Discussion
Pneumocystis pneumonia in AID is unusual in the absence of steroid treatment. Corticosteroids have been recently administered in over 90% of cases in most series [
14,
17,
18,
22] and were the sole immunosuppressant in 17–28% of AID patients [
18,
22]. The median duration of treatment prior to the diagnosis of PCP is three to four months. Occurrence within a month of starting treatment is uncommon, with the exception of inflammatory myopathies [
28,
32]. Most cases have taken prednisolone in excess of 15 mg per day, or equivalent doses of corticosteroid. Notable is the profound inter-subject variation in response to standard steroid doses as measured by in vitro inhibition of lymphocyte proliferation [
33], indicating that host factors are likely to have a significant, but as yet ill-defined role. Several mechanisms have been postulated to explain the role of steroids in promoting the development of P. jiroveci including CD4+ lymphocyte depletion and immune dysfunction [
17,
30]. Porges et al. found an association between the risk of PCP and the dose of prednisone used in SLE [
32]. Similarly, Hellman et al. found an association between prednisolone dosage and risk of fatal opportunistic infection in SLE, the commonest cause of which was P. jiroveci [
26]. However, other studies have failed to show an association between cumulative steroid dose and risk of PCP [
34].
A number of cytotoxic and other immunosuppressive agents commonly used in the treatment of AID are frequently associated with PCP, including cyclophosphamide, azathioprine, methotrexate and ciclosporin [
14]. Cyclophosphamide is routinely used in the treatment of Wegener's granulomatosis and has transformed the previous one year survival figure of 20% [
35] into the present eight year survival of 80% [
36]. Godeau et al. [
34] showed a significant association between cyclophosphamide cumulative dose and the risk of PCP. However, this was not an independent factor in multivariate analysis when lymphopenia was taken into account [
19]. In one series involving 180 patients with Wegener's granulomatosis, no cases of PCP were identified amongst patients on cytotoxic therapy alone (although the authors did not specify the numbers involved), suggesting a permissive role for corticosteroids [
30].
Data on PCP associated with AID indicates lymphocytopenia (<1,000 cells/mm
3) is almost a prerequisite, with 91% of patients exhibiting a low lymphocyte count. Fifty percent of such PCP patients have total lymphocyte counts of <400 cells/mm
3 [
22]. The pre-treatment lymphocyte count and lymphocyte counts during the first three months of immunosuppressive treatment in Wegener's granulomatosis have been shown to be predictive for PCP in multivariate analysis. A total lymphocyte count <600 cells/mm
3 was recorded in ten (83%) of 12 patients with PCP, but such a low lymphocyte count was recorded in 11 (34%) of 32 with Wegener's unaffected by PCP [
34]. A similar association was found in a prospective study involving patients with SLE [
37]. Porges et al [
32] proposed a cut off of total lymphocyte count of <350 cells/mm
3 which captured 4 out of 6 cases with PCP and SLE, but only 1 of 20 patients with SLE unaffected by PCP.
Information on CD4+ counts, which have been shown to be highly predictive of the risk of PCP in HIV infected individuals [
38], is less well documented in AID patients. The issue was addressed by Mansharamani et al [
19] who prospectively observed 171 patients in various risk categories for PCP, including 22 patients with active PCP. They found that patients who were at high risk of PCP had significantly lower CD4+ counts than patients at low risk. They noted that 91% of cases of PCP had CD4+ counts <300 cells/mm
3 at the time of diagnosis. Their findings are echoed by an increased risk of respiratory colonisation by P.jiroveci in HIV negative patients with CD4+ counts of <400 cells/ mm
3 [
39].
Kadoya et al [
37] in their study on the occurrence of PCP in 75 patients with inflammatory myopathy and SLE, noted a significant association between radiological interstitial pulmonary fibrosis (IPF) and the risk of PCP (8.8% IPF in non-PCP vs 100% IPF in PCP, p < 0.001). In contrast, PCP has only rarely been reported in idiopathic pulmonary fibrosis [
15], indicating that more than systemic steroids and pulmonary fibrosis are required to put patients at excessive risk of PCP.
Prevention
Co-trimoxazole is commonly used for PCP prophylaxis in Wegener's granulomatosis when CD4+ counts are <300 cells/mm
3 [
15] or even with normal counts in some centres [
21]. This combination of antibiotics has been shown to be effective prophylaxis when used daily or thrice weekly at a dose of 960 mg in HIV positive patients [
40]. Adverse effects occur in less than 20% of patients, usually manifesting as a rash, which resolves on temporary discontinuation and often does not recur on re-challenge [
41,
42].
Apart from Wegener's granulomatosis, identifying patients with AID who are at risk of PCP has proved a challenge, as the overall incidence is low. Nevertheless it remains an important issue, as AID patients contribute a considerable proportion of cases of PCP in HIV negative patients (up to 36%) and have a particularly poor prognosis, as discussed earlier.
Any method used to select patients for prophylactic treatment needs to be assessed against set standards and have a high sensitivity and specificity. The standards should address the percentage of PCP cases captured by the selection criteria (ideally 100% but in practice >80%) and the risk of the condition in the selected group (which should be significant). In HIV patients who meet the criteria for PCP prophylaxis as set out by the US Public Health Service [
43], the annual risk of PCP is 18% [
38]. As the mortality from PCP in HIV negative patients is approximately double that of HIV-positive patients [
14‐
18,
27], we would suggest that an annual risk >9% of PCP would be sufficient to justify prophylaxis.
In the study by Mansharamani et al, their proposed cut off of <300 CD4+ cells/mm
3 would capture 91% of cases of PCP in all HIV negative patients, but would also include 39–46% of patients on systemic steroids, most of whom would be unaffected by PCP. Administering prophylaxis to such large numbers of patients would unnecessarily expose patients to drug side-effects and potentially encourage drug resistance. However, analysis of their data reveals that the subgroup of patients with AID who developed PCP had CD4+ counts of <250 cells/mm
3 and six out of eight had counts <200 cells/mm
3 [
19].
Given the laboratory costs, we would argue in favour of performing CD4+ counts after one month's immunosuppression only on patients who satisfy the following three screening criteria:
• Steroid dosage >15 mg prednisolone or equivalent/day
• >three months corticosteroid treatment proposed
• total lymphocyte count <600 cells/mm3
A CD4+ count <200 cells/mm3 might then warrant the use of prophylactic co-trimoxazole, if the annual risk of PCP in these patients is greater than 9%. Most cases of PCP in patients with AID would be captured by these criteria, according to published series.
Clearly, further prospective investigation is required to gather sufficient data to validate any selection method. To justify our proposed threshold for prophylaxis we would need to know the risk of PCP for patients on steroid-based immunosuppression for AID with CD4+ counts of <200 cells/mm3, information which is currently unavailable.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AJ Carmichael initiated the discussion, appraised results, lead departmental debates and helped revise the manuscript. ES carried out the literature search, presented at meetings and wrote the original manuscript. All authors read and approved the final manuscript.