Thoughtful assessment of a patient before an anesthetic event is crucial to creating a safe and effective anesthetic plan. Preanesthetic assessment consists of reviewing the patient's medical records, acquiring a thorough patient history, performing a physical examination, and performing any other tests or evaluations pertinent to the planned procedure.1 A thorough physical exam is the most important part of preoperative assessment and should be performed on all patients before an anesthetic event. Evaluation of bloodwork before an anesthetic event to screen for systemic derangements is also a very common practice, although this recommendation is still somewhat controversial.1–3 Based on physical exam findings and bloodwork results, other diagnostics may be recommended for a patient before undergoing general anesthesia.
Some clinicians recommend or require preoperative thoracic radiographs before anesthesia, especially in older patients. This recommendation is to screen for any abnormalities that may affect the anesthetic plan and/or change the plan to proceed with surgery. There are currently no data available to support this recommendation in veterinary medicine. The American Animal Hospital Association (AAHA) Anesthesia and Monitoring Guidelines for Dogs and Cats only recommend a physical exam and bloodwork within 3 to 6 months for patients undergoing anesthesia who have no risk factors or specific concerns from history or exam.1 The AAHA Senior Care Guidelines for Dogs and Cats do not recommend screening thoracic radiographs for older animals unless specifically clinically indicated.4 In human medicine, preoperative thoracic radiographs are only recommended if patient or clinical factors indicate a specific risk factor that makes it likely to yield an abnormal result.5 Furthermore, routine preoperative tests in general are not recommended in people, yet it is recommended that preoperative diagnostics are performed on a selective basis.5
The purpose of this study was to evaluate the utility of thoracic radiographs as a routine preoperative diagnostic in dogs without identified risk factors for or indications of intrathoracic disease. We hypothesized that there would not be a large number of clinically relevant abnormalities found on preoperative radiographs. We also hypothesized that there would be significant agreement in the interpretation of thoracic radiographs between the clinician (surgeon) and radiologist.
Methods
This is a prospective cross-sectional study of dogs that underwent an elective tibial plateau leveling osteotomy across 3 hospitals within the Chesapeake Veterinary Surgical Specialists practice group between January 1, 2022, and December 31, 2022. All dogs of any age that had a tibial plateau leveling osteotomy during this time period were eligible for inclusion. Clients completed a consent form and a questionnaire (Supplementary Material S1) used to screen for any history of cardiopulmonary disease or neoplasia. A 3-view thoracic radiographic study, including right lateral, left lateral, and either dorsoventral or ventrodorsal projections, was performed for each patient before anesthesia. Patients were excluded if they had any history or clinical signs suggestive of intrathoracic or cardiopulmonary abnormalities or disease, history or clinical signs of malignant neoplasia, incomplete client consent form, incomplete disease history questionnaire, incomplete 3-view thoracic radiographic study, or incomplete clinician assessment of radiographs. Clinical signs consistent with intrathoracic or cardiopulmonary disease included heart murmur, arrhythmia, increased respiratory effort, cough, auscultated crackles or wheezing, dyspnea, or any other abnormal respiratory pattern consistent with pulmonary disease.
Medical records of all patients who were eligible for inclusion were reviewed for exclusion criteria. Patients were excluded from the study if their medical records noted cardiopulmonary abnormalities, malignant neoplasia, or diagnosis of any thoracic structural abnormalities or diseases. For those included, when available in the record, data recorded from the day of radiographic assessment included signalment, heart rate, respiratory rate, temperature, body weight, owner-reported history, physical examination abnormalities, abnormalities in CBC, and abnormalities in serum biochemistry profile.
All radiographs performed for this study were initially evaluated by the attending board-certified or residency-trained veterinary surgeon and an assessment/interpretation of the radiographs was written for each case. All radiographs were later evaluated by a single board-certified veterinary radiologist (EKR) for the purpose of this study and an assessment/interpretation of the radiographs was written for each case. The radiologist was blinded to the surgeons’ radiographic findings. Radiographic abnormalities were classified as those pertaining to the cardiac silhouette, pulmonary parenchyma and pleural space, or extrathoracic structures. Radiographic abnormalities were then assessed for significance. A clinically significant radiographic finding was defined as one that warrants further investigation, such as repeat or follow-up diagnostics, but the degree of its significance is unclear. Examples of clinically significant findings are cardiomegaly and hepatomegaly. A surgically significant radiographic finding was defined as one that the surgical team interpreted as having such an impact on anesthetic safety and/or prognosis that we would recommend not proceeding with this elective surgical procedure. Examples of surgically significant findings include suspected neoplasia and pneumonia.
