olfactory dysfunction parkinson's disease - jnnp.bmj.com · olfactory dysfunction...

46ournal ofNeurology, Neurosurgery, and Psychiatry 1997;62:436-446

Olfactory dysfunction in Parkinson's disease

C H Hawkes, B C Shephard, S E Daniel

AbstractObjective-To evaluate olfactory functionin Parkinson's disease.Methods.-A standardised odour identifi-cation test was used, together with anevoked potential assessment with hydro-gen sulphide. In addition, histologicalanalysis was performed on the olfactorybulbs of cadavers who died fromParkinson's disease.Results-Over 70% of patients studied (71of 96) were outside the 95% limit of nor-mal on the identification test in an agematched sample and there was an unusualpattern of selective loss to certain odours,not hitherto described. The evoked poten-tials were significantly delayed but ofcomparable amplitude to a controlmatched population. Of the 73 patientsstudied only 37 had a technically satisfac-tory record containing a clear response toboth gases and of these, 12 were delayed.For H,S there was more delay on stimulat-ing the right nostril than the left. Somepatients with normal smell identificationtest scores had delayed evoked potentials.In the pathological examination of olfac-tory bulbs from eight brains, changescharacteristic of Parkinson's disease(Lewy bodies) were seen in every olfactorybulb, particularly in the anterior olfactorynucleus, and were sufficiently distinct toallow a presumptive diagnosis ofParkinson's disease.Conclusions-Olfactory damage inParkinson's disease is consistent andsevere and may provide an important clueto the aetiology ofthe disease.

Department of ClinicalNeurology, IpswichIP4 5PD, UKC H HawkesB C ShephardInstitute of Neurology,Queen Square, LondonWClN 3BG, UKC H HawkesDepartment ofNeuropathology,Parkinson's DiseaseSociety Brain Bank,Institute ofNeurology,1 Wakefield Street,London WClN 1PJ, UKS E DanielCorrespondence to:Dr C H Hawkes, 22 HenleyRd, Ipswich IPI 3SL, UK.Received 28 May 1996and in revised form21 October 1996Accepted 6 November 1996

(7 Neurol Neurosurg Psychiatry 1997;62:436-446)

Keywords: olfactory evoked potentials; smell identifica-tion; Parkinson's disease

There is now good evidence from a wide varietyof quantitative psychophysical tests that theability to smell is substantially affected inParkinson's disease. For example, Doty andcoworkers showed significantly impaired olfac-tion relative to matched controls in a group of81 patients with Parkinson's disease.' All hadnormal cognitive function on formal tests andwere assessed by smell identification andodour detection threshold tests. To date,olfactory evoked potential (OEP) techniques,which are not influenced by odour naming or

recall ability, have not been systematically

applied to patients with Parkinson's disease.Such techniques were pioneered by Kobal andPlattig2 and adopted by us.'The basal ganglia have been the subject of

intense pathological study in Parkinson's dis-ease, but the rhinencephalon has not beeninvestigated systematically. Chui et a14 exam-ined four patients with Parkinson's diseasewith dementia and in one brain found anAlzheimer type change in the amygdala, adja-cent anterior temporal cortex, and CA2 sectorof the hippocampus. The hippocampus wasnormal in the remaining three cases. It isuncertain whether all the central olfactoryareas were examined. Furthermore, the caseswere complicated by the presence of dementia.We have undertaken a multidisciplinary

study of olfactory identification, olfactoryevoked potential, and pathological examina-tion of the olfactory bulb in Parkinson's dis-ease.

MethodsAfter local ethics committee approval andinformed consent of patients and controls weundertook the procedures described below.

SMELL IDENTIFICATION TESTThese measurements were carried out usingthe University of Pennsylvania smell identifi-cation test (UPSIT).5 This test uses strips ofpaper impregnated with microencapsulatedodours which are released on scratching thestrip with a pencil. There are 40 differentodours and a forced choice is made from fourpossible answers.

ControlsNormative UPSIT data are available forAmericans but because some odours are unfa-miliar to British subjects we obtained our ownnormal values for 96 controls derived fromhealthy members of hospital and BritishTelecom staff. Before testing, all patients wereasked to estimate their sense of smell on a sim-ple six point scale (very good; good; average;below average; poor; very poor). The nose waschecked for patency and a questionnaire wasadministered to all subjects to probe for his-tory of nasal disease, head injury, use or mis-use of drugs, and other conditions (forexample, endocrine or hepatic disease) whichare occasionally associated with hyposmia.

PatientsPatients were obtained consecutively fromneurological inpatients and outpatients. All

436 on 9 A

pril 2019 by guest. Protected by copyright.

http://jnnp.bmj.com

/J N

eurol Neurosurg P

sychiatry: first published as 10.1136/jnnp.62.5.436 on 1 May 1997. D

ownloaded from

http://jnnp.bmj.com/


were examined neurologically at least once byone of us (CHH) and were considered to haveidiopathic Parkinson's disease although weare well aware of the roughly 25% fallibility ofsuch classification.6 All patients had intactnasal passages on routine bedside examina-tion and scored 27/30 or more on the minimental test. If questionnaire analysis sug-gested nasal disease or any other conditionthat might impair olfaction (for example, dia-betes, severe head trauma, alcoholism) theywere excluded. Despite these precautions weare aware that nasal disease may still be pre-sent even in apparently healthy people.7Nearly all the patients with Parkinson's dis-ease were receiving levodopa and selegilinebut these probably do not affect the ability tosmell.89 We made no specific tests for depres-sion. Six patients were taking long term tri-cyclic antidepressants in small doses and twowere on long term lithium carbonate for affec-tive disorder. Impairment of sense of smellhas not been found in patients with depres-sion (Amsterdam et al10 and our own find-ings). The effect of depression on olfactoryevoked potentials is not known but we suspectthat it is unimportant as long as the patient iscooperative.

