A guide to clinical epidemiology for radiologists: Part II statistical analysis

Computerised tomography in the staging of Hodgkin's disease and non-Hodgkin's lymphoma

The last 25 years have seen major changes in the imaging investigation and subsequent management of patients with Hodgkin's disease (HD) and non-Hodgkin's lymphoma (NHL); accurate staging is vital for prognostication and treatment in both, and particularly in HD. The choice of imaging modality for staging depends on its accuracy, impact on clinical decision-making, and availability. Modern CT scanners fulfil most of the desired criteria. The advent of CT scanning, along with the development of ever more effective chemotherapeutic regimens, has resulted in the virtual demise of bipedal lymphangiography (LAG) as a staging tool in patients with lymphoma. It has rendered superfluous a battery of other tests that were in routine use. This contribution reviews the evidence for the use of CT in preference to LAG. CT accurately depicts nodal enlargement above and below the diaphragm, has variable sensitivity for intra-abdominal visceral involvement and is generally outstanding in depicting the extent of disease, especially extranodal extension. Despite the advances in CT technology, there are still areas where CT performs less well (e.g. disease in normal-sized lymph nodes, splenic and bone marrow infiltration). The influence of technical factors, such as the use of intravenous contrast medium, is discussed. In some instances, CT is not the imaging modality of choice and the place of newer techniques such as MRI and endoscopic ultrasound will be reviewed.

Bronchiolitis obliterans syndrome in heart-lung transplant recipients: diagnosis with expiratory CT

PURPOSE

To determine the test performance and longitudinal evolution of air trapping for diagnosing bronchiolitis obliterans syndrome (BOS).

MATERIALS AND METHODS

Over 7 years, 111 combined inspiratory and expiratory computed tomographic examinations were performed in eight healthy control subjects and 38 heart-lung transplant recipients. Functional impairment was assessed with the BOS classification. Receiver operating characteristic (ROC) analysis was performed to determine the optimal threshold of air trapping to distinguish between patients with and those without BOS and to compute sensitivity and specificity for diagnosing BOS.

RESULTS

The extent of air trapping increased with BOS severity (P =.001). A threshold of 32% of air trapping is optimal for distinguishing between patients with and those without BOS and provides a sensitivity of 83%, a specificity of 89%, and an accuracy of 88%. The prevalence of BOS and positive predictive value of air trapping increased with postoperative time, but the negative predictive value of air trapping remained high throughout the study. Patients without BOS who had air trapping exceeding 32% of the parenchyma were at significantly increased risk of developing BOS (P =.004).

CONCLUSION

At the threshold of 32%, air trapping is sensitive, specific, and accurate for diagnosing BOS. Patients with air trapping below 32% are unlikely to have BOS. Air trapping exceeding 32% may be an early indicator of future BOS.

Accuracy of axillary MR imaging in treated breast cancer for distinguishing between recurrent tumour and treatment effects: does intravenous Gd-DTPA enhancement help in cases of diagnostic dilemma?

AIM

To evaluate the sensitivity and specificity of axillary magnetic resonance imaging (MRI) in symptomatic patients, who had previously been treated for breast cancer, compared with clinical outcome after a minimum of 1 year.

METHODS

One hundred and five patients underwent axillary MRI examinations and were diagnosed as axillary tumour, metastatic tumour, treatment effect or normal.

RESULTS

At MRI, 48 patients had axillary tumour, 51 had metastatic tumour (37 had both), 27 had treatment effect and 22 were normal. At outcome (median follow-up, 484 days), 54 patients were positive for axillary tumour, 59 for metastatic disease (40 had both), 21 had treatment effect alone and 18 were clear. Magnetic resonance imaging showed 89% sensitivity, 100% specificity and 94% accuracy for recurrent axillary tumour, and 85% sensitivity, 98% specificity and 90% accuracy for metastatic tumour. Soft tissue plaques were the commonest axillary disease pattern seen (37). Small volume soft tissue plaques gave the most diagnostic difficulty. Non-dynamic enhancement with intravenous Gadopentetate dimeglumine (Gd-DTPA) in a subset of 34 patients improved sensitivity for axillary tumour from 40 to 74%, and improved diagnostic confidence in 11 patients (32%). Magnetic resonance imaging had a positive management impact leading to treatment alteration in 45 patients, 43 of whom had recurrent axillary and/or metastatic tumour.

