Chest radiography cannot predict diaphragmfunction

Alfredo Chettaa,b,*, Amer K. Rehmana, John Moxhamc, Denis H. Carra,Michael I. Polkeya

aRespiratory Muscle Laboratory and Radiology Department, Royal Brompton Hospital, London, UKbClinical Sciences Department, Section of Respiratory Diseases, University of Parma, ItalycRespiratory Muscle Laboratory, Kings College Hospital, London, UK

Received 23 January 2004; accepted 28 April 2004

Summary The finding of hemidiaphragm elevation on a chest radiograph, in absenceof an ipsilateral lung disease, is assumed to indicate severe hemidiaphragmdysfunction. To test this hypothesis we retrospectively reviewed chest radiographfindings and corresponding twitch transdiaphragmatic pressure (TWPDI) results from42 (17 female, age range 22–79 years) consecutive patients who underwent phrenicnerve stimulation studies. Chest radiographs were independently reviewed in a blindmanner by two radiologists. The interobserver agreement was moderate, the kappavalue ranging from 0.48 (left hemidiaphragm) to 0.59 (lung parenchyma).Hemidiaphragm dysfunction was diagnosed if TWPDI of corresponding hemidiaphragmwas less than 3.5 cm H2O. The prevalence of patients with an elevated unilateralhemidiaphragm on chest radiograph was 64% and of patients with unilateral paralysisjudged by TWPDI was 24%. Sensitivity, specificity, positive and negative predictivevalues for chest radiograph, as a diagnostic test for unilateral diaphragm dysfunctionwere 0.90, 0.44, 0.33 and 0.93, respectively. We conclude that the isolatedelevation of hemidiaphragm on chest radiograph is of little value in the diagnosis ofunilateral hemidiaphragm paralysis, though the condition is unlikely if diaphragmelevation is absent.& 2004 Elsevier Ltd. All rights reserved.

Introduction

The finding of a unilateral elevation of thehemidiaphragm on a chest radiograph, in absenceof an ipsilateral decrease in lung size, i.e. lobarcollapse, surgical resection or fibrosis, is assumed

to indicate severe dysfunction or paralysis of thehemidiaphragm.1–3 Classical radiologic features ofunilateral hemidiaphragm paralysis are consideredpresent when the elevated hemidiaphragm showsan accentuated dome configuration in posteroan-terior and lateral projections with the costophrenicand costovertebral angles tending to be deepened,narrowed and sharpened.1

Paralysis of a hemidiaphragm results from inter-ruption of transmission of the phrenic nerve andcan be due to infectious, iatrogenic or malignantcauses, although the most common is paralysis of

ARTICLE IN PRESS

KEYWORDS

Diaphragm;

Radiology;

Magnetic stimulation

*Corresponding author. U.O. Clinica Pneumologica, Diparmen-to di Scienze Cliniche, Azienda Ospedaliera e Universit!a diParma, Padiglione Rasori, Viale G. Rasori, 10-43100 Parma, Italy.Tel.: þ 39-0521-703-475; fax: þ 39-0521-292-615.E-mail address: chetta@unipr.it (A. Chetta).

0954-6111/$ - see front matter & 2004 Elsevier Ltd. All rights reserved.doi:10.1016/j.rmed.2004.04.016

Respiratory Medicine (2005) 99, 39–44

unknown aetiology.1 Definitive diagnosis of phrenicnerve dysfunction can be obtained by the phrenicnerve electric stimulation combined with themeasurement and twitch transdiaphragmatic pres-sure (TWPDI).

4,5 Recently the electrical stimulationtechnique has been successfully substituted bymagnetic stimulation of the phrenic nerve stimula-tion, which avoids the use of inaccurate surfaceelectrodes or painful needle electrodes.6,7

Early studies investigated the value of the chestradiograph, as a screening test of unilateralhemidiaphragm paralysis,8–10 however up to nowno study has related chest radiograph findings todiaphragm function as judged by phrenic nervestimulation. Therefore to test the sensitivity,specificity, and predictive values for chest radio-graph to diagnose severe hemidiaphragm dysfunc-tion, defined by TWPDI, we reviewed the data fromconsecutive patients who had undergone in ourlaboratory full respiratory muscle assessment overa 4-year period.

Methods

Subjects

We retrospectively reviewed the data from 112consecutive patients who underwent full respira-tory muscle assessment from January 1999 toDecember 2002. The respiratory muscle studiesand chest radiographs had been undertaken forclinical reasons at the request of the patient’sclinician and approval to report these data hasbeen given by the chair of our ethical review board.