OLFACTORY EVOKED POTENTIALSThe term olfactory evoked potential is used inthis article to refer to both H2S and CO2responses. CO2 has no odour and is a stimulantof trigeminal nerve endings in the nose.The olfactory stimulator we used is similar in

construction and design to that described byKobal" except that our solenoid valves have aslower response time (see below). The stimula-

tor overcomes the problem of inadvertenttrigeminal stimulation. Anosmic patients testedwith this device show no response to primarilyolfactory stimulants such as H,S (Kobel andHummel"2 and our own findings). Olfactorystimuli are embedded in an odourless bacterio-logically pure carrier gas. This is achieved bypassing compressed air through a series of fivefilters. The gas flows at 140 ml/s and is heatedand humidified to match the nasal environ-ment. The principle of stimulus generation is tohave two identical gas flows-only one ofwhichcontains the odorous substance. Either of theflows can be directed to the nose (fig 1). The airstream is delivered to the nose by means of ateflon tube with a nozzle (internal diameter4 mm) inserted about 1 cm into one nostril.Between stimuli, only clean air enters the nasalcavity and the air stream containing the odouris vented outside the recording room.Stimulation is carried out by switching the twoflows for a preset time using two valves (FestoLtd) with a fast response time. This eliminatesdetectable pressure change at the nose andtherefore avoids trigeminal activity due to pres-sure changes. The duration of stimulus con-taining the test gas is set to 200 ms in all cases.The characteristics of the pulse have beenascertained by a low flow meter (Si-Plan UKLtd). The rise time measured at the end of thenasal tubing is approximately 34 ms. This is acomposite figure derived from the switchingperiods of the valves (approximately 10 ms);travel along the insulated tubing from the noseto the intranasal cannula (approximately20 ms); and response time of the flow meteritself of 4 ms. The last figure when subtractedfrom 34 ms gives a true rise time of around

Airflow

Air i Filters Flow r Hcompressor '5 regulator

_ ffl~~~~~~~~~~~~'Airflow/ 1I

Flow~,regullator .| HumidifierregLiI

f fl:~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~t

Flow_.I regulatorAN

Gas cylinkias cy ill"Cee

Valve V2 7 fValve aJ1

SuCtion 2 Su]ction 1

--. ----BB_- N,ase

A

-J..

Heater

Teflor=== Nylor

Si. ico'l

Vent

Figure 1 Principle of olfactory stimulation. Here the olfactometer is in stimulus mode. Valve V2 is open and Vl is closed.Pure airflow (2) is diverted by suction 2 and the H2S/air mixture is delivered to the nose. During the interstimulus intervalVl is open while V2 is closed; the H2S/air mixture is diverted, and only airflow 2 reaches the nose. Before use theapparatus is calibrated so that no H,S enters the nose during the interstimulus period. To ensure that all airflow 2 isremoved by suction 2 when the olfactometer is in stimulus mode, the tubing is temporarily blocked at B and suction 2 isadjusted until theflow at the nose is zero. In interstimulus mode V2 is closed and Vl open; flow is blocked at A and suction1 is adjusted so thatflow at the nose is zero.

437

on 9 April 2019 by guest. P

rotected by copyright.http://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp.62.5.436 on 1 M

ay 1997. Dow

nloaded from


Hawkes, Shephard, Daniel

Figure 2 (A) CO2evoked responses in fourhealthy subjects. Ages andsex are given in the rightmargin. Traces werederivedfrom Al-CZ inresponse to a 200 mS pulseof50% CO2. Filters wereset at 1-50 Hz. Squaresrepresent 12 5 pVonvertical axis and 200 mSon horizontal axis. (B)H2S evoked responses infour healthy subjects.Parameters as for CO2except that responses werederivedfrom Al-PZ andamplitude on the verticalaxis is 625 gV. The 200mS pulse contained H2S ata concentration of20 ppm.

A

B

30 ms. Pressure change on switching the gasstreams is small and verified by the fact that noevoked potential has been detected by us inresponse to a pure air pulse. Objective measure-ment of the transient pressure change with theflow meter above is difficult because of therapidity of change at onset and offset of thepulse, but approximates to a difference of flowrate of 1 1/min. Subjects breathe through themouth to avoid variation in stimulus concentra-tion through nasal breathing. Breathing is mon-itored by a thermistor taped close to the mouthso that variation of intranasal gas concentrationdue to inadvertent nasal breathing can beavoided and such potentials rejected. The olfac-tory stimulant in the present study is H,S(20 ppm by volume). Although its odour issomewhat unpleasant, it is a primary olfactoryagent which does not elicit a trigeminalresponse in anosmic patients"3 The thresholdlimit volume is 10 ppm for continuous expo-sure in the workplace. As we use 20 ppm for 11stimuli (10 averages) of 200 ms duration anyirritant effects are most unlikely. Furthermore,all unwanted gases are removed from therecording environment by use of a powerfulextractor fan. This concentration is higher thanthat used routinely by Kobal1' (2ppm). Carbondioxide (50% by volume) is used in an identicalmanner. This gas, which is odourless and

slightly painful, allows strong nasal trigeminalstimulation and provides an evoked potential ofshorter latency than H,S. For H,S an interstim-ulus interval (ISI) of 60 seconds is used and forCO2 40 seconds.13The cerebral response is recorded at CZ and

PZ using the standard 10/20 system referencedto Al. The electro-oculogram is simultaneouslyaveraged and any trial contaminated by eyeblinks or other muscle artefacts is discarded.Typically a 2048 ms epoch is used ofwhich 600ms is presignal. Filters are set at 1-50 Hz andamplification at 10 MV/cm. Ten to 16 artefactfree signals are averaged on a Nihon-KohdenNeuropak-4 machine, which uses 256 datapoints per channel. The averaged trace issmoothed by one pass of a nine point filter(equivalent to a low pass filter of 27-5 Hz).Latency measurements are unaffected by thisfilter setting but there is slight reduction ofamplitude. Latencies are measured from thestart of the trigger pulse. The stimulus reachesthe nostril approximately 10 ms afterwards;because this interval is short by comparisonwith the initial Ni response, it is ignored inlatency calculations. A simple visual trackingtask is performed to maintain a constant stateof alertness. Subjects use a joystick to centre asmall square in a slowly moving large squaredisplayed on a 14 inch computer monitor at adistance of about 1 metre. Headphones deliver-ing white noise are worn to mask out auditoryclues from switching devices. Figure 2 shows asample of normal CO2 and H2S averagedresponses derived from four healthy subjects.The P1 and N2 responses are often unde-tectable and Ni is sometimes indistinct even inhealthy people. Latencies are measured to thefirst negative peak (Ni) and second positivetrough (P2). If the Ni response is not clearlyvisible, its position is estimated from the pointof downward (positive) deviation from baselineof the P2 response. Stimuli are presented uni-laterally; the most patent nostril is usuallyselected for testing but if results are obtainedfrom both nostrils then the mean latency andamplitude is used unless specified otherwise.Amplitude is measured from the peak of Ni tothe trough of P2.

