Prediction of Cardiovascular Events by Using Non-Vascular Findings on Routine Chest CT
et al. (2011) Prediction of Cardiovascular Events by Using Non-Vascular Findings on
Routine Chest CT. PLoS ONE 6(10): e26036. doi:10.1371/journal.pone.0026036
Prediction of Cardiovascular Events by Using Non-Vascular Findings on Routine Chest CT
Pim A. de Jong 0
Martijn J. A. Gondrie 0
Constantinus F. M. Buckens 0
Peter C. Jacobs 0
Willem P. T h. M. Mali 0
Yolanda van der Graaf 0
the PROVIDI study group 0
Ewout W. Steyerberg, University Medical Center Rotterdam, The Netherlands
0 1 Department of Radiology, University Medical Center Utrecht , Utrecht , The Netherlands , 2 Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht , Utrecht , The Netherlands
Background: Routine computed tomography (CT) examinations contain an abundance of findings unrelated to the diagnostic question. Those with prognostic significance may contribute to early detection and treatment of disease, irrelevant findings can be ignored. We aimed to assess the association between unrequested chest CT findings in lungs, mediastinum and pleura and future cardiovascular events. Methods: Multi-center case-cohort study in 5 tertiary and 3 secondary care hospitals involving 10410 subjects who underwent routine chest CT for non-cardiovascular reasons. 493 cardiovascular hospitalizations or deaths were recorded during an average follow-up time of 17.8 months. 1191 patients were randomly sampled to serve as a control subcohort. Hazard ratios and annualized event rates were calculated. Results: Abnormalities in the lung (26-44%), pleura (14-15%) and mediastinum (20%) were common. Hazard ratios after adjustment for age and sex were for airway wall thickening 2.26 (1.59-3.22), ground glass opacities 2.50 (1.72-3.62), consolidations 1.97 (1.12-3.47), pleural effusions 2.77 (1.81-4.25) and lymph-nodes 2.04 (1.40-2.96). Corresponding annual event rates were 5.5%, 6.0%, 3.8%, 10.2% and 4.4%. Conclusions: We have identified several common chest CT findings that are predictive for future risk of cardiovascular events and found that other findings have little utility for this. The added value of the non-vascular predictors to established vascular calcifications on CT remains to be determined.
The use of diagnostic imaging is increasing rapidly: in the United
States of America approximately 62 million computed tomography
(CT) examinations  are obtained annually in a population of
some 300 million. Chest CT images exhibit an abundance of
findings unrelated to the clinical problem which challenge
physicians as they often lead to further testing, patient anxiety
and unnecessary procedures for findings without relevance. On the
other hand unrequested findings also provide an opportunity for
early disease detection and prevention of symptoms and mortality
[2,3]. Therefore, reliable, reproducible and meaningful prognostic
models need to be developed for treatable diseases and the impact of
such models on patient outcome needs to be evaluated .
We hypothesized that certain findings in the lung, pleura and
mediastinum on routine chest CT examinations might contain
prognostic information for cardiovascular events. The hypothesis
is based on the fact that airway obstruction is consistently found to
be independently associated with cardiovascular mortality  and
lung diseases are now coming to be seen as a risk factor for
cardiovascular mortality [6,7]. Several mechanisms including
ongoing pulmonary inflammation [5,8], repair mechanisms ,
misbalances between proteases and anti-proteases  and
autoimmune phenomena  that damage the lung also appear to
adversely affect the cardiovascular system. The advances and
increased use of chest CT make the pulmonary manifestations of
these processes increasingly accessible by direct in vivo
visualization of the morphological signs  of obstructive pulmonary
disease [10,1216] and chest inflammation [10,1218].
We believe that for an important subgroup of patients who
undergo chest CT the cardiovascular risk is unknown and
therefore that a cardiovascular risk calculation based on
unrequested chest CT findings may reveal hidden cardiovascular
disease. Cardiovascular diseases, despite their high prevalence 
and high awareness of risk reduction measures, often manifest
suddenly and unexpectedly in a large portion of patients with no
prior cardiovascular disease history . Identification of at-risk
patients using unrequested findings on otherwise routine chest CT
leading to timely diagnosis and concomitant treatment may be
relevant as early treatment has potential to reduce risk of future
cardiovascular events [21,22].