Patients were only included if: (1) they had noother cause for hemi-diaphragm elevation, such asfibrosis, lobar collapse or surgical resection; andeither (a) they had unequivocal evidence ofunilateral severe diaphragm weakness, assessedby TWPDI; or (b) they were studied because of thefinding of elevated hemidiaphragm on chest radio-graph. We excluded all patients, whose hemidiaph-ragms were non interpretable on chest X-raysbecause of the presence of scoliosis or pleuraleffusion or parenchimal lung disease. The patientswere in a clinically stable condition and chestradiograph and respiratory muscle study wereperformed within two weeks.

Chest radiograph assessment

Chest radiographs (posteroanterior and lateralprojections) were obtained with the patient’supright, holding their breath at full inspiration.

Two radiologists independently reviewed the chestradiographs and reported whether the hemidiaph-ragm elevation was present or absent as well aswhether lung parenchyma was normal or not. Theradiologists were blinded to both the clinicaldetails of the patient and the function of phrenicnerve and diaphragm. Only reports in agreementbetween the radiologists were analyzed.

Pulmonary function tests and pressuremeasurements

Lung volumes and spirometry were measuredaccording to the standard guidelines11 and ex-pressed as percentages of the predicted values.12

Maximal static inspiratory (PIMAX) and expiratory(PEMAX) pressures were measured from residualvolume and total lung capacity,13 respectively, andthe highest recorded pressures maintained for 1 swere used for analysis and expressed as percent ofpredicted.14

After insertion of oesophageal and gastric bal-loon catheters, oesophageal pressure (PES), gastricpressure (PGAS), and transdiaphragmatic pressure(PDI) were measured in cm H2O with differentialpressure transducers (Validyne, Northridge, CA)and amplified before passing the data to acomputer (Apple Computers, Cupertino, CA). Dataacquisition and analysis were performed with Lab-view 4.1 (National Intruments, Austin, TX).

Hemidiaphragm dysfunction was diagnosed bymagnetic phrenic nerve stimulation as previouslydescribed.7,15 Briefly, a Magstim 200 (high power)magnetic stimulator was used to power themagnetic coils (Magstim Co., Whitland, Dyfed,UK). Unilateral stimulation was applied with thesubjects breathholding at resting end expirationagainst a closed airway. The phrenic nerve wasstimulated at the posterior border of the sterno-cleiodomastoid muscle at the level of the cricoidcartilage. Supramaximal stimulation, manifestedby a levelling off of TWPDI with increasing stimulusintensities, was achieved in all subjects. Unilateraldiaphragm dysfunction was diagnosed if right orleft TWPDI was less than 3.5 cm H2O. This cut-offlevel was arbitrarily chosen because cardiac con-traction may cause intrathoracic pressure fluctua-tions of up to 2 cm H2O.

Statistical analysis

Values are presented as mean7standard deviation(SD), unless otherwise specified. The agreementbetween radiologists and between diagnostic testswas evaluated using kappa agreement test for

ARTICLE IN PRESS

40 A. Chetta et al.

dichotomic variables; so that a kappa value of 1indicates complete agreement while as a kappavalue of 0 indicates disagreement.16 Standardmethods were used to calculate sensitivity, speci-ficity, and predictive values for chest radiographs,as a diagnostic test.17 The receiver operatingcharacteristic (ROC) curve method18 was used toevaluate the screening cut-off for TWPDI withrespect to sensitivity and specificity of chestradiograph.

Results

In the overall patient population, a moderateinterobserver agreement was found in the chestradiograph reading (Table 1). In 21 out of 112patients, hemidiaphragms were judged as non-interpretable on chest radiograph. In addition, 49patients did not meet inclusion criteria, in parti-cular 13 patients had bilateral elevation of hemi-diaphragms, 4 had bilateral diaphragm paralysisjudged by phrenic nerve stimulation, 11 hadunilateral elevation of the hemidiaphragm alongwith an ipsilateral lung disease, and 21 hadinconsistent reports by the two radiologists. Thus42 patients (17 female, age range 22–79 years)remained for further analysis.

The 42 remained patients had a variety of clinicalproblems and according to their primary clinicaldiagnosis, were divided in four groups: elevatedhemidiaphragm with no other symptoms (n¼ 24),neuromuscular disorders (n¼ 5), unexplained dys-pnoea (n¼ 7), and other (n¼ 6) (Table 2).