Statistical analysis
Descriptive statistics of patient variables were computed. Categorical variables were described with numbers, percentages, and 95% CI. Confidence intervals were calculated using an online calculator.6 To evaluate the association between patient age and the probability of a case having a surgically significant abnormality found via thoracic radiographs, a logistic regression model with a binary distribution and logit link was fitted. Univariable logistic regression models assessing the association between age, sex/intact status, breed, and presence of abnormal blood work with the probability of a case having a clinical or surgically significant abnormality diagnosed by thoracic radiographs were then fitted. Age was assessed for linearity due to its continuous distribution using a locally weighted regression of the outcome on a logit scale on age. If it was not linearly associated with the modeled outcome, it was categorized, and the logistic model was reevaluated. To estimate agreement between surgeons and radiologists beyond that due to chance, the Cohen κ value was calculated for each abnormality type. All analyses were performed with the R language,7 version 4.3.2, and the significance threshold for all models was set at P ≤ .05. The estimates of the logistic regression models were presented as ORs and mean percentages, with their corresponding 95% CI.
Results
A total of 866 cases were presented for elective tibial plateau leveling osteotomy during the study period. A total of 281 cases met the inclusion criteria for the study. There were 148 spayed females, 7 intact females, 120 castrated males, and 6 intact males. There were 45 different breeds and their mixes in this population. The most common breeds were Labrador Retrievers, pit bull–type dogs, mixed-breed dogs, Poodles and Poodle mixes, Golden Retrievers, Mastiffs, and Boxers. The median weight of this population was 29.8 kg (range, 4.5 to 76 kg). The median age of this population was 7.1 years (range, 1 to 14.1 years). Preanesthetic CBC and serum biochemistry profiles were performed in 280 cases. Of those cases with bloodwork performed, 57 (20%) had at least 1 clinical pathology abnormality and 7 (2.5%) had more than 1 abnormality. Ninety-six dogs (34.2%) had a radiographic abnormality noted. The other 185 dogs (65.8%) did not have any radiographic abnormalities noted. Twenty-four dogs (8.5%) had cardiovascular abnormalities noted on thoracic radiographs. The most common abnormalities were generalized enlargement of the cardiac silhouette (15 dogs) and bulges around the cardiac silhouette on the ventrodorsal or dorsoventral projection (8 dogs). Thirty-three dogs (11.7%) had pulmonary abnormalities noted on thoracic radiographs. The most common pulmonary abnormality was soft tissue nodules (18 dogs), which included suspected neoplastic pulmonary nodules, likely end-on vessels, and nodules most consistent with subcutaneous masses or nipples. Seven dogs had a bronchial pulmonary pattern, and 1 dog had a collapsed lung lobe without associated clinical signs. Fifty-four dogs (19.2%) had extrathoracic abnormalities noted on thoracic radiographs. The most common extrathoracic findings were material in the lumen of the stomach (n = 16 dogs), hepatomegaly (11), and microhepatica (9).
Thirty dogs (10.7%) were determined to have at least 1 clinically significant radiographic abnormality (Table 1). Six dogs had more than 1 abnormality determined to be clinically significant. One dog had 3 abnormalities determined to be clinically significant. For 10 of the 281 (3.6%) dogs, radiographic examination revealed surgically significant abnormalities: suspected neoplastic pulmonary nodules (n = 5 [1.8%]), cranial abdominal masses (5 [1.8%]), and collapsed lung (1 [0.4%]). One of these 10 dogs had a pulmonary nodule and a cranial abdominal mass. There was no statistical association between sex, intact status, and the presence of an abnormality on bloodwork and finding a clinically or surgically significant radiographic abnormality (Table 2). The mean age and weight of dogs diagnosed with a radiographic abnormality were 7.5 years and 29 kg, respectively. The mean age of dogs diagnosed with a surgically significant radiographic abnormality was 8.5 years. There was a borderline but not statistically significant association between age and the probability of diagnosing a clinically significant radiographic abnormality (P = .06). There was no statistical association between patient age and the probability of diagnosing a surgically significant radiographic abnormality (P = .2). Six dogs (2.1%) had a surgically significant abnormality found in the thorax, and 5 dogs (1.8%) had a surgically significant abnormality found in the abdomen on thoracic radiographs.