CONCLUSIONS

Tumour plaques were the commonest pattern of recurrent axillary disease. Forty-eight percent of the patients had metastatic deposits identified by MRI. Magnetic resonance imaging had excellent specificity (100%) and good sensitivity (89%) for recurrent axillary tumour compared with outcome at 1 year, which was improved by non-dynamic administration of Gd-DTPA in 32% of the subset who received it.

Comparison of plain radiographs and magnetic resonance images in the evaluation of periosteal reaction and osteoid matrix in osteosarcomas

To determine the relationship and the degree of agreement between radiographs and MR images for the existence of osteoid matrix and periosteal reactions in the initial diagnosis of osteosarcomas, the plain radiographs and MR studies of 54 patients with proven osteosarcoma were retrospectively evaluated. In each tumor the visualization and type of osseous matrix, periosteal reaction and Codman angle were recorded independently for both techniques and by consensus between two radiologists. In 37 tumors agreement existed regarding osteoid matrix and in 31 cases regarding periosteal reactions. The Kappa statistic showed a significant relationship between both tests (0.49 and 0.44, respectively). Both techniques were also not statistically different in the proportion of findings with the McNemar test. Therefore, the ability of MR images seems important in reporting the MR features of bone tumors. Identification of osteoid mineralization and periosteal reaction can also be used with MR in the diagnosis of osteosarcoma.

Pulmonary nodules: experimental and clinical studies at low-dose CT

PURPOSE

To compare the number of pulmonary nodules detected at helical low- and standard-dose computed tomography (CT) and to investigate the diagnostic value of low-dose CT with a radiation exposure equivalent to that used at chest radiography.

MATERIALS AND METHODS

Two radiologists recorded pulmonary nodules at standard-dose (250 or 100 mA, pitch of 1; 200 mA, pitch of 2) or low-dose CT (50 or 25 mA, pitch of 1 or 2) in five postmortem specimens and 75 patients. Nodules were assessed by size (5 mm or smaller, 6-10 mm, or larger than 10 mm) and by diagnostic confidence ("definite nodule," "definite lesion, not classic nodule," or "questionable lesion, possibly representing a vessel") with the Wilcoxon signed rank test. Artifacts depicted at low-dose CT were recorded.

RESULTS

There were no statistically significant differences in the number of nodules detected at standard- or low-dose CT except in nodules 5 mm or smaller that were assessed as definite nodules at standard- or low-dose CT (25 mA, pitch of 2) (472 vs 397, P < .05). Artifacts that possibly interfered with nodule detection were observed exclusively at CT with 25 mA and a pitch of 2.

CONCLUSION

Pulmonary nodules were detected reliably at CT with 50 mA and pitch of 2 or with 25 mA and a pitch of 1. However, further reduction of the dose to that used at chest radiography was associated with a significant decrease in the number of nodules 5 mm or smaller that were detected, possibly due to artifacts.

Comparative evaluation of conventional and dynamic magnetic resonance imaging of the pituitary gland for the diagnosis of Cushing's disease

OBJECTIVES

The ability of MRI to detect pituitary ACTH-secreting adenomas in patients with Cushing's disease is limited. Owing to different dynamics of contrast enhancement between adenomas and normal pituitary tissue, it has been suggested that obtaining images within seconds after gadolinium (Gad) injection using dynamic procedures increases the sensitivity of MRI in the detection of pituitary microadenomas. The objective of this study was to compare the ability of conventional magnetic resonance imaging (CMRI) and dynamic MRI (DMRI) to detect ACTH-secreting pituitary adenomas.

