1%VC for white men

and women as a function of a population of average height. With few exceptions the FEV1%VC, computed in keeping with the publication, falls well below 70%. It is also obvious that airway obstruction will be frequently overlooked in young adults if the 70% GOLD criterion is applied. Conclusion 1 The GOLD criterion for esta blishing airway obstruction is definitely not suitable for indiscriminate application to any population, as it will be associated with false negative results in young adults, and with false positives in older adults. How does the 70% cut-off perform in a healthy reference population (negative respiratory history, lifelong nonsmoker)? Assuming that the author has properly defined the LLN, then convention has it that 5% of a reference population

Are they really sick?

may produce a result below that LLN. In the example on the previous page. comprising 2269 healthy nonsmoking English men, this is indeed the case: 5.24% prevalence of airway obstruction. The LLN is therefore appropriate for this population (Falaschetti et al. Eur Respir J 2004; 23: 456-463). In a random sample of the population the 70% GOLD criterion, which neglects the age and height dependence of the LLN, produces false negative (blue) and false positive results (red). Just to clarify: if only the authors criteria (LLN) had been applied, then the blue circles would have been green (below the LLN) and the red ones would have been black (above the LLN). The LLN is therefore just above the blue and below the red circles. It is unlikely that a GP will perform spirometry on subjects without any likelihood of lung disease. What are the findings in a random sample of the population in England, hence including subjects with respiratory disease?

The picture is essentially the same. The prevalence of airway obstruction has doubled, 13.2%. A nice feature of a higher prevalence is that the percentage of false positives drops. The predictive value of a positive test has increased by only a small amount. The above population is more representative of the people seen in a GPs office. What about patients seen at a hospitals clinical and outpatient departments? Those are the patients the GP has referred to the chest physician. The illustration on the next page shows the findings in 2393 patients at Dijkzigt
December 2, 2006 T Schermer & Ph Quanjer

Are they really sick?

Academic Hospital, Rotterdam. The prevalence, and the predictive value of a positive rest result, is higher again. Needless to say that the larger the a priori likelihood of the presence of airway obstruction, such as in this selected patient population, the smaller the percentage of false positives. We shall see shortly that this statement has to be taken with a grain of salt. Using the GOLD guidelines it is not possible to classify all subjects. In a large Dutch population study (Vlaardingen-Vlagtwedde) 5.70% of subjects could not be classified. Khler et al. (Thorax 2003; 58: 825) could not classify 10% of 1,000 patients clinically treated for COPD; previous recommendations of ERS and ATS did not lead to this problem, the authors found.

What if I apply a set of prediction equations that has not been derived from the
December 2, 2006 T Schermer & Ph Quanjer

population I am going to apply them to? The above figure illustrates findings in 2406 Dutch males, a random sample from the population. The blue line re presents the LLN according to ECCS/ERS. If we would slightly elevate the level of the LLN, for example as high as the yellow line, this would obviously increase the number of false negatives and decrease the number of false positives. At this stage the predicted values due to Hankinson are widely used in the United States (Am J Respir Crit Care Med 1999; 159: 179-187). The predicted value for the LLN for FEV1%FVC is only slightly higher than that of ECCS/ERS. How does substitution of Hankinsons equations work out in Dutch men and women? Predicted value Increase positives FP%pos FN%neg ECCS Man Women 31.6% 0.0% 42.3% 0.3% Hankinson Man Women +16.6% +16.6% 17.9% 10.6% 3.2% 5.8%

FN = false negative, FP = false positive

In spite of the rather small change in the LLN the number of people identified as having airway obstruction with Hankinsons prediction equation increases by 1/6, and accordingly the percentage of false positives diminishes. Obviously the trend is reversed with regard to false negatives. It is therefore useful to have a closer look at the comparibility of predicted values and how differences will affect your decisions.

Are they really sick?

