METHODS

The study group comprised 29 patients attending St Helens and Knowsley Trust Hospitals rheumatology outpatient departments. These patients with FA were identified from a previous study,⁴ when a cross section of 150 patients with RA underwent HRCT scanning, irrespective of the presence or absence of respiratory symptoms. This hospital is based in northwest England and its catchment population is from an industrialised area. The patients were diagnosed with definite RA as defined by the American Rheumatism Association 1987 criteria⁵ and were diagnosed with FA based on HRCT diagnosis.

The research ethics committee of St Helens and Knowsley Hospitals approved the study.

Clinical assessment

A questionnaire was completed for each patient by JKD. This noted duration of RA, extra-articular complications, current and previous disease modifying drugs, corticosteroid use, early morning joint stiffness, and patient assessment of disease activity. Each patient filled in the Modified Stanford Health Assessment Questionnaire to assess functional impairment. Respiratory questions were asked about previous chest disease, cough, dyspnoea, sputum production, chest pain, weight loss, and risk factors for respiratory disease such as smoking, drugs, domestic pets, and occupation. Cigarette consumption was evaluated in pack years (one pack year—20/day for one year). Current smokers were those who had smoked during the previous six months: non-smokers had smoked fewer than 20 packets of cigarettes during their lifetime. A detailed clinical examination was performed. Table 1 shows the patients' details.

All patients had venous blood taken for full blood count, plasma viscosity, renal and liver function, C reactive protein, and plasma proteins. Immunological investigations included immunoglobulins, rheumatoid factor (latex agglutination, positive at 1/20 dilution), and antinuclear antibody. All patients underwent echocardiography, electrocardiography, chest radiography, HRCT, and full pulmonary function testing.

Pulmonary function testing

Lung function was measured by a standard protocol and included spirometry (Sensor Medics water spirometer, California, USA), lung volumes (Sensor Medics helium dilution analyser), and transfer factor measured by single breath diffusing capacity (Sensor Medics autolink breath system). The highest of three reproducible measurements was used, and expressed as the percentage predicted for age, height, and sex according to standardised tables.⁶ One senior technician performed the measurements in the cardiorespiratory department at Whiston Hospital.

Pulmonary function tests were performed within four weeks of the HRCT scan. Full pulmonary function tests were carried out four monthly.

High resolution computed tomography

All the study group underwent chest HRCT scanning, which was performed on a Siemens Somatom hiQ scanner. Scanning time was 1.3 s. Supine and prone views were taken. Serial slices were taken 2 mm in width and 10 mm apart. All images were obtained at window levels appropriate for lung parenchyma settings (window width 1300 HU; window level 600 HU) and mediastinum (window width 350 HU: window level 40 HU). A chest radiograph was taken at the same time as the HRCT scan. The HRCT scans were interpreted and graded for FA by consultant radiologists JD and HEF, who were unaware of the clinical details. Each radiologist reviewed the scans independently of the other and a consensus opinion between JD and HEF was taken in the event of disagreement.

Interpretation of the HRCT scan

A ground glass pattern was defined as a patchy or diffuse increase in lung density that did not obscure pulmonary vasculature. A reticular pattern was defined as the presence of intersecting lines that formed anything from a fine network to frank honeycombing; this was defined as FA if the appearance was thought to be typical of usual interstitial pneumonia.⁷ A κ score for the level of agreement between radiologists on the presence of FA was 0.741. Other lung disease present on HRCT was systematically noted.

The scans were analysed according to the type, distribution, and extent of FA as described by Wells et al.⁸ HRCT was repeated yearly.