Clinically significant abnormalities found on preoperative screening thoracic radiographs in a population of dogs without history or physical exam findings suggestive of thoracic disease undergoing a routine tibial plateau leveling osteotomy procedure over a period of 1 year.
| Abnormality | Number of cases | Percentage of population (95% CI) |
|---|---|---|
| Pulmonary soft tissue nodule | 15 | 5.3 (3.0–8.7) |
| Bulge of the cardiac silhouette | 6 | 2.1 (0.8–4.6) |
| Cranial abdominal mass | 5 | 1.8 (0.6–4.1) |
| Material in gastric lumen | 3 | 1.1 (0.2–3.1) |
| Alveolar pulmonary pattern | 3 | 1.1 (0.2–3.1) |
| Hiatal hernia | 2 | 0.7 (0.1–2.6) |
| Generalized enlargement of the cardiac silhouette | 2 | 0.7 (0.1–2.6) |
| Collapsed lung lobe | 1 | 0.4 (0–2.0) |
Total number of cases and percentage of the total study population with 95% CI are reported.
Model adjusted mean percentage, OR, and P value of the association of sex, intact status, and bloodwork abnormalities and the probability that an abnormality is surgically or clinically significant for the population of dogs as described (Table 1).
| Abnormality diagnosed by a radiologist is surgically significant | Abnormality diagnosed by a radiologist is clinically significant | |||||
|---|---|---|---|---|---|---|
| Model-adjusted mean percentage (95% CI) | OR (95% CI) | P value | Model-adjusted mean percentage (95% CI) | OR (95% CI) | P value | |
| Sex | .22 | .38 | ||||
| Female intact | 33.3 (4.3–84.6) | Ref | 66.7 (15.4–95.7) | Ref | ||
| Female spay | 10.7 (4.9–21.9) | 0.2 (0.02–5.6) | 33.9 (22.8–47.2) | 0.3 (0.01–2.8) | ||
| Male intact | 50.0 (5.9–94.1) | 2.0 (0.04–120.3) | 50.0 (5.9–94.1) | 0.5 (0.008–23.5) | ||
| Male neuter | 5.7 (1.4–20.2) | 0.1 (0.007–3.2) | 22.9 (11.9–39.5) | 0.1 (0.006–1.7) | ||
| Bloodwork | .43 | .72 | ||||
| Abnormal | 14.3 (5.5–32.4) | 1.7 (0.4–6.6) | 28.6 (15.0–47.6) | 0.83 (0.3–2.1) | ||
| Normal | 8.8 (4.0–18.3) | Ref | 32.4 (22.3–44.3) | Ref | ||
There was poor agreement between the surgeon and the radiologist for identifying cardiovascular (κ = 0.04, P = .36) and pulmonary (κ = −0.04, P = .37) abnormalities. There was a slight agreement between the surgeon and the radiologist for identifying extrathoracic abnormalities (κ = 0.11, P = .01). The radiographic findings of the surgeon and radiologist differed 9.6%, 13.9%, and 18.9% of the time for cardiovascular, pulmonary, and extrathoracic abnormalities, respectively. The surgeon did not identify a lesion that was identified by the radiologist 95.8%, 100%, and 90.7% of the time for cardiovascular, pulmonary, and extrathoracic abnormalities, respectively. Assuming the findings of the radiologist are accurate, these would represent false negative rates or miss rates of the surgeon evaluating thoracic radiographs. The surgeon identified a lesion that was not identified by the radiologist 1.5%, 2.4%, and 1.7% of the time for cardiovascular, pulmonary, and extrathoracic abnormalities, respectively. Assuming the findings of the radiologist are accurate, these would represent false positive rates or false alarm rates of the surgeon evaluating thoracic radiographs.
Discussion
Our study found few abnormalities on preoperative screening thoracic radiographs that would change the plan to proceed with elective surgery in this cohort of dogs without history or physical exam findings suggestive of thoracic or cardiopulmonary disease. Some abnormality was found in 34% of cases, but many of these findings were not determined to be clinically significant, such as end-on vessels. Our study found that 10.7% of cases had abnormalities determined to be clinically significant. Furthermore, in 3.6% of cases, an abnormality was found that was likely to affect the decision to proceed with surgery. The aim of this study is not to make a recommendation regarding screening preoperative radiographs but to provide data to clinicians to help them make their own informed decision. As we hypothesized, the utility of preoperative screening radiographs in dogs with no clinical indications for this test appears to be low.