The aim of this study was to assess the association between
unrequested chest CT findings in lungs, mediastinum and pleura
and future cardiovascular events in a cohort of routine diagnostic
chest CTs performed on adults in the Netherlands.
This study was approved by the ethical review board of the
University Medical Center Utrecht (number 06/193), data were
analyzed anonymously, informed consent was waived.
The rationale and design of the PROVIDI (PROgnostic Value of
unrequested Information in Diagnostic Imaging) study has been
described in detail and is presented in Figure 1 . Briefly, we
identified 23443 subjects 40 years of age and older who underwent
chest CT in eight hospitals. Subjects who had lung cancer or
metastasis were excluded (N = 9077) because of the poor prognosis.
Patients who were suspected of or whom were known to have the
outcome of interest (cardiovascular disease) on the pre-CT referral
forms were excluded (N = 2303). Finally, for future external validation
studies, 1653 subjects from one centre were excluded. For the present
study the cohort consisted therefore of chest CT examinations from
10410 individuals. Cases consisted of those subjects who experienced a
cardiovascular event hospitalization or died from cardiovascular
disease during follow-up. Cardiovascular diseases included
cerebrovascular disease, coronary heart disease, peripheral artery disease,
heart failure, aortic aneurysm and non-rheumatic heart disease.
A case-cohort design (Figure 1)  was used to limit the
number of CT examinations that had to be scored by the observers,
while still enabling absolute risk estimations. The random sample of
the cohort (subcohort) included 1285 individuals and 94 (7.3%) of
these were excluded because the CT examination could not be
retrieved, leaving a subcohort of 1191. The cardiovascular
endpoints were recorded during an average follow-up time of
17.8 months. The endpoints were obtained through linkage with the
National Death Registry and the National Registry of Hospital
Discharge Diagnoses . In cases with multiple valid
endpoints, cause of death prevailed over hospital admissions; otherwise
the first hospital discharge diagnosis to occur was used.
Chest CT scoring
All chest CT examinations were obtained using multi-detector
CT (264 detector rows) of different vendors according to the
prevailing routine clinical protocols of the participating hospitals.
All CT examinations (all images from all CT data acquisitions) were
scored by one of two observers for abnormalities in the
mediastinum, lungs and pleura (Table 1) that were considered to
be potentially related to inflammation and/or obstructive
pulmonary function. For emphysema the scoring system from the National
Emphysema Treatment Trial  was used, other pulmonary
abnormalities were scored as absent or present in 5 lobes.
Abnormalities were defined according to the Fleischner Society
criteria . A random set of 150 examinations were scored by both
observers independently to evaluate the inter-observer agreement.
For bronchiectasis, airway wall thickening, ground glass and pleural
effusion score 1 was defined as mild, score 2 as moderate and score 3
and higher as severe disease. For emphysema score 13 was defined
as mild, score 36 as moderate and score 7 and higher as severe
disease. For lymph-adenopathy a short axis diameter between 6 and
10-mm was defined as mild and greater than 10-mm as moderate
disease. In routine clinical practice lymph nodes ,10-mm are seen
as normal, but we argued that these small lymph nodes may still be
predictive for cardiovascular disease.
Observer agreement was calculated using a weighted Kappa
statistic or an intra-class correlation coefficient (icc). Crude and
Emphysema for five lobes (score 020)
Bronchiectasis for five lobes (score 05)
Airway wall thickening for five lobes (score 05)
Ground glass for five lobes (score 05)
Consolidation for five lobes (score 05)
Pleural effusion (score 06) *
Pleural plaques (score 02)
Mediastinal lymph node (if $5 mm)
*For pleural effusion, the thickness of the fluid layer was measured on axial images. For bilateral effusions the most severe side was scored.
age and sex adjusted hazard ratios for cardiovascular events were
estimated by using Cox proportional-hazards regression.
Interaction with CT indication and slice thickness was tested for CT
findings that were significantly associated with any cardiovascular
event. Annualized event rates were calculated for each CT
abnormality . Missing values for CT scores (all ,4%) were
imputed using regression methods in SPSS (SPSS 14.0, Chicago,
During follow-up we recorded 559 cases of incident
cardiovascular disease, 66 (11.8 %) of these were excluded because the CT
examination could not be retrieved. Further details of the
subcohort and cases are presented in Table 2.