ControlsWe used 47 controls (aged 17-74; 18 men, 29women; mean age 45-6 years) who hadresponses to both gases derived from the 96making up the UPSIT group. Only 71 of 96agreed to be tested and of these 24 had unreli-able tracings. The main reason was that sub-jects found that stimulation with 50% CO2 wastoo uncomfortable although they could toleratelower concentrations and showed a normalevoked response.

PatientsA total of 74 patients with Parkinson's diseasewere examined who were also extracted fromthe UPSIT group. For statistical analysis weused only those patients (37 altogether) with aclear response to both gases. There were 19men and 18 women aged 27-77 years, meanage 62 years.

438



j.com/

J Neurol N


ay 1997. Dow

nloaded from


Olfactoty dysfunction in Parkinson's disease

Figure 3 Logittransformation of UPSITscore plotted against age.0 Actual scores forcontrols; A actual scoresfor patients.

40 F- 0 0

0 0 00

- .0 0 000 0 0 00

o ControlsA Parkinson's disease

0

0 0

o ooo o

o oo o o o0 oo o 0 o o oA o o 0ob0 0

o ooo oo o o _ oo ooo o oooAvo~~ ~ ~~~~o oA

A* A Regression line= *~~~~'

A o o 'for controls) A

IAA 0 A

A

A

A AA AAA -AoA A

A A A AA AA A A

A A A AA A

A AA

A A

0

A

i Limit of 95%A prediction intervali for controlsAAAA 1--_= Regression lineAA A for Parkinson's

A A'A 'A'A

A

15 25 35 45 55 65 75 85Age (y)

OLFACTORY BULB PATHOLOGYOlfactory bulbs and tracts were removed fromformalin fixed half brains of eight patients witha clinical and pathological diagnosis ofParkinson's disease. All were taken from tissuein the United Kingdom Parkinson's DiseaseBrain Bank. None had olfactory testing by us.

There were four male and four female patientsaged 67-76 (mean 72-25 years) and threemale and five female controls aged 58-80years (mean age 74-13 years). Tissue wasprocessed in paraffin wax and serial horizontalsections of the entire nerve were cut at 7 u.

The first and every 10th section were stainedwith haematoxylin and eosin and the immedi-ately adjacent sections were stained with antis-era for ubiquitin (Dako, polyclonal 1: 150) andmodified Bielschowsky silver impregnation.Sections were examined "blind" to clinicalinformation. Cases were classified as "proba-ble Parkinson's disease" or "non-Parkinson'sdisease" on the basis of whether Lewy bodieswere seen, and results subsequently comparedwith the diagnosis after full neuropathologicalexamination.

ResultsSMELL IDENTIFICATION TESTS (uPsrr)Altogether we tested 96 patients withParkinson's disease (aged 27-81 years, 49 men,47 women; mean age 57 0 years) and 96 con-

trols (aged 18-78 years; 39 men, 57 women;mean age 41-7 years). For UPSIT scores theproportion of correct answers (p) was firsttransformed using the logit transformation,log(p/(I - p)), because p was skewed towards 1,especially for the controls. The UPSIT scoresfor patients with Parkinson's disease were sig-nificantly lower than those for controls(P < 0-0001). Only 26% (25 of 96) of the

patients with Parkinson's disease had a scorewithin the level expected for 95% of our healthycontrols (fig 3). There was no evidence of qua-dratic effects.We examined whether any of the 40 different

odours in the UPSIT gave greater difficulty inidentification for patients compared with con-trols. From 96 patients and 96 controls wematched 70 controls and 70 patients on thebasis of sex and age ± 3 years. There were 35 ofeach sex. The mean age of controls was 54-7years and of patients 55-4 years. We calculatedthe percentage of correct scores for each odourin the control group and the patient group anddetermined the difference between the two per-centages (fig 4). The vertical line is set at 31%,which is the average of all 40 differences in pro-portions. The confidence intervals are derivedfrom McNemar's test for differences betweentwo paired proportions with the confidencelevel adjusted to allow for the fact that 40 statis-tical tests were being carried out. If any of thesecorrected confidence intervals wholly exceededthe average difference of 31 %, the differencebetween control and patient responses on thisparticular component would be in excess of thenatural (random) component to componentvariation, which would be expected. The mostsignificant results are shown at the top, the leastsignificant at the bottom.

Figure 5 is a plot of sensitivity against speci-ficity for all 40 odours in the UPSIT. A combi-nation of pizza and wintergreen was the bestdiscriminator, with a sensitivity of 90% andspecificity of 86%. Inclusion of a third odourdid not improve the separation of controls frompatients. Because subjects made a forced choiceof one answer from four possibilities, 25% ofanswers will be correct by chance alone. Hencethe sensitivity value (76%) for pizza (percentageof patients scoring incorrectly), in the presence

c)0Co* X0)0

° 35

EEx 30

Ea, 25o0X 20

CI)OL 15

10

5

439



j.com/

J Neurol N


ay 1997. Dow

nloaded from



Figure 4 Results ofUPSIT test to all 40odours compared withcontrols. The differencesare shown, with associated95% confidence intervals,between the percentages ofcontrols and patients withcorrect answers for each ofthe 40 individualcomponents of the UPSIT.The vertical line is set at31%, which is the averageof all 40 differences inproportions. The mostsignificant results areshown at the top, the leastsignificant at the bottom.See textforfullerexplanation of method.