Scoring of distribution of FA on initial HRCT scans

The mean extent of abnormal lung was calculated for two levels in the upper part of the lung (the mid-arch of the aorta, the carina) and separately for two levels in the lower part of the lung (the pulmonary venous confluence, 1 cm above the right dome of the diaphragm). The overall distribution of disease was expressed as the ratio of the extent of disease in the upper zone to the extent of disease in the lower zone. For each HRCT scan, the mean figure for the two observers was used,

Extent of FA on initial HRCT scans

HRCT scans were evaluated at five levels (the origin of the great vessels, the mid-arch of the aorta, the carina, the pulmonary venous confluence, and 1 cm above the right dome of the diaphragm) for extent of disease. At each level, the extent of abnormal lung was visually estimated to the nearest 5%. An estimate of overall lung involvement was obtained by applying a weighting factor to correct for differences in lung volumes at each HRCT level

The ratios of the lung volumes for the five levels are 0.129 for level 1, 0.190 for level 2, 0.222 for level 3, 0.228 for level 4, and 0.230 for level 5. The percentage of abnormal lung at each level was multiplied by the corresponding ratio; addition of the five adjusted figures gave an estimation of the overall percentage of abnormal lung. For each HRCT scan, the mean figure for the two observers was used.

Follow up assessments

Clinical

The patients were assessed clinically every three months by JKD. A standard questionnaire was completed on each occasion. The patients were monitored over two years. Pulmonary function tests were repeated four monthly by the same technician. Chest HRCT was repeated at baseline, one, and two years.

Definitions

Criteria used to define significant change in pulmonary function measurements were (a) improvement, a rise of more than 15% from baseline value of the carbon monoxide transfer factor (Tlco); (b) stability, no change or a change of ±15% from the baseline value of Tlco; and (c) deterioration, a decline of >15% from the baseline value of Tlco.⁸

Increasing FA on HRCT was based on HRCT scans examined in sequence. The pattern, extent, and distribution of abnormal HRCT findings were compared with findings in the same region on previous and subsequent HRCT scans.³

Progression of FA was defined as significant deterioration in pulmonary function tests with increased FA demonstrated on HRCT scan, or death as a result of respiratory failure before investigations were repeated.

Statistical analysis

The Mann-Whitney U test was used to compare quantitative data. χ² with Yates's correction was used to compare frequencies. Student's t test was used to compare normally distributed quantitative data.

Potentially significant parameters were tested for possible interrelationship by multiple logistic regression analysis.

Interobserver variability, before consensus agreement, was evaluated for the grading of CT appearance, the assessment of whether the extent of disease had changed at follow up scanning, and the categorisation of which CT pattern had changed in extent. Agreement between observers was expressed as a κ value. Values of κ 0.40–0.75 were taken to indicate fair to good agreement between observers.

Receiver operator curves were used to identify sensitivity and specificity of cut off points for investigations predicting progression.

RESULTS

Clinical events

Twenty nine patients had typical FA on chest HRCT. Three (10%) had a predominantly ground glass pattern, three (10%) had equal predominance of ground glass and reticular pattern, and 23 (79%) presented with a predominantly reticular pattern FA. Twelve (41%) patients had apical emphysematous bullae in addition to FA.

Over two years, 10 (34%) patients showed a significant deterioration in pulmonary function as a result of FA and four (14%) patients died in hospital of FA with respiratory failure. One patient developed lung carcinoma; she was also a cigarette smoker.

Two patients did not continue to attend for serial investigations, two patients died of causes not related to FA—one of ischaemic heart disease, and one of disseminated non-pulmonary malignancy.

HRCT

Two patterns of deterioration were seen, increasing reticular pattern involvement of the lung (five patients) and also sudden deterioration with occurrence of a new ground glass pattern disease (two patients). The mean increase in extent of FA was by 9.4%. Two patients with predominantly ground glass pattern FA had spontaneous improvement in lung involvement without specific treatment.

Treatment of FA

Six patients were treated after significant deterioration in their lung condition. In two patients this was after the two year study period had ended. Three patients were male with a mean (SD) age 61.5 (9.6) years, two had finger clubbing, three had coexistent emphysematous bullae on HRCT, and three had predominantly reticular pattern FA. Table 2 gives further details of their pulmonary condition at the onset of treatment. In one patient with an acute ground glass exacerbation and one with predominantly ground glass pattern FA a severe deterioration in their clinical condition occurred requiring admission to hospital. An initial response to high dose corticosteroids was seen in all three patients with predominantly ground glass pattern FA or ground glass exacerbations, but this was only maintained in two.

