PATIENTS AND METHODS

Patients

The study started in July 1998 and finished in December 2000. Patients were studied during one year after having given their written informed consent. Initially, 84 patients with SLE and cutaneous lupus (CL) were included in the study; in the end, however, three of them were excluded because they did not complete the follow up. They all came from the Systemic Autoimmune Disease Unit, Department of Internal Medicine, “Virgen de las Nieves” University Hospital, Granada, Spain. Seventy of the 81 patients who had been finally selected had SLE, and fulfilled four or more of the American College of Rheumatism criteria¹⁸; the other 11 had CL, which was diagnosed by observing clinical and anatomopathological data. The study also included a control group made up of 92 subjects who were healthy or had general complaints unrelated to lupus, urinary pathology, or diabetes. All of them were taken from the researcher’s milieu and had a distribution of sex and age similar to that of the lupus group. Eventually, six of the 92 healthy subjects had to be excluded because they did not complete the study.

Definitions

Sexual activity was considered a UTI risk factor when the patient had sexual intercourse four times a month.¹⁹ The diagnosis of chronic constipation was based upon Rome’s criteria.²⁰ Low urine frequency was considered when the subject urinated four or fewer times a day.²¹ Complicated UTI was considered in cases of repeated UTIs, with signs and symptoms in the upper tract, and/or an anatomical or functional alteration in the urinary tract, immunological deficits, pregnancy, and antibiotic resistant uropathogens.²¹ Uncomplicated UTI was diagnosed in cases of positive urine culture (⩾10² colony forming units/ml urine), urinary symptomatology, and lack of UTI predisposing factors (excluding UTI in sexually active women).²¹ For the diagnosis of asymptomatic bacteriuria, two consecutive >10⁵ cfu/ml urine cultures had to be recorded.²² The total number of UTI episodes throughout the one year follow up was calculated by adding up the number of complicated and uncomplicated UTI episodes, and asymptomatic bacteriuria episodes. Repeated UTI was considered to occur when the subject had had three or more episodes over the course of one year. The diagnosis of detrusor instability (or hyperactive bladder) was confirmed when the detrusor showed involuntary contractions, with variable frequency and intensity higher than 15 cm of H₂O, which were not inhibited during vesical filling and were frequently related to urinary urgency.²³ Hypoactive bladder was diagnosed when there was a large filling volume, delay in urinary voiding, or low detrusor pressure with low urinary flow.²⁴ Female urinary incontinence was identified by taking into account clinical criteria and urodynamic studies.²⁵ The diagnosis of lower urinary tract obstruction (prostatism) in men was determined in accordance with clinical manifestations, rectum examination, isolated flowmetry, and vesicoprostatic echography. The total accumulated drug dosage was calculated by multiplying the daily dose given by the total number of treatment days.

Protocol

During the one year follow up, patients were examined on five occasions: at the beginning of the study, and in the 1st, 3rd, 6th, and 12th months.

Lupus patients

On the first visit the following variables were recorded on a database (Microsoft Access 98): age, sex, duration of the disease, lupus type (CL, SLE), previous cases of UTI, number of admissions to hospital due to UTI, UTI predisposing risk factors (Sjögren’s syndrome included), current treatment for lupus, and total accumulated dosage of danazol, prednisone, antimalarial drugs (chloroquine and hydroxychloroquine), and immunosuppressive agents (azathioprine, methotrexate, cyclophosphamide, cyclosporin A, and mycophenolate mofetil). On the next occasions, patients were asked about their urinary habits, and were required to record their urinary frequency, diuresis, and approximate daily volume of liquid drunk in the week before the visit. They were then asked about their potential urinary incontinence, the severity of which was measured according to the following scale: 0 = no loss of urine; 1 = light loss of urine; 2 = mild loss of urine; and 3 = loss of urine with minimal activity, or gravitational incontinence.²⁵ Lupus disease activity was established in accordance with the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) score²⁶; and organic damage, according to the Systemic Lupus International Collaborating Clinics/American College of Rheumatism (SLICC/ACR).²⁷ A full physical examination of the patients was also carried out; they were asked questions about clinical data in the 20 days before the visit. This questionnaire included the following items: arthralgias, arthritis, myalgias, photosensitivity, butterfly rashes, oral aphthas, asthenia, costal pain, dyspnoea, temperature, anorexia, loss of weight, convulsions, diarrhoea, vomiting, abdominal pain, ocular symptoms, pollakiuria, dysuria, suprapubic pain, tenesmus, changes in the colour and smell of urine, foaming urine, pain in renal fossa, and shivering.