Few previous studies in veterinary medicine have evaluated the utility of thoracic radiographs in patients without specific indications for this test. In a study8 of a cohort of dogs admitted to a tertiary-care veterinary hospital for noncardiopulmonary disease, 27.7% had some abnormality found on screening thoracic radiographs. However, only 4.8% of cases had abnormalities found that resulted in an alteration in the clinical plan. A study9 evaluating the utility of screening thoracic radiographs in dogs with immune-mediated hemolytic anemia found abnormalities on thoracic radiographs in 32% of dogs. In 20% of cases, the findings changed the clinician's diagnostic and treatment plans; however, many of those abnormalities were not related to immune-mediated hemolytic anemia, and some of those dogs had history or exam abnormalities suggestive of cardiopulmonary disease. In a study10 of preoperative thoracic radiographs in dogs with gastric dilatation with volvulus, the most common reason for thoracic radiographs was screening for unsuspected underlying disease. The presence of incidental findings was low, with pulmonary nodules found in 4% of cases. That study did find evidence of aspiration pneumonia in 14% of cases, but many of those dogs had exam findings suggestive of pulmonary disease. In addition, that study found a relationship between radiographic cardiomegaly and outcomes, suggesting that preoperative thoracic radiographs may be warranted in this population. In human medicine, abnormalities on chest radiographs in asymptomatic or nonselected patients have been reported in between 0.3% and 60.1% cases.5 Of cases with an abnormality found, that abnormality led to postponement of the procedure or changes in case management in 0.6% to 20.3% of cases.5 Overall, these studies would support our findings that preoperative screening radiographs have low utility without a specific indication. Furthermore, in human medicine, the American Society of Anesthesiologists Task Force on Preanesthesia Evaluation produced a practice advisory for preanesthesia evaluation, which only recommends preoperative thoracic imaging when the patient has history or exam abnormalities or specific risk factors that make finding an abnormality likely.5
This study did not find that older dogs had a higher incidence of clinically significant radiographic abnormalities or abnormalities that would affect the decision to proceed with surgery. The median age of dogs with an abnormality on thoracic radiographs was approximately 8 years, but our data do not support using this as a criterion or cutoff for recommending screening thoracic radiographs. It is possible that an association with age would have been found if a different or larger patient population had been evaluated. The recommendation for screening preoperative radiographs in geriatric dogs is based on the perceived higher risk of underlying disease, such as neoplasia. Although older patients are suspected to be more likely to have some underlying disease process compared to younger patients, this increased risk has not been well evaluated or associated with a specific age. Furthermore, geriatric dogs have not been shown to have a higher risk of specifically cardiopulmonary or thoracic disease compared to disease in other systems or parts of the body. Whereas more thorough screening may be logical in older patients, there are not sufficient data to suggest more than a thorough physical assessment and screening bloodwork is warranted.2,11 The AAHA Senior Care Guidelines for Dogs and Cats recommend annual CBC, biochemistry profile, T4 level, N-terminal pro-B–type natriuretic peptide, urinalysis, fecal testing for parasites, tick-borne testing, and blood pressure measurement for dogs in the latter 25% of their expected lifespan.4 Imaging of a region is only recommended if clinically indicated or if warranted based on breed-specific variables. It would appear that if no specific indication for thoracic imaging is present on these recommended diagnostics, then there is not a clear indication for preanesthetic screening thoracic imaging even in an older patient. If the pretest probability of a specific diagnosis is very low based on history and clinical signs, then the posttest probability of making that diagnosis will also be low.12
Although this study evaluated thoracic radiographs as a screening tool for cardiopulmonary and thoracic abnormalities that would affect the decision to proceed with surgery, we found nearly as many surgically significant abnormalities in the included cranial abdomen as we did in the entirety of the thorax. In this study, thoracic radiographs were not performed because of any clinical factors but rather based on common preoperative recommendations. It has been reported that most decisions to perform screening thoracic radiographs in dogs without a specific indication are based on the personal preference and discretion of the clinician and not based on owner request or patient factors.10 These recommendations are therefore arbitrary as there is not a patient risk factor or clinical finding to suggest disease in this particular anatomic site. These data suggest that there may be a similar or possibly larger number of significant abnormalities found if abdominal imaging was performed for preoperative screening. If one were to consider including imaging as a preoperative screening tool, abdominal radiographs, ultrasound, or CT may serve as an equal or higher yield diagnostic. A study13 evaluating screening abdominal ultrasound before pursuing advanced neurologic diagnostic tests showed that only 1.3% of dogs did not have advanced neurodiagnostics because abnormalities were found on ultrasound. Further studies on the utility of abdominal imaging in dogs with no clinical indication have not been evaluated in veterinary medicine.