-

I__

I_

I_

I_I_I_I_I_I_I_I_I_

I~~~~~~

i~~~~~~~I~~~~~~

---I1-H--

- Hi

i ---i~~~~~~~~~~~~~~~~~~~~~~

- --30% -20% -10% 0% 10% 20% 30% 40% 50%

Control % - PD % (for each test)60% 70% 80% 90%

of 90% specificity (percentage of controls scor-ing correctly) for this odour indicates anosmiato pizza for patients with Parkinson's disease.

There was no correlation between durationof disease and UPSIT score (r = 0 074).Patients were asked for a subjective estimate oftheir sense of smell and this correlated broadlywith their UPSIT score (r = 0 47, P < 0 001).

EVOKED POTENTIALSA response was considered absent if no visible

Figure 5 Plot ofsensitivity (percentages ofcases answeringincorrectly) againstspecificity (percentage ofcontrols answeringcorrectly) for each of the 40test odours in UPSIT. Thespecificity axis is reversedto give a graph similar tothe receiver operatingcharacteristic curve. A"perfect" discriminatorytest will have a sensitivityand specificity of 100%and will be located in thetop left corner of the graph.A poor test will be locatedin the bottom right corner.Discriminant analysissuggests that pizza andwintergreen in combinationprovide the bestdiscrimination with asensitivity of 90% andspecificity of 86%.

100

90

80

70

60

50

40

30

- Pizza + WintergreenA

= Pizza0~~~

.0-Wi ntergreen *0

/ 0.

-Clove* Tu

600 0

*4- 0*#

0 e

irpentine

tnion

20

l0tL

100 90 80 70 60 50 40 30 20

Specificity

waveform was detectable in an artefact freerecording. Forty seven controls were used forboth gases as described in the methods section.

Carbon dioxide (50% concentration volumelair inall cases)Sixty six patients with Parkinson's diseasewere tested initially but six were rejected: fivehad unclear recordings and one was absent.This left 60 patients and there were only 37who responded to both CO2 and H2S. Figure6A shows the olfactory evoked response at CZdue to stimulation by CO2 for a selection ofpatients with Parkinson's disease.

Hydrogen sulphide (20 ppm in all cases)There were 73 patients with Parkinson's dis-ease tested in whom 21 responses were absent,and six unclear. This left 46 results with 37patients who had clear responses to both CO2and H2S. Figure 6B shows the olfactory evokedresponse at PZ due to stimulation by H2S for aselection of patients with Parkinson's disease.The Ni and P2 distributions were normal

but the amplitude distribution was skewed. Itwas decided to use loge transformation on allthree variables as this not only normalised theamplitude distribution but stabilised the vari-ances between the two groups for Ni and P2.For each transformed variable, regressionanalysis was carried out adjusting for age tosee if any significance could be attributed todifferences due to diagnosis-that is, whetherParkinson's disease or control. For H2S (butnot CO2) a highly significant difference existed

PizzaWintergreenCloveLemonBananaBubble gumLimeCoconutGrassCinnamonRosePaint thinnerCherryRoot beerCedarGingerbreadLicoriceCheeseGrapeMentholOrangePineapplePineDill pickleGasolineWatermelonStrawberryMintMotor oilChocolateSoapLilacPeanutPeachNatural gasLeatherSmokeTurpentineFruit punchOnion

:LI

ina)U)

0 .

440



j.com/

J Neurol N


ay 1997. Dow

nloaded from



Figure 6 (A) CO,evoked responses in fourpatients with Parkinson'sdisease; ages and sexes aregiven in the right margin.Traces were derivedfromAl-CZ in response to a200 mS pulse of50%C02. Filters were set at1-50 Hz. Squaresrepresent 12-5 MVonvertical axis and 200 mSon horizontal axis. (B)H2S evoked responses infour patients withParkinson's disease.Parameters as for C02except that responses werederivedfrom A1-PZ andamplitude on the verticalaxis was 6-25 ,uV. The200 mS pulse containedH,S at a concentration of20 ppm.

A

M 70

M 75

Table 1 Differences in regression lines for CO2 and H,Slatency and amplitude between patients with Parkinson'sdisease (PD) and controls

Difference(PD-control) P value

CO,:N1 9 ms 0-26P2 32ms 0-88Amp -7-6 1iV 0-07

H2S:Ni 64ms 0-96P2 100 ms 0.01*Amp 1 7uV 0 5

F 57 Results were obtained by regressing latency/amplitude on agefor each year. Amp = amplitude.*Significant.

Table 2 Parkinson's disease (PD) - control differencesF 72 for H2S between left and right nostril regression lines after

allowingfor age

M 43

M 67

F 59

F 71

between diagnostic groups for P2 latency butnot NI latency or N1-P2 amplitude (table 1).We were able to test 10 patients with nor-

mal UPSIT scores using olfactory evokedpotentials. These patients therefore scoredwithin the 95% age adjusted limits of ourhealthy control group. One of the 10 patientshad absent H2S responses and three had sig-nificantly prolonged latencies after stimulationof one or both nostrils with H2S. All had intactCO2 responses.

There were 13 patients with Parkinson'sdisease and nine unmatched controls who hadall shown clear responses from both nostrils.The nostril first stimulated was randomlyassigned to either side (that is, right then leftor left then right). Regression analyses wereundertaken separately for Ni and P2 latencyand N1-P2 amplitude to see if the slopes ofthe latency and amplitude against age differedsignificantly between patients and controls. Inno instance was there a difference in slopesbetween patients and controls. The differencebetween these regression lines was tested forany disparity in the magnitude of latencies oramplitudes between patients and controls afterallowing for age (table 2). Table 2 shows thatH2S signals for stimulation of the right nostril inthe Parkinson's disease group are significantly

Difference(PD-control) P value

Left:N1 30ms 0 53P2 11 ms 0 79Amp -1 6,uV 0 34

Right:Ni 140 ms 0.001*P2 124 ms 0.0001*Amp 23 /uV 0.91

Data are for both nostrils from 13 patients with Parkinson'sdisease (mean age 57 years) and nine controls (mean age51 years). Amp = amplitude.*Significant.