Fourteen patients had an elevated position ofright hemidiaphragm on chest radiograph, 13 had

ARTICLE IN PRESS

Table 1 Interobserver agreement in chest radio-graph interpretation.

No. Chest radiographfindings

Percent ofagreement (%)

k value

112 Lung parenchima 0.59Normal 83Abnormal 77

91 Righthemidiaphragm

0.53

Normal 90Elevated 63

91 Lefthemidiaphragm

0.48

Normal 84Elevated 64 Ta

ble

2Charac

teristicsofpatients.

Prim

arydiagn

osis

No.

Gender

(F/M

)Age

(Year)

TLC

(%pred)

FVC

(%pred)

FEV1

(%pred)

FEV1/F

VC

(%pred)

PI M

AX

(%pred)

PEMAX

(%pred)

Eleva

tedhemidiaphragm

248/

1660

713

75714

64720

57720

69710

70742

63738

Neuromusculardisorders

53/

240

716

87714

68720

71722

88710

43733

35718

Unexp

laineddyspnoea

73/

451

719

99724

103732

98720

78716

78735

66730

Other

63/

351

718

82724

82726

78715

8077

43715

55711

Total

4217

/25

55716

82718

72725

67724

74712

66738

60733

TLC

¼Totallungca

pac

ity,

FVC¼Fo

rcedvitalca

pac

ity,

PI M

AX¼Max

imal

inspiratory

pressure,PEMAX¼Max

imal

exp

iratory

pressure.

Chest radiography cannot predict diaphragm function 41

an elevated position of left hemidiaphragm, and 15patients had normal position of both hemidiaph-ragms; the prevalence of patients with an elevatedunilateral hemidiaphragm on chest radiograph was64%. Considering the 27 elevated hemidiaphragms,the mean value (range) of ipsilateral TWPDI was5.6 cm H2O (0–19.4 cm H2O) for the right hemi-diaphragm and 4.2 cm H2O (0–21 cm H2O), for theleft hemidiaphragm.

Thirteen patients had hemidiaphragm dysfunc-tion, (seven right-sided and six left-sided) and 29patients had no evidence of diaphragmatic dysfunc-tion; the prevalence of patients with unilateralhemidiaphragm dysfunction paralysis judged byTWPDI was 24%.

When we related the chest radiograph findings toTWPDI results, we found that the chest radiographhad a false-negative and false-positive rate of 10%and 56%, respectively. Moreover, sensitivity, speci-ficity, positive and negative predictive values forchest radiograph as diagnostic test were respec-tively 0.90, 0.44, 0.33, and 0.93. In addition, wefound fair agreement between the two diagnostictools by kappa statistics (Table 3).

Finally because of concern we failed to demon-strate a value for chest radiography because ourthreshold TWPDI of 3.5 cm H2O had been incorrectlychosen we constructed a ROC curve to assessdifferent threshold values of TWPDI with respect tohemidiaphragm elevation on plain chest radiograph(Fig. 1). The area under the ROC curve was 0.76 andchanging the threshold value for TWPDI did notsignificantly improve the value of the test.

Discussion

In this study, we found that the finding of anelevated hemidiaphragm on plain chest radiographis moderately sensitive, but poorly specific for thediagnosis of diaphragm dysfunction, at least inpatients referred to a tertiary pulmonary centre,even when patients with conditions known to causevolume loss are excluded on the basis of clinicalhistory and the plain radiograph; the diagnostic

value of the chest radiograph would have beeneven worse had we not excluded from our analysisthe 21 of 112 patients (19%) in whom theradiologists failed to agree.

The usefulness of chest radiograph as a screeningtest for unilateral hemidiaphragm paralysis iswidely assumed in clinical practice and it is alsoclaimed that the diagnosis of hemidiaphragmparalysis can be radiologically made in most cases,even though reasonable care is requested.1 How-ever, early studies reported discordant results onthis topic. Wynn–Williams described 15 cases ofparalysis without any marked rise in level and foundthat the right diaphragm is normally 1.3 cm higherthan left, with a range from zero to 2.5 cm.9

Previously, Freedman suggested a wider normalrange of 3.8 cm above Wynn–Williams’.8 Thus,seven out of Wynn–Williams’ 15 cases fall withinFreedman’s normal range. Moreover, another studyreported that only one out of six cases of malignantdiaphragmatic paralysis showed elevation outsideFreedman’s normal range.10 Recently, a substantialvariability in diaphragm position and shape was alsofound in adults with normal pulmonary functionwhich was related in part to weight, age andthoracic dimensions.19

To improve the diagnostic sensitivity of chestradiograph of diaphragm paralysis, the fluoroscopicor ultrasound screening of the diaphragm during amaximal sniff is sometimes suggested as a com-plementary test.1,2 However, sniffing can causeparadoxical motion of one hemidiaphragm in somehealthy people and patients with unilateral paraly-sis may use the abdominal muscles to elevate thediaphragm during expiration; accordingly false-positive and false-negative results have beenreported.10 Taken together these studies supportthe view that there is a potential pitfall in theradiological assessment of unilateral paralysis, asthe normal height of the hemidiaphragm mayconsiderably vary and a wide range of normal orabnormal circumstances in which paradoxicalmovement of diaphragm may be present.