In this study, there was rather poor agreement between the radiographic findings of the surgeon and that of the radiologist. More specifically, there were very few abnormalities found by the radiologist that were also noted by the surgeon. The superior detection accuracy of a veterinary radiologist compared to general practitioners has previously been reported.14 Another study15 showed improved performance and less interobserver variation with increased specialty radiology training and experience. These findings highlight the importance of routine review of radiographs by a board-certified radiologist.
In 5% of cases in this study, the surgeon described an abnormality that was not noted by the radiologist. These cases either represent false positives by the surgeon or false negatives by the radiologist. Follow-up confirmatory diagnostics were not performed. It is likely that at least some of these findings represent false positives. False positives in preoperative screening radiographs can lead to potentially unnecessary follow-up testing and/or a delay in the planned surgical procedure. This could lead to prolonged patient discomfort and dysfunction and added client costs. The potential for false positives has long been used as one rationale to advise against screening diagnostics.16–18 The excessive use of imaging can lead to false positive diagnoses, detection of incidental findings, and overdiagnosis, which all have potential negative effects on the patient.10 Alternatively, early detection of an abnormality with significant health and prognostic implications can allow for early treatment, which can positively affect the patient. The relative risks and benefits of these diagnostics should be considered for each individual patient when deciding to perform screening diagnostics.
The main limitation of this study is the low sample size, which could lead to a type II statistical error. It is possible that with a larger patient population, a higher percentage of significant abnormalities would be found. It is also possible that with a larger sample population, risk factors for finding an abnormality would be elucidated. The prospective unblinded nature of reviewing radiographs as part of this study could have introduced bias or affected how the radiographs were evaluated, interpreted, or reported. Another limitation is the classification of the abnormal findings into clinically and surgical significant categories based on the subjective assessment of the authors. It is possible that other veterinarians would classify the abnormalities differently.
Overall, we found low numbers of clinically significant abnormalities on preoperative screening thoracic radiographs, with very few findings being surgically impactful. It is for each individual veterinarian to interpret the results of this study and decide if it appears warranted to recommend routine preoperative screening thoracic radiographs in healthy patients without specific risk factors or clinical signs. We did not find age or any other patient factors as a risk factor that would increase the chance of finding an abnormality on screening thoracic radiographs. Further prospective studies evaluating screening preoperative thoracic radiographs in patients with and without specific pathologies are necessary to further define the utility of this practice and to make specific recommendations for or against its use in veterinary patients. We also found poor agreement between the findings of the primary clinician and the radiologist. This suggests that clinicians should consider a secondary review of screening thoracic radiographs by a radiologist.
Supplementary Materials
Supplementary materials are posted online at the journal website: avmajournals.avma.org.
Acknowledgments
The authors thank Drs. Andrea Dixon and Natalia Cernicchiaro with the Center for Outcomes Research and Epidemiology at the College of Veterinary Medicine at Kansas State University for their assistance in performing statistical analyses.
Disclosures
The authors have nothing to disclose. No AI-assisted technologies were used in the composition of this manuscript.
Funding
The authors have nothing to disclose.
ORCID
C. M. Gauthier https://orcid.org/0009-0002-4801-9715
References
- 1.↑
Grubb T, Sager J, Gaynor JS, et al. 2020 AAHA anesthesia and monitoring guidelines for dogs and cats. J Am Anim Hosp Assoc. 2020;56(2):59–82. doi:10.5326/JAAHA-MS-7055
- 2.↑
Joubert KE. Pre-anesthesia screening of geriatric dogs. J S Afr Vet Ass. 2007;78(1):31–35. doi:10.4102/jsava.v78i1.283
- 3.↑
Mitchell K, Barletta M, Quandt J, Shepard M, Kleine S, Hofmeister E. Effect of routine pre-anesthesia laboratory screening on pre-operative anesthesia-related decision-making in healthy dogs. Can Vet J. 2018;59(7):773–778.