Table 3 Differences between left and right nostril Nl andP2 latencies and N1-P2 amplitude (H,S) for nine controlsand 13 patients

Difference P value(left-right) (paired t test)

PD:Ni -54 ms 0-21P2 -38 ms 0 28Amp 044,V 0 79

Control:Ni 58ms 019P2 82 ms 0.04*Amp 2-3 MV 0-27

Mean ages of patients with Parkinson's disease was 57 yearsand that of the controls was 56 years. Amp = amplitude.*Significant.

delayed in comparison with the controls. Thedifferences in latency and amplitude betweenleft and right nostrils were tested using apaired t test separately for patients and con-trols. Within the Parkinson's disease or controlgroups there were no significant differencesexcept for slight prolongation of P2 latency incontrols after left nostril stimulation comparedwith the right (table 3).

For CO2 and H2S in controls there was sig-nificant correlation of UPSIT score with Ni-P2 amplitude (P < 0006, r = 037 for CO2;P < 0-001, r = 0-44 for H2S) and for latency(P < 0-05, r = -0-29 for CO2 to P2 only;P < 0-05; r = -0-25 for H2S to NI only). ForCO2 in patients there was no correlation ofUPSIT score with latency or amplitude mea-surements. For H2S in patients there wascorrelation of Ni and P2 latency (P < 0-001r = -0 54; P < 0-001, r = -0-62 respec-tively) but not N1-P2 amplitude. (Spearman'srank correlation coefficient for all analyses).

PATHOLOGICAL STUDYBy examining the olfactory bulb and tract all

441



j.com/

J Neurol N


ay 1997. Dow

nloaded from



Figure 7 (a) Theanterior olfactory nucleusin Parkinson's disease.Intraneuronal Lewy bodies(some arrowed) are easilyidentified using ubiquitinimmunocytochemistry.Occasional neurites(arrowheads) are alsoimmunoreactive. Avidin-biotin-peroxidaselhaematoxylin, bar =50,u. (b) Nerve cells of theanterior olfactory nucleuscontain Lewy bodies(arrows) which arehomogeneous hyalineinclusions resembling thosefound in the cerebralcortex. Haematoxylin-eosin, bar = 25g4. (c)Occasional Lewy bodiesin the olfactory bulb showa classic "cored"appearance.Haematoxylin-eosin, bar= lop.

e :.

;e: t.;h: : . : . : . . . .:!: .:

6 !. Yim_o ,+ 1Ex._nX. '.P.:.}: . X

*. :. _

eight cases were correctly diagnosed "probable

Parkinson's disease."

There were no Lewy bodies in any control

sample. Lewy bodies were best identified

using ubiquitin immunocytochemistry (fig 7a)

and were most numerous in the anterior olfac-

tory nucleus; they were also found occasion-

ally in mitral cells. The morphology of Lewy

bodies at this site resembled their conticalcounterparts (fig 7b) and inclusions showing a

classic trilaminar structure were rare (fig 7c).

In two cases Lewy bodies were plentiful and

...

associated with distended ubiquitin immuno-reactive neurites. In one case there were alsomany neurofibrillary tangles and a few senileplaques involving the anterior olfactorynucleus in a patient who had Alzheimer's dis-ease. Full histological examination of thebrains confirmed the diagnosis of Parkinson'sdisease. In the two brains with numerous olfac-tory Lewy bodies and swollen ubiquitinimmunoreactive neurites, contical Lewy bodiesnotably in the anterior cingulate gymus andparahippocampus, were also plentiful. Many

442

'A

1

I

,`.ss.

AF

Alr

...AI

.f....



j.com/

J Neurol N


ay 1997. Dow

nloaded from



distended neurites in the CA2 region of thehippocampus said to be characteristic for dif-fuse Lewy body disease were also identified.Age related degenerative changes with numer-ous corpora amylacea were identified in allbulbs; in one elderly control there were occa-sional neurofibrillary tangles in the anteriorolfactory neurons.

DiscussionOur data have substantiated earlier reports ofabnormalities in smell identification inParkinson's disease and additionally showabnormal olfactory evoked potentials and spe-cific pathological changes in the olfactory bulb.

SMELL IDENTIFICATIONThe UPSIT data for Parkinson's disease aresimilar to those described by Doty and col-leagues' except that our series is slightly largerand contains more young patients-down to27 years. These authors suggested that olfac-tory dysfunction was unrelated to odour type,did not depend on disease duration, and didnot correlate with motor function, tremor, orcognition, as was also found by others.9 Dotyet a18 also showed that the deficit was of thesame magnitude in both nostrils, and was notinfluenced by antiParkinsonian medication. Asreported by Doty et al' we also found no corre-lation of UPSIT with duration of disease.Whereas there was clearly a backgrounddepression of olfactory identification, superim-posed on this was a degree of selective odourdeficit-something that was not documentedby Doty's group. Odours that were most read-ily misidentified were lemon, pizza, winter-green, rose, and clove. Pizza was the bestsingle discriminant odour with pizza and win-tergreen in combination better still. Thusa subject could be suspected of havingParkinson's disease if both pizza and winter-green were inaccurately identified and wouldprobably not have the disease if both of theseodours were positively smelt correctly (not byrandom guessing). Because subjects made aforced choice of one from four possibilities,25% will be correct by chance alone. Hence asensitivity value of 76% for pizza in the pres-ence of 90% specificity indicates anosmia topizza for patients with Parkinson's disease. Itis difficult to explain why this should be.Controls and patients were matched for ageand sex as these variables influence the UPSITscore in Parkinson's disease.'4 We did not havesufficient numbers to match for other factorswhich might influence the result, such as intel-ligence or social class; nor were we able toallow for variation in odour intensity due toerrors in manufacture. Although lemon, win-tergreen, and clove have some trigeminal stim-ulation, pizza (oregano smell) and certainlyrose are mainly olfactory. Whatever the mech-anism, these findings raise the possibility thatthere may be a congenital or acquired selectivehyposmia in Parkinson's disease comparablewith androstenone smell blindness whichaffects 20%-47% of healthy people.'5 Of par-ticular relevance is a study in which rats were

exposed to 44 inhaled vapours for severalweeks.'6 For each odour there was a selectivepattern of mitral cell loss in the olfactory bulbwhich was unrelated to odour concentration.It is thus conceivable that the olfactory dam-age in Parkinson's disease may be caused byexposure to a neurotoxic vapour which selec-tively injures part of the olfactory system.Speculatively it should be possible to "workback" from the olfactory bulb pathology anddescribe the chemical characteristics of a possi-ble neurotoxin.