Lung abnormalities were found to be common with an overall
prevalence between 26% and 30% for the various abnormalities in
the subcohort. Lymph-nodes .10-mm were seen in 20% whilst
pleural abnormalities were less common (14 and 15% for effusion
and plaques respectively). There was minimal overlap between
these CT findings; it was more common to have either airway wall
thickening (42%) or ground glass (36%) than having both (21%).
This pattern was similar for the subcohort and the cases and for
the other CT findings. Observer agreement was good for pleural
effusion (kappa 0.89) and emphysema (icc 0.80), moderate for
consolidations (kappa 0.45), lymphadenopathy (kappa 0.54) and
plaques (kappa 0.42) and fair for airway wall thickening (kappa
0.33), bronchiectasis (kappa 0.31) and ground glass (kappa 0.38).
Prediction of cardiovascular events
Several CT findings predicted cardiovascular events. Compared
to subjects without the given CT abnormality the hazard for any
cardiovascular event was significantly increased in patients with
airway wall thickening in more than 1 lobe, ground glass in more
than 2 lobes, any consolidation, more than a small unilateral
pleural effusion and mediastinal lymphadenopathy .10-mm, even
when age and sex were taken into account. The strongest adjusted
hazard ratio was for moderate pleural effusion (adjusted hazard
ratio 2.77). This corresponds to an annual event rate of 10%, while
the annual event rate was 2.3% for the group with no effusion.
On the other hand several other CT items were not predictive
for cardiovascular events. The hazard ratios confidence interval
for emphysema, bronchiectasis and pleural plaques covered the
value of 1.0 corresponding to no increase in hazard compared to
the group without the abnormality.
No significant interaction was found between slice thickness and
CT indication and the hazards ratios for airway wall thickening,
ground glass, consolidation, pleural effusion and
lymph-adenopathy. Such interaction was tested in the age and sex adjusted
models. The hazard ratios are presented in Table 3 and
annualized mortality and age and sex adjusted Kaplan Meier
curves are presented in Table 4 and Figure 2.
In a cohort of patients who underwent routine chest CT
examinations we showed that unrequested findings in the lungs,
mediastinum and pleura are common and that airway wall
thickening, ground glass opacities, areas of consolidation, pleural
effusion and mediastinal lymph nodes predict cardiovascular
events. On the other hand emphysema, bronchiectasis and pleural
plaques did not predict cardiovascular events.
In the present study we propose a novel strategy which is the
reason why it is not possible to calculate incremental value over
known cardiovascular risk factors. We will now illustrate our
design to explain why calculating incremental value is not possible
and conform clinical practice from the perspective of a radiologist.
We see the clinical scenario as follows: A patient is submitted for a
chest CT. The radiologist, parallel to the diagnostic procedure,
calculates the cardiovascular risk based on CT findings, age and
sex. In clinical routine the radiologist does not have access to
cardiovascular risk factors and these are even unknown for a large
proportion of patients referred for chest CT. If based on the CT
findings the cardiovascular risk reaches a certain level this is
reported to the referring physician, who can next evaluate
conventional risk factors (as long as there is no evidence that CT
findings can be treated). The patient is subsequently managed and
treated based on current guidelines. The patient may benefit from
this CT based strategy as cardiovascular death often occur sudden
in patients unknown with cardiovascular disease and even patients
known with cardiovascular disease can be undertreated.
This type of prediction research based on routine CT imaging
findings has gained interest recently [24,31] and opens a wide
field of imaging research with new hypotheses to be tested . The
ultimate goal is to determine whether routine CT findings can
detect and successfully treat patients at high risk for treatable
disease . Currently imaging findings are often summarized in
detail in the radiologists written report while the meaning to
referring physicians can be unknown. This potentially leads to
unneeded further work-up while on the other hand potentially
Table 2. Characteristics of sub-cohort and cases.
Sub-cohort (N = 1191)
Cardiovascular disease cases (N = 493)
useful findings are ignored. It would be more meaningful to report
only relevant unrequested findings for which the prognostic value
is known and incorporate absolute risk estimates in the report,
especially for outcomes of which the prognosis can be altered.
Pragmatically, these risk estimation should be based on
information that is readily available to radiologists such as age, sex,
CT only (crude)
CT, age and sex
CT, age and sex
CT, age and sex
Data given are hazard ratios (95% confidence interval). The hazard ratio for CT severity absent is 1.0. N.A. = not applicable.