In previous text and illustra tions we displayed false negatives as a % of all nega tive test results, and false positives similarly as a % of all positive test results. That is not efficient. After all, false negatives are limited to the younger adults, and false positives to the older ones. The latter is usually regarded as the target group in the case of COPD. It is therefore more useful to use age-specific percentages. This was done in the illustration on the right for the large NHANES III population. Subjects were centered by age; thus age 50 means that persons were aged between 47.5 and 52.5 year. The message is quite bewil dering: above age 55 or 60 year the GOLD criterion leads to doubling the prevalence of airway obstruction (COPD). Hence for individuals in this age range in whom airway obstruction is diagnosed because FEV1%FVC < 70%, in about 50% of cases this is scientifically untenable. In particular the GOLD guidelines have been responsible for alarmist messages about the high prevalence of COPD in the population, and the future increase in the burden of disease because ever more people tend to grow older. Indeed the figure shows (red columns) an upward trend in the point prevalence of airway obstruction with age, and the percentages for age cohorts are indeed high, but the by and large equal percentage of false positive findings should in fact be added to the true negatives. These findings relate to a random sample of the American population. However, random samples from an English and Dutch population lead to quite similar findings. Conclusion 2. There surely is an upward trend with age in the prevalence of airway obstruction, but from age 50 in men, and age 60 in women, the prevalence is artificially doubled because the GOLD guideline fails to take into account that FEV1%FVC varies with both age and standing height. As the GOLD criterion: FEV1%FVC < 70% = airway obstruction (= COPD) leads to excessive errors, this criterion is untenable.
December 2, 2006 T Schermer & Ph Quanjer

Are they really sick?

10

Grading the severity of airway obstruction is conventionally based on FEV1%pred. The use of this index for clinical purposes is standard practice. Being sanctioned by use, is there any problem? Well, think about this. No author of predicted values has ever recommended to express FEV1 as a percentage of the predicted value for whatever evaluation of spirometric test results. Similarly no standardisation group has ever made such a recommendation. On the contrary, just like authors of predicted values such committees recommend to compute the LLN of FEV1 (and FVC, and other indices) lege artis, i.e. using appropriate and standard statistical techniques. Do consult:
ATS Statement. Lung function testing: selection of reference values and interpretative strategies. Am Rev Respir Dis 1991; 144: 1202-1218. Quanjer PhH, Tammeling GJ, Cotes JE, Pedersen OF, Peslin R, Yernault JC. Lung volumes and forced ventilatory flows. Eur Respir J 1993; 6 suppl. 16: 5-40. (ECCS/ ERS statement). Pellegrino R, Viegi G, Brusasco V, Crapo RO, Burgos F, et al. Interpretative strategies for lung function tests. Eur Respir J 2005; 26: 948-968. (ATS/ERS statement).

So what is wrong, and what are the consequences? Do please look at the graph on the right. It shows for men of 1.90 m height, and for small men of 1.65 m height, how according to the ECCS/ERS prediction equations FEV1 relates to age. Extending from the predicted value is a colored zone ending in red. The lower border of that zone represents the lower limit of normal, the 5 percentile: below that zone is 5% of a healthy non-smoking population from whom the equation was derived. The yellow interrupted line represents 80% of the predicted value. The smaller the predicted FEV1, the closer the line gets to the predicted value. The use of percentages is correct if the scatter about the predicted value is proportional to that value, so that the scatter becomes smaller as the predicted value becomes smaller. But that is not what is experimentally observed, however contraintuitive this may be. The graph shows clearly that the use of % predicted leads to an important age and height related bias. The proportion of false positive findings increases with age, with diminishing height, and is particularly large if people are both small and aged. We know that older generations were not as tall when they were young as younger generations, and that the ageing process is associated with diminishing standing height, so that both factors will aggravate the age-related bias. Conclusion 3. The use of % predicted leads to false positive findings at an older age, in small people, and especially in elderly small people. Use of percentage predicted has been sanctioned by usage, but its use is limited to the domain of lung diseases. One cannot complain that there are no publications on how to properly compute the LLN, but apparently this is not the literature favored by clinicians.
December 2, 2006 T Schermer & Ph Quanjer

Are they really sick?

11

What then is the proper way to establish the LLN? Let us consider an example.The illustration below depicts the FEV1 obtained in 832 Dutch

women who never had chronic respiratory symptoms and who were lifelong nonsmokers. The scatter about the mean (SD) comes to 610 mL. Differences in FEV1 between subjects are due in part to differences in standing height and in age. By performing regression of FEV1 on height and age we can establish the relationship and explain part of the differences. If we now use the equation we derived to correct measured values for differences in age and height you obtain the illustration on the right. The SD of FEV1 is now 477 mL, it decreased by 22%. The remaining scatter is called residual scatter, a measure of which is the residual standard deviation (RSD). The equation we derive is of the following form: FEV1 = a + bHeight + cAge +