In this study, we related the chest X-rayappearances to the twitch transdiaphragmaticpressure, which we assume to be the ‘goldstandard’ test of hemidiaphragm function. Directproof that this assumption is true is difficult toacquire in vivo but we have previously demon-strated that TWPDI varies appropriately in responseto interventions such as fatigue,20 lung volumechange21 and drugs22 which are known to affectskeletal muscle function and also that TWPDI canpredict respiratory failure in patients with condi-tions such as Amyotrophic Lateral Sclerosis23 wherediaphragm involvement has been pathologically

ARTICLE IN PRESS

Table 3 TWPDI and chest radiograph agreement in42 patients.

Hemidiaphragmfunction by TWPDI

Percent ofagreement (%)

k value

Normal 93 0.31Right paralysis 21Left paralysis 46

42 A. Chetta et al.

recognised. Thus we believe that the use of TWPDIas a reference standard is justified.

We provide evidence that chest radiographs cangive a considerable number of false positives, butfewer false negatives when compared to TWPDIdata. Use of the ROC plot demonstrated that toreduce the number of false positives a cut-off levelof TWPDI49 cm H2O was required, which is consis-tent with a normal function of a hemidiaphragm.24

However, at this TWPDI value sensitivity was verylow with a great number of false negatives.

We found that the proportions of patients withnegative and positive test results that werecorrectly diagnosed were quite different, amount-ing to 0.93 and 0.33, respectively. The predictivevalue of any diagnostic test depends not only on thetest characteristics but also on the prevalence ofthe disease within the studied population.17 Thus,the relatively low prevalence of unilateral paralysisthat we found in our selected population empha-sizes the negative predictive value of chest radio-graphs to exclude hemidiaphragm paralysis rather

than its positive predictive value to find anabnormal patient. Though a normal diaphragmposition on a chest radiograph makes hemidiaph-ragm paralysis unlikely our results show that thechest radiograph cannot be considered as analternative tool to TWPDI to confirm hemidiaphragmparalysis.

As far as we know, this is the first study thatevaluated the interobserver agreement to detectthe presence of unilateral elevation of hemidiaph-ragm on chest radiograph. Interestingly, the agree-ment was higher to detect parenchymalabnormalities than the elevation of unilateralhemidiaphragm. As a whole, the interobserveragreement was moderate, the kappa value rangingfrom 0.48 to 0.59, this finding being in line withprevious reports. A study evaluating the interob-server variability on the presence of an infiltrate onthe chest radiograph found a kappa value between0.37 and 0.51.25 A wider interobserver variabilitywas reported to discriminate normal from abnormalradiographs26 and to detect lung cancer.27

ARTICLE IN PRESS

Figure 1 ROC curve for TWPDI as screening cutoff for chest radiograph. There is no acceptable cut-off point on thecurve.

Chest radiography cannot predict diaphragm function 43

In summary, our study showed that amongpatients referred to hospital, hemidiaphragm ele-vation on plain chest radiograph is an unreliabletest to diagnose hemidiaphragm dysfunction. How-ever, a normally positioned hemidiaphragm makeshemidiaphragm dysfunction relatively unlikely.

References

1. Fraser RS, Muller NL, Colman N, Pare PD. eds. Thediaphragm. In: Diagnosis of diseases of the chest, Vol. IV.Philadelphia: WB Saunders Company; 1999. p. 2987–3010.

2. Slone RM. Chest wall and diaphragm. In: Slone RM, GutierrezFR, Fisher AJ, editors. Thoracic imaging. A practicalapproach. New York: McGraw-Hill; 1999. p. 191–3.

3. Tarver RD, Godwin JD, Putman CE. Symposium on non-pulmonary aspects in chest radiology. The diaphragm. RadiolClin North Am 1984;22:615–31.