Ground glass (%/year)
Pleural effusion (%/year)
Pleural plaques (%/year)
Lymphadenopathy (%/year) 2.1
N.A. = not applicable.
probably some medical information (previous malignancy) and
imaging findings. The referring physician can then assess, based
on the by the radiologist reported risk, whether this reveals hidden
disease or whether the patient was already known with increased
cardiovascular risk. Our hypothesis that unrequested findings in
the lung, pleura and mediastinum might harbor possible
prognostic factors was borne out by the observation that several
chest diseases are associated with cardiovascular events. Our study
may serve as a starting point for including findings in the lungs,
mediastinum and pleura in a comprehensive prognostic model
that includes known prognostic cardiovascular disease factors such
as coronary calcifications. Although our hypothesis was based on
potentially causal associations between abnormalities in the lungs
[5,6,810,12,3234], mediastinum and pleura , our study
cannot establish causality between the morphological findings and
future cardiovascular events. Possibly, the significant findings in
our study are simply a consequence of (subclinical) heart failure
[35,36]. Nevertheless we would like to stress that for the predictive
purposes of this study, merely the strength of the predictor and the
absolute risk that is associated with that predictor are of interest
while the underlying mechanisms are of secondary interest.
Our study has limitations. First, we cannot be certain that there
was no history of cardiovascular disease or known cardiovascular
risk profile amongst our patients, although we minimized this
limitation by excluding all CT examinations obtained for a
cardiovascular indication. Also cardiovascular events still occur
often sudden and unexpected . Because of these 2 arguments,
we believe that although misclassification is an issue this does not
invalidate the study. Second, the events are registered on a
national basis and there may be errors leading to misclassification
on the outcome. Given the strength of the CT predictors in our
study, some misclassification would not have seriously altered our
conclusion. Third, we cannot be certain that our CT findings were
truly incidental or unrequested in all subjects.
In conclusion, unrequested imaging findings in the lungs,
mediastinum and pleura are common on diagnostic chest CT.
Some CT examinations exhibit prognostic information while
others do not. Improving understanding of this information can
lead to more intelligent imaging, chest CT reporting, prevention of
unnecessary further testing and costs for those findings without
prognostic relevance and efficient use of prognostic findings to
prevent diseases. CT findings in lungs, mediastinum and pleura
may facilitate the detection of subjects at risk for cardiovascular
events, especially the detection of airway wall thickening in more
than 1 lobe, ground glass in more than 2 lobes, any area of
consolidation, more than a small unilateral pleural effusion and
mediastinal lymphadenopathy .10-mm. Logical further studies
would be the development of a prognostic model that includes
vascular calcifications and the evaluation of such models regarding
impact on patient management and outcome is needed before
We acknowledge the help with data collection and administrative support
by Cees Haaring (UMCU, Utrecht, The Netherlands), Karin Flobbe
(AZM, Maastricht, The Netherlands), Mirjam Borg (VUMC, Amsterdam,
The Netherlands), Pascal Verzijl (St Antonius Hospital, Nieuwegein, The
Netherlands), Jan Wolfers (AMC, Amsterdam, The Netherlands), Peter
van Ooijen (UMCG, Groningen, The Netherlands), Willie Donkers
(Elkerliek Hospital, Helmond, The Netherlands), and Robert Dragt (Gelre
Hospital, Apeldoorn, The Netherlands).
Members of the PROVIDI study group: Johan Lameris (Dept. of
Radiology, Academic Medical Center, Amsterdam, The Netherlands),
Cees van Kuijk (Dept. of Radiology, VU Medical Center, Amsterdam, The
Netherlands), Willem ten Hove (Dept. of Radiology, Gelre Hospital,
Apeldoorn, The Netherlands), Matthijs Oudkerk (Dept. of Radiology,
University Medical Center Groningen, The Netherlands), Ayke L. Oen
(Dept. of Radiology, Elkerliek Hospital, Helmond, The Netherlands),
Joachim Wildberger (Dept. of Radiology, Academic Hospital Maastricht,
Maastricht, The Netherlands), Johannes van Heesewijk (Dept. of
Radiology, St Antonius Hospital, Nieuwegein, The Netherlands), Mathias
Prokop (Dept. of Radiology, University Medical Center Nijmegen,
Nijmegen, The Netherlands).