= residual standard deviation (RSD)

When using %pred we want to know how far the observed value is removed (in %) from the predicted one. Similarly we wish to establish how many RSD the measured value differs from the predicted one. If the residual scatter is normally distributed the LLN is at: predicted 1.645RSD In the case of 80% predicted we (incorrectly) assume that the LLN is 20% below the predicted value. However, the appropriate index is to compute how many times the RSD the measured value differs from the predicted one, and we call the result the standard deviation score (SDS), also called z-score or standardized residual. For normal distributions the LLN is at SDS = -1.645. Computing the SDS is therefore a simple matter: SDS = (observed predicted)/RSD

December 2, 2006

T Schermer & Ph Quanjer

Are they really sick?

12

Interpreting the SDS is an easy matter. observed = predicted FEV1 at 5 percentile FEV1 at 212 percentile SDS 0 -1.645 -1.96

If the author has not indicated a different way to define the LLN, one can use SDS = -1.645 as the 5 percentile for any index in children and adults, men and women. Or SDS = -1.96 as the 212 percentile. Which prediction equations fit a Dutch population? The prediction equations almost universally used in Europe are those due to ECCS/ERS. If we apply these to a representative sample of the Dutch population (no symptoms, lifelong nonsmokers), this is what we find (in liters): Women Men Index FEV1 FEV1 FVC FVC Gender women men women men FEV1 2.06 4.49 Correction + 8% + 8% + 15% + 10% FEV1-predicted 0.22 0.30 FVC 3.71 5.65 FVC-predicted 0.52 0.60

Predicted values, except those for FEV1%VC, are too low. But this can be redressed by applying the corrections in the table on the left.

Comparison of the ECCS/ERS prediction equations with other ones discloses that the ECCS values are low; this has indeed been documented in a number of publications. After the above corrections the predicted values fit a healthy Dutch population. The figure on the right shows predicted values for women and men of average height, age 60 year; the adjusted ECCS/ ERS predicted values are now in line with a series of other predicted values.

December 2, 2006

T Schermer & Ph Quanjer

Are they really sick?

13

What is a valid age range for ECCS/ERS predicted values? Below is a quotation from the original ECCS publication (1983, p. 50): The summary equations were derived for an age range of 25-70 years, and for a height range of 1.55-1.95 m in men, and 1.45-1.80 m in women. According to various sources predicted mean values in cross-sectional studies do not decline with age between 18-25 years of age, so that they are the same as for a subject aged 25 years. For subjects older than 70 years, and for subjects outside the above height ranges, it is permissible to obtain predicted mean values by extrapolation from the regression equations. The prediction equations were not changed in the 1993 ECCS/ERS update. The two graphs below (FEV1 uncorrected for differences in height) show that indeed no harm is done in extending the valid age range to 80 yr. Obviously this still leaves the correction, alluded to on the previous page, to be attended to.

December 2, 2006

T Schermer & Ph Quanjer

Are they really sick?

14

What replaces % predicted? The severity of airway obstruction is conventionally gauged from FEV1%pred. But that index is biased by age and standing height and should, as explained, be abandoned in favor of the z-score (SDS). If we wish to maintain the arbitrary GOLD stages we must find a way to define limits that closely fit the ones defined by the GOLD group. This can be done by regressing FEV1%pred on SDS, or the other way around. In over 16,000 measurements in de NHANES III study this gives rise to the following relationship: SDS = -6.880 + 0.0686FEV1%pred Correlation 0.977, RSD 0.20. Substitution gives rise to the following table: FEV1%FVC Obstruction none mild moderate severe very severe % >70 <70 SDS >-1.645 <-1.645 FEV1 %voorspeld >80 <80 <50 <30 GOLD SDS >-1.645 <-1.645 <-3.44 <-4.82 new

GOLD

new

If an author has defined the LLN differently, obviously the recommended method replaces SDS= -1.645. The figure below shows how to easily work with the SDS. It shows a normal distribution of observations. The yellow numbers indicate the percentage of observations under the curve going from left to right. Thus 50 (median) marks the point where half of the observations are associated with an SDS < 0, and 50% > 0. 95% of observations is between -2 and +2 RSD, and 90% between -1,645 and +1,645 RSD (95% and 90% confidence interval, respectively). Below -1,645 are only 5% of observations in a healthy, nonsmoking population, and below -2 comprises only 2,3%. Thus -3,44RSD is equivalent to a chance of 3 out of 10,000 that the observation is compatible with one made in the reference group, and -4,82RSD is equivalent to <5 in 100,000 observa tions. A bit unusual, maybe and therefore at first sight a bit daunting, but in fact all it entails is replacing one figure (%pred) with another one (SDS), which is not too hard.
December 2, 2006 T Schermer & Ph Quanjer

Are they really sick?