The treatment was tailored to the patient's condition and concomitant drugs; patients were treated with 30–40 mg of prednisolone, which was reduced to a maintenance dose of 10 mg daily, or equivalent, once a satisfactory response was seen. Azathioprine was instituted in combination with prednisolone wherever possible. The pattern of FA on HRCT at the start of treatment did not seem to predict response to conventional treatment, with one of three patients with predominantly ground glass FA and one of three patients with reticular pattern FA dying of their lung disease. The four surviving patients had been stable with treatment for 18 months at the time of submission of this paper. Unfortunately, the number of patients who were treated for their lung disease was small and meaningful statistical analysis cannot be undertaken. Treatment was also given at different stages of the disease process, which further limits our ability to draw firm conclusions from the data available.

Predicting progressive FA

Ten patients had significant deterioration in their FA over two years' follow up. These patients were compared with 15 patients with stable lung disease who continued to the end of the study. Tables 3 and 4 give a comparison of their clinical features. Clinical features associated with the progression of FA were the presence of bibasal crackles (p=0.04), a reduced Tlco (p<0.001), extent of FA on HRCT (p=0.026), and distribution of involvement of FA on HRCT (p=0.031). On forward conditional logistic regression only Tlco remained significant (p=0.024, estimated odds ratio 0.85, 95% confidence interval 0.74 to 0.98). The level of Tlco was investigated with receiver operator curves to identify clinical predictors of progressive FA (fig 1). It was found that a 54% percentage of predicted Tlco cut off gave 80% sensitivity and 93% specificity.

DISCUSSION

This is the largest prospective HRCT based study of FA in patients with RA to date. With this study we have been able to assess the natural history of FA and to determine over the short term what proportion of patients have rapidly progressive FA and how many have a more benign course. The FA was mainly reticular pattern (80%) and coexistent emphysema was present in 41%. We found in this unselected cohort of patients with RA and FA that over a two year period pulmonary function deteriorated significantly in 10/29 (34%) patients, including four (14%) patients who died of respiratory failure. HRCT has shown that progression can occur with increased reticular pattern or ground glass exacerbations. Predominantly ground glass pattern FA involvement can spontaneously improve. When the presenting clinical features were examined, progressive FA was associated with the presence of bibasal crackles, Tlco, and extent and distribution of lung involvement on HRCT at initial presentation. However, Tlco is the most reliable independent predictor of progression. With a Tlco <54% we found in this group, with 93% specificity, that significant progression of their lung disease occurred. Measured Tlco will vary between centres, but it might be reasonably assumed that with a 5% standard deviation⁹ a trial of treatment with corticosteroids should be considered when the Tlco is consistently between 50% and 60% of the predicted value.

It is possible to question whether lung biopsy should have been undertaken on these patients, but now that it has been prospectively shown that HRCT correlates with lung biopsy in 90% of cases in diagnosing FA this would seem to be unnecessary when considering a diagnosis of FA.¹⁰ HRCT has been shown to be equal to, if not better than, lung biopsy in predicting survival in idiopathic pulmonary fibrosis.¹¹ It is the extent of fibrosis present that determines outcome in FA, which is where HRCT gives a particularly accurate, global assessment.^{7, 8} However, the exact nature of the ground glass pattern interstitial lung disease (ILD) cannot be known with certainty with HRCT. This might be due to fine fibrosis or cellular infiltration as found in desquamative interstitial pneumonia, non-specific interstitial pneumonia and, to a lesser extent, in usual interstitial pneumonia. Certainly, those patients with the remitting ground glass pattern ILD are likely to have had non-specific or desquamative interstitial pneumonia cellular infiltrations. The extended classification of lung histology now described in cryptogenic FA⁷ needs to be evaluated in RA associated FA.