Conceived and designed the experiments: PAdJ MJAG CFMB PCJ
WPThMM YG PROVIDI Study Group. Performed the experiments:
PAdJ MJAG. Analyzed the data: PAdJ MJAG CFMB. Contributed
reagents/materials/analysis tools: MJAG PCJ WPThMM YG PROVIDI
Study Group. Wrote the paper: PAdJ MJAG CFMB PCJ WPThMM YG.
Approved the final version of the manuscript: PAdJ MJAG CFMB PCJ
WPThMM YG PROVIDI Study Group.
1. Brenner DJ , Hall EJ ( 2007 ) Computed tomography--an increasing source of radiation exposure . N Engl J Med 357 ( 22 ): 2277 - 84 .
2. Lee CI , Forman HP ( 2010 ) What we can and cannot see coming . Radiology Nov 257 ( 2 ): 313 - 4 .
3. Gondrie MJ , Mali WP , Jacobs PC , Oen AL , van der Graaf Y , et al. ( 2010 ) Cardiovascular disease: prediction with ancillary aortic findings on chest CT scans in routine practice . Radiology Nov 257 ( 2 ): 549 - 59 .
4. Gondrie MJ , Mali WP , Buckens CF , Jacobs PC , Grobbee DE , et al. ( 2010 ) The PROgnostic Value of unrequested Information in Diagnostic Imaging (PROVIDI) Study: rationale and design . Eur J Epidemiol Oct ; 25 ( 10 ): 751 - 8 .
5. van Eeden SF , Sin DD ( 2008 ) Chronic obstructive pulmonary disease: a chronic systemic inflammatory disease . Respiration 75 ( 2 ): 224 - 38 .
6. McAllister DA , Maclay JD , Mills NL , Mair G , Miller J , et al. ( 2007 ) Arterial stiffness is independently associated with emphysema severity in patients with chronic obstructive pulmonary disease . Am J Respir Crit Care Med 176 ( 12 ): 1208 - 14 .
7. Sin DD ( 2209 ) Is COPD really a cardiovascular disease ? Chest Aug; 136 ( 2 ): 329 - 30 .
8. Macnee W , Maclay J , McAllister D ( 2008 ) Cardiovascular injury and repair in chronic obstructive pulmonary disease . Proc Am Thorac Soc Dec 1 ; 5 ( 8 ): 824 - 33 .
9. Man SF , Xing L , Connett JE , Anthonisen NR , Wise RA , et al. ( 2008 ) Circulating fibronectin to C-reactive protein ratio and mortality: a biomarker in COPD? Eur Respir J Dec; 32 ( 6 ): 1451 - 7 .
10. Lee SH , Goswami S , Grudo A , Song LZ , Bandi V , et al. ( 2007 ) Antielastin autoimmunity in tobacco smoking-induced emphysema . Nat Med 13 ( 5 ): 567 - 9 .
11. Jacobs PC , Prokop M , Oen AL , van der Graaf Y , Grobbee DE , et al. ( 2008 ) Prevalence of incidental findings in computed tomographic screening of the chest: a systematic review . J Comput Assist Tomogr 32 ( 2 ): p. 214 - 21 .
12. Williams FL , Lloyd OL ( 1989 ) Cerebrovascular disease in Scotland during 1959 to 1983: its geographical distribution and associations . Community Med Nov ; 11 ( 4 ): 306 - 15 .
13. Keistinen T , Saynaja kangas O , Tuuponen T , Kivela SL ( 1997 ) Bronchiectasis: an orphan disease with a poorly-understood prognosis . Eur Resp J 10 : 2784 - 2787 .
14. Akalin F , Ko roglu TF , Bakac S, Dagli E ( 2003 ) Effects of childhood bronchiectasis on cardiac functions . Pediatrics International 45 : 169 - 174 .
15. Liu SF , Liu JW , Lin MC ( 2005 ) Characteristics of patients suffering from tuberculous pleuritis with pleural effusion culture positive and negative for Mycobacterium tuberculosis, and risk factors for fatality . Int J Tuberc Lung Dis Jan ; 9 ( 1 ): 111 - 5 .
16. Wells DK , Ward MM ( 2010 ) Nephritis and the risk of acute myocardial infarction in patients with systemic lupus erythematosus . Clin Exp Rheumatol May-Jun ; 28 ( 2 ): 223 - 9 .
17. Badui E , Garcia-Rubi D , Robles E , Jimenez J , Juan L , et al. ( 1985 ) Cardiovascular manifestations in systemic lupus erythematosus . Prospective study of 100 patients. Angiology Jul ; 36 ( 7 ): 431 - 41 .