- 4.↑
Dhaliwal R, Boynton E, Carrera-Justiz S, et al. 2023 AAHA senior care guidelines for dogs and cats. J Am Anim Hosp Assoc. 2023;59(1):1–21. doi:10.5326/JAAHA-MS-7343
- 5.↑
Apfelbaum JL, Connis RT, Nickinovich DG, et al. Practice advisory for preanesthesia evaluation: an updated report by the American Society of Anesthesiologists Task Force on Preanesthesia Evaluation. Anesthesiology. 2012:116(3):522–538 doi:10.1097/ALN.0b013e31823c1067
- 6.↑
Kohn MA, Senyak J. Sample size calculators. Clinical & Translational Science Institute, University of California-San Francisco. September 28, 2024. Accessed December 29, 2024. https://sample-size.net/confidence-interval-proportion/
- 7.↑
R Core Team. R: a Language and Environment for Statistical Computing. R Foundation for Statistical Computing. 2023. Accessed April 8, 2024. https://www.R-project.org/
- 8.↑
Keyserling CL, Buriko Y, Lyons BM, Drobatz KJ, Fischetti AJ. Evaluation of thoracic radiographs as a screening test for dogs and cats admitted to a tertiary-care veterinary hospital for noncardiopulmonary disease. Vet Radiol Ultrasound. 2017;58(5):503–511. doi:10.1111/vru.12519
- 9.↑
Andres M, Hostnik E, Green E, et al. Diagnostic utility of thoracic radiographs and abdominal ultrasound in immune-mediated hemolytic anemia. Can Vet J. 2019;60(10):1065–1071.
- 10.↑
Green JL, Cimino Brown D, Agnello KA. Preoperative thoracic radiographs findings in dogs presenting for gastric dilatation-volvulus (2000–2010): 101 cases. J Vet Emerg Crit Care (San Antonio). 2012;22(5):595–600.
- 11.↑
Díaz M, Kaartinen J, Allison A. Preanesthetic blood tests in cats and dogs older than 8 years: anesthetists’ prediction and peri-anesthetic changes. Vet Anaesth Analg. 2021;48(6):854–860. doi:10.1016/j.vaa.2021.04.010
- 12.↑
Lamb CR. Veterinary diagnostic imaging: probability, accuracy and impact. Vet J. 2016;215:55–63. doi:10.1016/j.tvjl.2016.03.017
- 13.↑
Tong NM, Zwingenberfer AL, Blair WH, Taylor SL, Chen RX, Sturges BK. Effect of screening abdominal ultrasound examination on the decision to pursue advanced diagnostic tests and treatment in dogs with neurologic disease. J Vet Intern Med. 2015;29(3):893–899. doi:10.1111/jvim.12602
- 14.↑
Reese DJ, Green EM, Zekas LJ, et al. Intra- and interobserver variability of board-certified veterinary radiologists and veterinary general practitioners for pulmonary nodule detection in standard and inverted display mode images of digital thoracic radiographs of dogs. J Am Vet Med Assoc. 2011;238(8):998–1003. doi:10.2460/javma.238.8.998
- 15.↑
Froes TR, Zwingenberger AL, Sato A, et al. Interobserver agreement in interpretation of radiographic pulmonary changes in dogs in relation to radiology training. Semin Cien Agr. 2014;35(5):2513–2526. doi:10.5433/1679-0359.2014v35n5p2513
- 16.↑
Joo HS, Wong J, Naik VN, Savoldelli GL. The value of screening preoperative chest x-rays: a systematic review. Can J Anaesth. 2005;52(6):568–574. doi:10.1007/BF03015764
- 17.
Rucker L, Frye EB, Staten MA. Usefulness of screening chest roentgenograms in preoperative patients. JAMA. 1983;250(23):3209–3211. doi:10.1001/jama.1983.03340230061032
- 18.↑
Rees AM, Roberts CJ, Bligh AS, Evans KT. Routine preoperative chest radiography in non-cardiopulmonary surgery. Br Med J. 1976;1(6021):1333–1335. doi:10.1136/bmj.1.6021.1333