The presence of emphysema complicates the study, but there is increasing evidence that smoking is associated with RA.^12–14 Recently, Rajasekaran et al found that active smoking is higher in patients with RA related ILD than those with cryptogenic FA¹⁵; it is therefore not surprising that emphysema and FA occur in RA associated FA. In our patients, HRCT showed apical emphysematous bullae that did not interfere with the diagnosis or grading of the scans for FA. In anticipation of emphysema and smoking producing a progressive fall in Tlco that might be falsely attributed to FA, evidence of HRCT progression of FA was included in the definition of progression. We also included the presence or absence of bullae in the multiple logistic regression analysis, and this was not associated with progression at two years. We suggest that if a more sensitive scale for the HRCT assessment of emphysema were developed it might be possible to improve prediction of progression in patients with RA and FA even further.

Akira et al specifically reported the course of CT diagnosed lung disease in patients with RA.³ They described the clinical events occurring in 29 patients with RA who had lung disease diagnosed by chest HRCT between 1986 and 1997. Nineteen patients had features in keeping with FA, of whom 10 patients died of respiratory failure, and one died of small cell lung carcinoma. Mean survival of those dying was three years (range four months to seven years). This study was not able to correlate CT and lung histology with clinical features or pulmonary function tests and did not attempt to find predictors of the course of HRCT diagnosed FA in patients with RA. Progression of disease on HRCT showed a similar pattern to that found in our patients. They had a higher percentage of patients with rapid deterioration with the appearance of multifocal areas of ground glass attenuation (38%), and unfortunately all these patients died. Thirty one per cent of their patients showed extension of honeycombing from peripheral to central parts of the lung on HRCT. The difference in frequency of deterioration probably reflects the retrospective nature of the study of Akira et al.

Remy-Jardin et al included follow up CT data on four patients.¹⁶ HRCT was repeated owing to clinical deterioration, rather than at regular intervals. One patient had progression in the extent of honeycombing from the peripheral to central portions of both lungs. In three patients CT showed rapid changes in lung abnormalities, with extension of honeycombing centrally and concurrent replacement of ground glass attenuation by honeycombed lung. This occurred within four months in two patients and six months in the other patient. Again all three patients with this rapid deterioration died.

Fujii et al repeated HRCT at one year in 28 patients with RA in whom ILD had been diagnosed at the first examination.¹⁷ Similar results to those of our study were found, with 57% of patients having no change in HRCT findings; however, progression was seen in nine patients. In three patients the HRCT appearance improved, all these patients had ground glass pattern ILD. Interestingly, like our patients they had not been treated with corticosteroids between HRCT assessments and so it seems the cellular component of the disease can spontaneously regress without specific treatment.

For idiopathic pulmonary fibrosis (or cryptogenic FA), predictors of progression have been studied more extensively. It is important to study predictors of progression of lung disease and not to confuse this with response to treatment and survival, as current treatment options are often unsatisfactory.¹⁸ The extent of HRCT lung involvement,¹⁹ which has been shown to correlate with the pulmonary function parameters, forced vital capacity and Tlco,²⁰ has also been shown to be associated with progressive disease in idiopathic pulmonary fibrosis. Currently, several different scoring systems are available for assessing the type, extent, and distribution of FA on HRCT. We did not find the Wells systems of classification to be independently better than Tlco in predicting progression of FA, in our patient group.

Although the initial number of patients with RA associated FA in our study was reasonably large, the number with progression at two years was low and so we would suggest further studies should recruit larger numbers of patients, incorporate each type of pattern of FA, and use all the different HRCT scoring systems. As mentioned previously, a formal score for coexistent emphysema may also improve the accuracy of the HRCT assessment.

Now that we have identified a readily available predictor of progression in FA associated with RA, a standardised approach to monitoring patients and initiation of treatment can be taken in clinical practice and in controlled treatment trials.