18. Vehmas T , Hiltunen A , Kivisaari L , Leino-Arjas P ( 2008 ) Atherosclerotic and pleural calcifications are related among asbestos-exposed workers . Eur J Cardiovasc Prev Rehabil Oct ; 15 ( 5 ): 599 - 601 .
19. Mathers CD , Loncar D ( 2006 ) Projections of global mortality and burden of disease from 2002 to 2030 . PLoS Med 3 ( 11 ): e442 .
20. Fox CS , Evans JC , Larson MG , Kannel WB , Levy D ( 2004 ) Temporal trends in coronary heart disease mortality and sudden cardiac death from 1950 to 1999: the Framingham Heart Study . Circulation 110 ( 5 ): 522 - 7 .
21. Critchley JA , Capewell S ( 2003 ) Mortality risk reduction associated with smoking cessation in patients with coronary heart disease: a systematic review . JAMA 290 ( 1 ): 86 - 97 .
22. LaRosa JC , Grundy SM , Waters DD , Shear C , Barter P , Fruchart JC , et al. ( 2005 ) Intensive lipid lowering with atorvastatin in patients with stable coronary disease . N Engl J Med 352 ( 14 ): 1425 - 35 .
23. Pentice RL ( 1986 ) A case-cohort design for epidemiologic cohort studies and disease precention trials . Biometrika . 73 : 1 - 11 .
24. De Bruin A , Kardaun JW , Gast A , Bruin E , van Sijl M , Verweij G ( 2004 ) Record linkage of hospital discharge register with population register: experiences at Statistics Netherlands . Stat J UN Econ Comm Eur ; 23 - 32 .
25. Paas GR , Veenhuizen KC ( 2002 ) Research on the validity of the LMR . Utrecht: Prismant.
26. ICD9 website . Available: http://icd9cm.chrisendres. com/index.php?action = contents . 2009 . Accessed 2011 Sept 21.
27. ICD10 website . Available: http://apps.who.int/classifications/apps/icd/ icd10online/. 2009 . Accessed 2011 Sept 21.
28. National Emphysema Treatment Trial (NETT) Research Group ( 2001 ) Patients at high risk of death after lung-volume-reduction surgery . N Engl J Med 345 : 1075 - 83 .
29. Hansell DM , Bankier AA , MacMahon H , McLoud TC , Mu ller NL , Remy J ( 2008 ) Fleischner Society: glossary of terms for thoracic imaging . Radiology 246 ( 3 ): 697 - 722 .
30. Mayne T , Whalen E , Vu A ( 2006 ) Annualized was found better than absolute risk reduction in the calculation of number needed to treat in chronic conditions . Journal of Clinical Epidemiology 59 ( 3 ); 217 - 223 .
31. Gondrie MJ , van der Graaf Y , Jacobs PC , Oen AL , Mali WP ; On behalf of the PROVIDI Study Group ( 2010 ) The association of incidentally detected heart valve calcification with future cardiovascular events . Eur Radiol May ; 21 ( 5 ): 963 - 73 .
32. Tiddens H , Silverman M , Bush A ( 200 ) The role of inflammation in airway disease: remodeling . Am J Respir Crit Care Med Aug 162 ( 2 Pt 2 ): S7 - S10 .
33. Patel BD , Coxson HO , Pillai SG , Agust AG , Calverley PM , Donner CF , et al. ( 2008 ) Airway wall thickening and emphysema show independent familial aggregation in chronic obstructive pulmonary disease . Am J Respir Crit Care Med Sep 1 178 ( 5 ): 500 - 5 .
34. Nakano Y , Muller NL , King GG , Niimi A , Kalloger SE , Mishima M , et al. ( 2002 ) Quantitative assessment of airway remodeling using high-resolution CT . Chest Dec 122 ( 6 Suppl): 271S - 275S .
35. Ahmad K , Gandhi MS , Smith WC , Weber KT ( 2008 ) Bilateral pleural effusions in congestive heart failure . Am J Med Sci 335 ( 6 ): 484 - 8 .
36. Ngom A , Dumont P , Diot P , Lemarie E ( 2001 ) Benign mediastinal lymphadenopathy in congestive heart failure . Chest 119 ( 2 ): 653 - 6 .