DESIGN

Twenty-six consecutive patients with ACTH-dependent Cushing's syndrome and 10 normal subjects were investigated. According to the results of inferior petrosal sinus sampling, 21 patients had Cushing's disease and five had ectopic ACTH syndrome. Patients with Cushing's disease were operated regardless of the results of imaging studies. All underwent identical MRI and DMRI procedures using a 1.0 T magnet. Image sampling time during DMRI was 19 sec. Scans were randomly mixed and analysed blind, retrospectively and independently by two experienced radiologists. The clarity of the images was assessed by the analysis of agreement among radiologists. MRI findings were compared to surgical and histopathological findings.

RESULTS

Surgical exploration identified three macrodenomas and 14 microadenomas. One microadenoma was found at pathological examination after subtotal hypophysectomy and no tumour was found in three cases. According to the combined opinion of radiologists, the three macroadenomas were identified equally well with CMRI and DMRI. Eight ACTH-secreting microadenomas were detected with CMRI and 11 with DMRI. The three microadenomas detected with DMRI only were visualized within 60 sec following Gad injection. No false positives occurred with CMRI. Three false positives were obtained with DMRI: one in a patient with ectopic ACTH syndrome while a silent microprolactinoma and normal tissue were found at the site of the radiological abnormality in two patients with Cushing's disease. In our study, the sensitivity of DMRI is greater than that of CMRI (0.67 vs. 0.52) but is associated with a loss in specificity (0.80 vs. 1.00). False positives may result from the increased sensitivity of DMRI which detects incidental pituitary lesions, technical artefacts or lowest clarity of images, as suggested by a lower observer agreement of DMRI (Kappa statistic 0.66 vs. 0.83). Overall, the two MR procedures had equivalent diagnostic power (0.72).

CONCLUSIONS

In our hands, dynamic procedures did not improve the usefulness of MRI in Cushing's syndrome.

Comparison of two oral evacuants (Citromag and Golytely) for bowel preparation before barium enema

Oral regimens for bowel preparation before barium enema examination are routinely used because of their convenience and simplicity, rather than the traditional method of colonic wash-out. We performed a prospective study comparing the side-effects and efficacy of two commonly used oral bowel evacuants (Citromag and Golytely) for bowel preparation before barium enema examination in 102 patients. The side-effects associated with the agents were assessed by analysing a questionnaire completed by the patients. There was no significant difference in the side-effects between the two agents although more patients taking Golytely (45.5%) deemed its taste unacceptable than those taking Citromag (25.9%). The results of the bowel preparation were assessed by two independent radiologists giving scores on the amount of faecal residue and the quality of mucosal coating. The amount of faecal residue was less in the Golytely group (p < 0.05). The quality of mucosal coating by barium was also better in the Golytely group than the Citromag group (p < 0.05).

A review of the statistical analysis used in papers published in Clinical Radiology and British Journal of Radiology

Statistical analysis such as significance testing have become essential features of published medical studies. This has resulted in an increased frequency with which statistics are used, making the interpretation of scientific publications more difficult. There is an extensive array of tests and techniques. The aim of this study is to identify which statistical tests are used in radiology publications. All major articles published in Clinical Radiology and British Journal of Radiology in one year were reviewed. The frequency of statistical methods used was as follows: no statistical method or descriptive statistics only 103 (47%), one type of statistical method 67 (31%), and two or more methods 47 (22%). Statistics dealing with basic inference, decisions, contingency tables or correlation/regression techniques were found in 124 (53%) in which a procedure had been used. Advanced statistics including receiver operating characteristics (ROC), odds ratio, regression techniques, multiway ANOVA, and nonparametric ANOVA studies accounted for only 41 (19%) in which a procedure had been used. We conclude that descriptive analysis and basic statistical techniques account for most of the statistical tests reported. Physicians should concentrate on improving their understanding of basic statistics but advice should be sought from professionals in the fields of biostatistics and epidemiology as to whether the use of more advanced techniques would be more appropriate.