15

Few people like change. So they usually object that these computations are unnecessarily complex. But that is not true. To compute %pred a division needs to be performed, just as in the case of the SDS; it is true that one also needs to know the RSD of the prediction equation. But is there any doctor who performs these computations himself? Of course not, it is all presented by computer, so why worry about the computational aspects? What happens to the severity grading if we replace %pred with the SDS? The figure below illustrates this for the large NHANES III population. The numbers 1-4 in the left figure represent the staging according to GOLD guidelines. On the right is the equivalent staging using the LLN and SDS. The number of people with airway obstruction falls from 2374 (GOLD) to 2010, i.e. by 15%. That, however, is a superficial conclusion. After all, in younger age groups a few hundred subjects are erroneously diagnosed as not having airway obstruction

when applying the GOLD criterion that FEV1%VC should be < 70%, and is appropriately identified by applying Hankinsons LLN. Furthermore a very large number of false positives is correctly staged by using the LLN. So overall there are very important age-related GOLD stage GOLD method LLN-SDS method shifts in the assessment I 1194 1490 of spirometric data, due II 933 261 to which mild obstruction 242 216 (GOLD stage I) becomes III 35 43 more prevalent, particularly IV among young adults, Total 2374 2010 GOLD II becomes much rarer, with only minor changes in the severer classes of airway obstruction. The GOLD guidelines, therefore, give rise to considerable inaccuracies in grading the less severe cases of airway obstruction. One can criticise these analyses for the fact that the GOLD guidelines require interpretation of FEV1%VC after bronchodilatation. None of the data presented comply with that requirement. However, according to a recent study by Johannessen et al. (Thorax 2005; 60: 842-847) this would reduce the prevalence of airway obstruction based on GOLD criteria by only 27%,
December 2, 2006 T Schermer & Ph Quanjer

Are they really sick?

16

which falls far short of reducing the gross overestimate of airway obstruction demonstrated in this manuscript. But apart from that, we are not aware of any evidence that an FEV1%VC ratio < 70% either prior to or after bronchodilata tion implies airway obstruction, or COPD. It is in addition a rather peculiar requirement. Must we really expose a population regarded as healthy to bronchodilator drugs? If the answer is affirmative, can we only arrive at the LLN for blood pressure in healthy people after administration of a drug that affects blood pressure and administration of a diuretic drug? Should we establish the LLN for the hemoglobin concentration in healthy subjects after administering supplementary vitamin B12, folic acid and iron? This gives rise to the following conclusions and recommendations with regard to measurement requirements, and the interpretation of measurements: Essential basic conditions: 1 Use appropriate equipment, regularly tested and calibrated. Generally few spirometers are sold with equipment (calibration syringe) required for checking linearity and calibrating output. Draw up a list of suitable hardware AND software. 2 Adhere to minimal requirements that have been internationally recom mended. Quality assessment during intervention studies reveals that frequently quality of spirometry is unsatisfactory. The following measures are suitable for making it difficult to neglect internationally accepted standards: 3 Quality control online, immediate feedback, via software. Apply the ATS/ERS (Eur Respir J 2005; 26:319-338) or NLHEP criteria (see footnote). Draw up a list of minimal requirements that software should comply with and inform dealers that henceforth only equipment and associated software that is on that list will be recommended to general practitioners. 4 Store quality assessment with the data. Thus it will always be clear how much weight should be attached to a measurement on file. It will not always be possible for a patient to comply completely (e.g. in the case of mental limitations, poor hearing, paralysis, pain, etc.). There is no reason to reject the good with the bad, as sometimes the measurements may yield clinically useful information even though they do not meet quality criteria. 5 Certify personnel and laboratories. Inevitable, because correctly administering spirometry is not easy, cannot be done casually, and requires professional training and maintenance of good laboratory practice, as well as postgraduate training. It should be possible to achieve this through cooperation with the national societies of chest physicians and lung function technicians. 6 Certification of laboratories to be periodically renewed. Certification for an indefinite period is not acceptable. The system will probably entail site visits. 7 Resolve the issue of predicted values and their lower limits of normal. This is an international challenge, important for daily medical practice. 8 Sick: FEV1, (F)VC, FEV1%(F)VC: Forget about the rest.
http://www.nlhep.org/resources-medical.html#review, click on Office Spirometry for Lung Health Assessment in Adults - A Consensus Statement from the NLHEP

December 2, 2006

T Schermer & Ph Quanjer

Are they really sick?