4. Bellemare F, Bigland-Ritchie B. Assessment of humandiaphragm strength and activation using phrenic nervestimulation. Respir Physiol 1984;58:263–77.

5. Mier A, Brophy C, Moxham J, Green M. Twitch pressures inthe assessment of diaphragm weakness. Thorax1989;44:990–6.

6. Similowski T, Fleury B, Launois S, Cathala HP, Bouche P,Derenne JP. Cervical magnetic stimulation: a new painlessmethod for bilateral phrenic nerve stimulation in conscioushumans. J Appl Physiol 1989;67:1311–8.

7. Mills GH, Kyroussis D, Hamnegard C-H, Wragg S, Moxham J,Green M. Unilateral magnetic stimulation of the phrenicnerve. Thorax 1995;50:1162–72.

8. Freedman B. Unilateral paralysis of the diaphragm andlarynx associated with inflammatory lung disease. Thorax1950;5:169–82.

9. Wynn-Williams N. Hemidiaphragmatic paralysis and paresisof unknown aetiology without any marked rise in level.Thorax 1954;9:299–303.

10. Alexander C. Diaphragm movements and diagnosis ofdiaphragmatic paralysis. Clin Radiol 1966;17:79–83.

11. British Thoracic Society. Guidelines for the measurement ofrespiratory function. Respir Med 1994;88:165–94.

12. Quanjer PhH, Tammeling GJ, Cotes JE, et al. Lung volumesand forced ventilatory flows. Eur Respir J 1993;6:5–40.

13. Black LF, Hyatt RE. Maximal respiratory pressures: normalvalues and relationships to age and sex. Am Rev Respir Dis1969;99:696–702.

14. Wilson SH, Cooke NT, Edwards RHT, Spiro SG. Predictednormal values for maximal respiratory pressures in Cauca-sian adults and children. Thorax 1984;39:535–8.

15. Mills GH, Kyroussis D, Hammegard C-H, Polkey MI, Green M,Moxham J. Bilateral magnetic stimulation of the phrenicnerves from an alterolateral approach. Am J Respir Crit CareMed 1996;154:1099–105.

16. Landis RJ, Koch CG. The measurement of observer agree-ment for categorical data. Biometrics 1977;33:159–74.

17. Griner PF, Mayewski RJ, Mushlin AI, Greenland P. Selectionand interpretation of diagnostic tests and procedures.Principles and applications. Ann Intern Med 1981;94:557–92.

18. Zweig MH, Campell G. Receiver-operating characteristic(ROC) plots: a fundamental evaluation tool in clinicalmedicine. Clin Chem 1993;39:561–77.

19. Suwatanapongched T, Gierada DS, Slone RM, et al. Variationin diaphragm position and shape in adults with normalpulmonary function. Chest 2003;123:2019–27.

20. Hamnegard CH, Wragg S, Kyroussis D, Mills GH, Polkey MI,Moran J, Road J, Bake B, Green M, Moxham J. Diaphragmfatigue following maximal ventilation in man. Eur Respir J1996;9:241–7.

21. Polkey MI, Hamnegard CH, Hughes PD, Rafferty GF, Green M,Moxham J. Influence of acute lung volume change oncontractile properties of human diaphragm. J Appl Physiol1998;85:1322–8.

22. Fauroux B, Cordingley J, Hart N, Clement A, Moxham J,Lofaso F, Polkey MI. Depression of diaphragm contractility bynitrous oxide in humans. Anesth Analg 2002;94:340–5.

23. Lyall RA, Donaldson N, Polkey MI, Leigh PN, Moxham J.Respiratory muscle strength and ventilatory failure inamyotrophic lateral sclerosis. Brain 2001;124:2000–13.

24. Mills GH, Kyroussis D, Hamnegard CH, Wragg S, Moxham J,Green M. Unilateral magnetic stimulation of the phrenicnerve. Thorax 1995;50:1162–72.

25. Albaum MN, Hill LC, Murphy M, Li YH, Fuhrman CR, BrittonCA, Kapoor WN, Fine MJ. Interobserver reliability of thechest radiograph in community-acquired pneumonia. PORTInvestigators. Chest 1996;110:343–50.

26. Tudor GR, Finlay D, Taub N. An assessment of inter-observeragreement and accuracy when reporting plain radiographs.Clin Radiol 1997;52:235–8.

27. Quekel LG, Kessels AG, Goei R, van Engelshoven JM.Detection of lung cancer on the chest radiograph: a studyon observer performance. Eur J Radiol 2001;39:111–6.

ARTICLE IN PRESS

44 A. Chetta et al.