It was noted' that only 28% of patients withParkinson's disease were aware of some olfac-tory or taste impediment on a simple yes or norating scale before formal testing. We foundthat patients had a reasonable idea of theirsmell sense as judged by a six point ratingscale compared with UPSIT score. These twofindings are comparable although differenttechniques of analysis were used. We did notmeasure motor disability on a rating scale suchas Hoehn and Yahr as it is a relatively crudeassessment and earlier findings had impliedthat there was no correlation between motorfunction and smell defect.' 9

EVOKED POTENTIALSOur main finding has been an absent ordelayed H,S evoked response in a few patientswith Parkinson's disease. The latency ratherthan amplitude abnormality is an unexpectedfinding of possible diagnostic value. Theincrease in H,S latency was only apparent toP2 not N1. This suggests that the main changeis a reduction of N1-P2 gradient causing abroader wave shape. In our earlier study ofmultiple sclerosis there was significant changein latency.' Because Parkinson's disease is not ademyelinating condition we did not expect tofind a latency change and can offer no expla-nation for this at present. Our patient and con-trol groups are not well matched for age. Asthere are more elderly patients than controlsthis might inflate any difference between thetwo groups, especially in the older age range.However, age was allowed for in the regressionanalysis. The large number of absentresponses (27 of 73) was partly due to technicaldifficulty in obtaining the olfactory evokedpotential, which is considerably greater thancomparable procedures such as visual, audi-tory, and somatosensory responses. Apartfrom the need for repeated checking of the sig-nal it is possible to average 20 signals at mostbecause of the necessarily prolonged interstim-ulus interval (60 seconds) required. It is notsurprising that so many were absent as thepatients with Parkinson's disease had a poorsense of smell generally and the few delayedH,S responses in those we could test may be areflection of a healthier population in theolfactory sense. Sometimes technical difficul-ties brought the recording session to a prema-ture halt or a problem was experienced withinvoluntary movement, making it difficult forthe patient to keep the tubing in the nose.Even in healthy controls six of 69 (7%) H,Srecordings were unsuccessful. For CO,,responses were unobtainable in six of 66 (9%)

443



j.com/

J Neurol N


ay 1997. Dow

nloaded from



patients and six of 56 (9%) controls. Thenumber of healthy controls used was reducedto 47 as we accepted only those with responsesto both gases. All controls who were rejectedtolerated 40% CO, but they had to beexcluded because of intolerance to 50% CO,.The UPSIT is the superior test with more than70% abnormality in Parkinson's disease. Thiscontrasts with 12 of 37 (32%) delayedresponses to H,S in patients. A record which islabelled absent may in fact contain a responsewhich is below the limit of detection imposedby averaging 16 signals. We have used onlyone odour whereas the UPSIT implements 40.If three or four different gases were used thesensitivity of olfactory evoked potentials mightwell increase.The correlation of UPSIT score with

latency and amplitude measurements in con-trols for H2S is to be expected as identificationmight well be easier with a more vigorous cere-bral signal. This association does not holdgood for patients exposed to H2S in whom thecorrelation is strong with latency but notamplitude.

In one study'3 eight patients withParkinson's disease were compared with agematched controls using a simple odour identifi-cation test and by measuring the olfactoryevoked potential after stimulation with vanillinand H2S. The olfactory evoked potentials toH,S and vanillin were of longer latency in theParkinson's disease group although these wereonly significant on left nostril stimulation;amplitudes did not differ. Our data confirmthis pilot study in that we discovered signifi-cant prolongation of latency to P2 in theParkinson's disease group but we foundincreased latency (without amplitude change)compared with controls for H2S when the rightnostril was stimulated compared with the left.In keeping with their studies in healthy sub-jects'3 we found more delay on right nostrilstimulation for H2S. It is not clear why there issuch a discrepancy but in both instances thenumbers studied were small and our resultsshould only be regarded as provisional.Furthermore we necessarily selected thosewith an H2S response which was not too diffi-cult to elicit. Ideally a right, left, left, right orleft, right, right, left nostril stimulationsequence should be used but it is unlikely thatpatients would tolerate such a lengthy test pro-cedure.Of particular interest is our finding of

abnormal H,S responses in four of 10 patients(one absent, three delayed) with normalUPSIT score. Viewed alone the UPSIT resultsplace 74% of patients outside the 95% limit forcontrols. When this information is combinedwith olfactory evoked potential data thefraction of olfactory abnormality exceeds80%. Such frequency is greater than that oftremor, which is usually quoted at 70%'7 andnearly equals that of rigidity and akinesia. Atleast for a hospital (rather than communitypopulation) olfactory dysfunction would seemto be as common and presumably as importantas the cardinal motor signs of Parkinson'sdisease.

OLFACTORY BULB PATHOLOGYThe finding of Lewy body damage in theolfactory bulb was first reported briefly by usearlier.'8 In every instance the Parkinson's dis-ease specimens were readily distinguishedfrom controls. Furthermore in two cases withsevere involvement of olfactory bulbs the cere-bral cortex was similarly affected and corticalLewy bodies were numerous. In one case therewas coexistent Alzheimer's disease with neu-rofibrillary tangles and senile plaques presentin the neocortex and olfactory bulb-asalready documented.19-21 In our Parkinson'sdisease series, Lewy bodies were most readilyidentified in the anterior olfactory nucleus.Detailed topographical analysis and cell count-ing will be required to determine whetherthese neurons are selectively vulnerable or ifadditional neuronal types-for example, mitraland granule cells-are similarly affected. Ourfindings suggest that inspection of the olfac-tory bulb allows preliminary diagnosis ofParkinson's disease. An initial diagnosis ofother neurodegenerative diseases may also bemade in this way-for example, Alzheimer'sdisease, Pick's disease (Yoshimura2 and per-sonal observations) and multiple system atro-phy,'8 a finding which may be particularlyuseful before whole brain postmortem studiesare undertaken. The importance of these mor-phological changes to the smell defect inParkinson's disease and the extent of centralolfactory involvement remain to be elucidated.Our data show that olfaction is discretely