17

Spirometric tests give rise to a plethora of indices. They have no added value above the above. 9 Who is ill? Do not treat laboratory findings. Illness is a clinical condition. 10 When are pharmaceutical interventions appropriate? Unfortunately the literature is not encouraging. Apart from smoking cessation it seems that so far COPD eludes medical intervention, except to some extent in the severest stages. That underlines the necessity to only perform spirometric tests if there is prior evidence of respiratory disease. The combination of recommendations 1-5 will be invaluable in improving the quality of spirometric tests if general practitioners learn to take due account of them when evaluating the patients condition. Interpretation of spirometric data: 1 The GOLD criterion FEV1%VC < 70% leads to gross overestimates in the prevalence of airway obstruction in people over age 50 yr because it is associated with a pronounced age and height bias. This in turn leads to very significant over-diagnosis of COPD in this age range. 2 The same criterion is responsible for under-diagnosis of airway obstruction in young adults. 3 FEV1%predicted similarly leads to a height and age bias. Arising from this the severity of airway obstruction is systematically overestimated in elderly and small subjects. 4 All published research and recommendations issued by expert committees recommend to use the lower limit of normal, none recommend the use of % predicted. 5 A scale is presented in which %pred is replaced by standard deviation score, allowing to assess the severity of airway obstruction in a scientifically correct manner. If people wish to adhere to the four GOLD stages there will always be a significant proportion of subjects whose measurements cannot be classified, i.e. in those who have no airway obstruction (FEV1%FVC > LLN) but FEV1 < LLN. Each scaling system has so far been arbitrary, and much more research is required to establish to what extent pulmonary function data are prognostic of increased morbidity and premature death, allowing the construction of a validated scale.

December 2, 2006

T Schermer & Ph Quanjer

Are they really sick?

18

Finally: It is a crucial problem that we do not know how to define COPD. The general pattern of reasoning is more or less as follows: Grass is green. Cows eat grass. Hence cows are green. In the context of COPD: COPD is associated with airway obstruction. Therefore airway obstruction is COPD. The medical profession has agreed that airway obstruction in COPD is typically hardly responsive to bronchodilator drugs. Hence irreversible airway obstruction is COPD. Free translation: we have little understanding of the nature of COPD, hardly comprehend what it is, everyone has seen it and has therefore formed an opinion about it. That is what the following poem is about.

The Blind Men and the Elephant It was six men of Indostan To learning much inclined, Who went to see the Elephant (Though all of them were blind), That each by observation Might satisfy his mind. The FIRST approached the Elephant, And happening to fall Against his broad and sturdy side, At once began to bawl: God bless me! but the Elephant Is very like a wall! The SECOND, feeling of the tusk, Cried, Ho! what have we here So very round and smooth and sharp? To me tis mighty clear This wonder of an Elephant Is very like a spear! The THIRD approached the animal, And happening to take The squirming trunk within his hands, Thus boldly up and spake: I see, quoth he, the Elephant Is very like a snake! The FOURTH reached out an eager hand And felt about the knee, What most this wondrous beast is like Is mighty plain, said he: Tis clear enough the Elephant Is very like a tree! The FIFTH, who chanced to touch the ear, Said: Een the blindest man Can tell what this resembles most; Deny the fact who can, This marvel of an Elephant Is very like a fan! The SIXTH no sooner had begun About the beast to grope, Then, seizing on the swinging tail That fell within his scope, I see, quoth he, the Elephant Is very like a rope! And so these men of Indostan Disputed loud and long, Each in his own opinion Exceeding stiff and strong, Though each was partly in the right, And all were in the wrong! Moral So oft in scientific wars The disputants, I ween, Rail on in utter ignorance Of what each other mean, And prate about an elephant Not one of them has seen!

John Godfrey Saxe

December 2, 2006

T Schermer & Ph Quanjer