impaired in 70%-90% of patients withParkinson's disease and that damage is presentin the initial part of the olfactory pathway-namely, the bulb. Why should a movementdisorder be so consistently associated withhyposmia? The olfactory identification defectin Alzheimer' disease is said to be identical toParkinson's disease8; there is profound distur-bance of cognition in Alzheimer's disease butonly in a minor proportion of patients withParkinson's disease. One possibility is thatParkinson's disease and perhaps Alzheimer'sdisease might be caused by a virus or chemicalagent that gains entry to the CNS via the nose.One group22 were able to show that HSV1virus placed intranasally in six week old micecould be detected in the brain stem trigeminalroot entry zone and the olfactory bulbs fourdays later. In some mice, viruses which hadentered the olfactory bulb had spread as far asthe temporal lobe, hippocampus, and cingu-late cortex. Another23 showed that horseradishperoxidase applied intranasally was trans-ported to the bulb, anterior olfactory nuclei,and transmitter specific projection neuronsfrom the diagonal band (cholinergic), raphe(serotonergic), and locus coeruleus (noradren-ergic). All these zones are known to projectwidely to non-olfactory areas of the CNS.Because of the ease with which a large macro-molecule such as horseradish peroxidase couldspread it was reasoned23 that environmentalcontaminants and perhaps drugs of misusecould gain relatively unimpeded access to theentire nervous system. A further possibility isthe effect of inhaled solvents such as

444



j.com/

J Neurol N


ay 1997. Dow

nloaded from



trichloroethylene, which may rekindle activityin latent viruses such as herpes simplex.24Support for the nasal entry theory might bestrengthened if there was a correlationbetween the olfactory and motor deficits-either in terms of severity or duration of symp-toms. However, this has not been found.' Atpresent there is no convincing evidence thatsubjects with anosmia later developParkinson's disease. It could be argued thatParkinson's disease is caused by a single majorCNS insult and that any progression is simplydue to age related neuronal attrition of thesubstantia nigra.2' The olfactory defect mightnot progress because of the regenerativecapacity of nasal olfactory neurons.An alternative hypothesis could be con-

structed around the finding that some patientswith Parkinson's disease exhibit a defect in theP-450 cytochrome CYP2D6-debrisoquinehydroxylase gene.'6 Mammalian P-450 depen-dent oxygenases provide a central line ofdefence against exogenous toxins and it hasbeen shown that the risk of Parkinson's diseaseis more than doubled for those with a P-450genetic polymorphism associated with defi-cient debrisoquine metabolism. The high con-centration of P-450 in hepatic microsomes iswell known and it has been shown that micro-somes in the olfactory epithelium of, forexample, the rat27 and rabbit28 have highconcentrations of P-450, sometimes in excessof those in the liver depending on the particularsubtype. The olfactory bulb of the monkey hasparticularly high concentrations of P-450 com-pared with other parts of its brain. There seemto be variable concentrations of P-450 in therest of the CNS but most areas are incom-pletely characterised. Numerous volatile com-pounds are readily absorbed by the nasalmucosa.29 Many non-volatile materials such asenvironmental pollutants can reach the nasalmucosa, probably by adsorption on to smallparticles in the air that deposit on the nasalepithelium.'0 Compounds absorbed by thenasal mucosa are actively metabolised in situ,sometimes detoxified, or they may be activatedto become more toxic or carcinogenic.3' 32Compounds that have been shown to bemetabolised in vitro by the nasal P-450 depen-dent monooxygenase system include nasaldecongestants, essences, anaesthetics, alco-hols, nicotine, cocaine, and many nasal car-cinogens.3 It has been shown that herbicidessuch as dioxins34 or chlorthiamid,'35 whethergiven intravenously or intraperitoneally, areselectively taken up by and harmful to theolfactory epithelium. Some compounds seemto be selectively toxic to the olfactory epithe-lium whereas others cause damage indirectlyby formation of toxic byproducts. The CNSdistribution of CYP2D6 is not known but if itwere to be concentrated in the basal gangliaand olfactory pathways the regular appearanceof pathology at these sites might be explained.We conclude that patients with Parkinson's

disease have a severe, apparently stable defectof smell sense which is as frequent as tremorand almost as common as bradykinesia. Whysuch disparate items such as movement disor-

der and anosmia should occur together is per-plexing. It may relate either to simple damageby an agent that enters the CNS via the noseor there may be a defective P-450 cytochromesystem common to the olfactory and basalganglia pathways. These two hypotheses arenot necessarily mutually exclusive.

SED is funded by a grant from the United KingdomParkinson's Disease Society. We are grateful to Peter Sacaresand Ann Petruckevitch, Institute of Neurology, for statisticalsupport.

1 Doty RL, Deems DA, Stellar S. Olfactory dysfunction inparkinsonism: a general deficit unrelated to neurologicsigns, disease stage or disease duration. Neurology 1988;38:1237-44.

2 Kobal G, Plattig KH. Methodische Anmerkungen zurGewinnung olfaktorischer EEG-Antworten des wachenMenschen (objektive Olfaktometrie). ZEEG-EMG 1978;9:135-45.

3 Hawkes CH, Kobal G, Shephard B. The olfactory evokedresponse and evidence of disordered olfaction in multiplesclerosis. J Neurol Neurosurg Psychiatry 1992;55:243.

4 Chui HC, Mortimer JA, Slager U, Zarow C, Bondareff W,Webster DD. Pathological correlates of dementia inParkinson's disease. Arch Neurol 1986;43:991-5.

5 Doty RL, Shaman P, Dann M. Development of theUniversity of Pennsylvania smell identification test: astandardised microencapsulated test of olfactory func-tion. PhysiolBehav 1984;32:489-502.

6 Hughes AJ, Daniel SE, Kilford L, Lees AJ. Accuracy ofclinical diagnosis of idiopathic Parkinson's disease: a clin-ico-pathological study of 100 cases. J Neurol NeurosurgPsychiatry 1992;55:181-4.

7 Levine HL. The office diagnosis of nasal and sinus disor-ders using rigid nasal endoscopy. Otolaryngol Head NeckSurg 1990;102:370-3.

8 Doty RL, Stern MB, Pfeiffer C, Gollomp SM, Hurtig HI.Bilateral olfactory dysfunction in early stage treated anduntreated idiopathic Parkinson's disease. J7 NeurolNeurosurg Psychiatry 1992;55: 138-42.

9 Quinn NP, Rossor MN, Marsden CD. Olfactory thresholdin Parkinson's disease. J Neurol Neurosurg Psychiatry1987;50:88-9.

10 Amsterdam JD, Gregg Settle R, Doty RI, Abelman E,Winookur A. Taste and smell perception in depression.Biol Psychiatry 1987;22: 1481-5.

11 Kobal G. A new method for determination of the olfactoryand the trigeminal nerve's dysfunction: olfactory (OEP)and chemical somatosensory (CSEP) evoked potentials.In: A Rothenberger, ed. Event related potentials in children.Amsterdam: Elsevier, 1982:455-61.

12 Kobal G, Hummel G. Cerebral chemosensory evokedpotentials elicited by chemical stimulation of the humanolfactory and respiratory nasal mucosa. ElectroencephalogrClin Neurophysiol 1988;71:241-50.

13 Kobal G, Hummel T. Olfactory evoked potentials inhumans. In: Getchell TV, Doty RL, Bartoshuk, Snow JB,eds. Smell and taste in health and disease. New York:Raven Press, 1991;13:269-70.

14 Doty RL, Bromley SM, Stem MB. Olfactory testing as anaid in the diagnosis of Parkinson's disease: developmentof optimal discrimination criteria. Neurodegeneration1995;4:93-7.

15 Amoore JE, Pelosi P, Forrester U. Specific anosmias to 5-androst-16-en-3-one and -pentadecalactone: the urinousand musky primary odours. Chemical Senses and Flavor1977;2:401-25.

16 Pinching AJ, Doving KB. Selective degeneration in the ratolfactory bulb following exposure to different odours.Brain Res 1974;82:195-204.

17 Hoehn M, Yahr MD. Parkinsonism: onset, progression andmortality. Neurology 1967;17:427-42.

18 Daniel SE, Hawkes CH. Preliminary diagnosis ofParkinson's disease using olfactory bulb pathology [let-ter]. Lancet 1992;340:186.

19 Esiri MM, Wilcock GK. The olfactory bulbs in Alzheimer'sdisease. J Neurol Neurosurg Psychiatry 1984;47:56-60.

20 Mann DMA, Tucker CM, Yates PO. Alzheimer's disease:an olfactory connection? Mech Ageing Dev 1988;42:1-15.

21 Yoshimura N. Olfactory bulb involvement in Pick's dis-ease. Acta Neuropathologica 1988;77:202-5.

22 Tomlinson AH, Esiri MM. Herpes simplex encephalitis:immnuno-histological demonstration of spread of virus viaolfactory pathways in mice. _J Neurol Sci 1983;60:473-84.

23 Shipley MT. Transport of molecules from nose to brain:transneuronal anterograde and retrograde labelling in therat olfactory system by wheatgerm agglutin-horseradishperoxidase applied to the nasal epithelium. Brain Res Bull1985;15:129-42.

24 Buxton PH, Hayward M. Polyneuritis cranialis associatedwith industrial trichloroethylene exposure. J NeurolNeurosurg Psychiatry 1967;30:51 1-8.

25 Caine DB, Eisen A, McGeer E, Spencer PS. Alzheimer'sdisease, Parkinson's disease and motor neurone disease;abiotrophic interaction between ageing and environment?Lancer 1 986;ii: 1067-70

445



j.com/

J Neurol N


ay 1997. Dow

nloaded from



26 Smith CAD, Gough AC, Leigh PN, Summers B, et al.Debrisoquine hydroxylase gene polymorphism and sus-ceptibility to Parkinson' disease. Lancet 1992;339:1375-7.

27 Hext PM, Lock EA. The accumulation and metabolism of3-trifluoromethylpyridine by rat olfactory and hepatic tis-sues. Toxicology 1992;72:61-75.

28 Ding X, Coon MJ. Immunochemical characterisation ofmultiple forms of cytochrome P-450 in rabbit nasalmicrosomes and evidence for tissue specific expression ofP-450s NMa and NMb. Cellular Pharmacology 1990;37:489-96.

29 Stott WT, Ramsey JC, McKenna MJ. Absorption of chemicalvapours by the upper respiratory tract of rats. In: CSBarrow, ed. Toxicology of the nasal passages. Washington,DC: Hemisphere Publishing Corporation, 1986:191-210.

30 Schlesinger RB. Comparative deposition of inhaledaerosols in experimental animals and humans: a review.Toxicol Environ Health 1985;15:197-214.

31 Bond JA. Bioactivation and biotransformation of xenobi-

otics in rat nasal tissue. In: CS Barrow, ed. Toxicology ofthenasal passages. Washington DC: Hemisphere PublishingCorporation, 1986:249-61.

32 Dahl AR. Possible consequences of cytochrome P-450dependent monooxygenases in nasal tissues. In: CSBarrow, ed. Toxicology of the nasal passages. WashingtonDC: Hemisphere Publishing Corporation, 1986:263-73.

33 Dahl AR. The effect of cytochrome P-450-dependentmetabolism and other enzyme activities in olfaction. In:Margolis FL, Getchell TV, eds. Molecular neurobiology ofthe olfactory system. New York: Plenum Press, 1988:51-70.

34 Gillner M, Brittebo EB, Brandt I, Soderkvist P, AppelgrenL-E, Gustafsson J-A. Uptake and specific binding of2,3,7,8-tetrachlorodibenzo-p-dioxin in the olfactorymucosa of mice and rats. Cancer Res 1987;47;4150-9.

35 Brittebo EB, Eriksson VF, Bakke J and Brandt I. Toxicityof 2,6-dichlorothiobenzamide (chlorthiamid) and 2,6-dichlorobenzamide in the olfactory nasal mucosa of mice.Fundam Appl Toxicol 1991;17:92-102.

446



j.com/

J Neurol N


ay 1997. Dow

nloaded from


olfactory dysfunction parkinson's disease - jnnp.bmj.com · olfactory dysfunction...

Documents