The laboratory assays carried out on the five visits included the following: standard biochemical test; haemogram; erythrocyte sedimentation rate; C reactive protein; urinary sediment; proteinuria in 24 hour urine; nDNA, and complements C3 and C4; urine culture (on cystine lysine electrolyte deficient medium); and antibiogram, through a quantitative method of minimal inhibitory concentration obtained by using WIDER 1 Difco panels (Dade Microsan Inc, West Sacramento, CA, USA) and following the current antibiotic susceptibility criteria for the uropathogens suggested by the National Committee for Clinical Laboratory Standards.²⁸ An immunological study was also carried out on the first visit which included: antinuclear antibodies (ANA), Sm, Ro/SSA, La/SSB, and antiribosomal P0. Furthermore, anticardiolipin antibodies (aCL), IgM and IgG, and lupus anticoagulant (LA) were tested on the first and third visits. Patients with UTIs were given antibiotics in accordance with the antibiogram and, after treatment, another urine culture was carried out. Patients with asymptomatic bacteriuria were given antibiotic prophylactic treatment under the following circumstances: pregnancy, immunosuppression (leucopenia, hypocomplementaemia, and immunosuppressant treatment), bacteriuria by Proteus, and persistent bacteriuria 3–5 days after an intravesical catheter.²⁸ When the urine culture remained positive on three consecutive visits despite the antibiotic treatment, the patient was treated with antibiotic prophylaxis for 6 months, with the urine culture being controlled in the first, third, and sixth months. Postcoital prophylaxis was given when we suspected UTI was due to sexual activity, whereas nocturnal prophylaxis was used in the other cases. Finally, a genitourinary echography was carried out in order to rule out structural alterations (either congenital or acquired), as well as a urodynamic study, to rule out functional alterations during urinary voiding. This urodynamic study included isolated flowmetry as a screening test, followed by cystometry and a pressure flow test in accordance with the following protocol: when the flowmetry showed normal parameters and the patient did not have urogenital manifestations (that is, stress incontinence, prostatism, or micturitional syndrome), the urogenital study was stopped. On the contrary, the study was continued in the three following situations: (a) when flowmetry was not normal regardless of the patient’s clinical manifestations; (b) when, despite flowmetry being normal, the patient had urogenital manifestations; and (c) when, after repeating a doubtful flowmetry, this continued to be doubtful or abnormal.

Control group

The control subjects were asked about previous cases of UTI, UTI risk factors, clinical manifestations, urinary habits, and incontinence in the same way and during the same period of time as the lupus patients. They were also subjected to the same supplementary tests except for the immunological study.

Statistical analysis

A descriptive analysis of variables (that is, median score and standard deviation) was performed for both the lupus and the control groups. Dichotomic variables were expressed in percentages. Mann-Whitney U and χ² tests were used as appropriate. The level of significance was p<0.05. To determine the factors that influence the onset of UTI on the five visits, we used the generalised estimation equation model (model GEE), with a non-structured correlation matrix for categorical data, and the binomial distribution and the “logit” as a natural link. Thus, the “natural” measure for the effect is the “cross-product ratio”. The appropriate goodness of fit test was used, which did not show significant results in any of the cases. Firstly, a comparative study was also carried out between the lupus and the control group; and, later on, within the lupus group itself. A three stage procedure was followed, which resulted in the adjustment of a multivariate model: (a) a bivariate analysis (model 0) was performed for each item in order to measure their individual effect on the response variable; (b) we took those significant variables from the previous analysis or those that were to be controlled but were not significant, and we produced a joint multivariate model (model I) in order to measure the net effects of such variables; (c) by eliminating the variables of model I that did not provide any information on the response variable and did not cause significant changes in the model coefficients, we produced a definitive model (model II). This last model was the one presented for the final study. This procedure aimed at avoiding having an excessive number of variables (parsimonious model), without eliminating any that might provide information on the response variable, directly or indirectly. The following statistical programs were used: SPSS 10.0, in the first part of the study; and STATA 6.0, in the different comparisons.

RESULTS

Descriptive analysis

Demographic data, UTI prevalence, type and risk factors in the lupus and control groups

Seventy three of 81 (90%) lupus patients were women and 8 (10%) men. The average age was 41.4 (SD 11.9) years (range 18–69). The average disease duration was 7.9 (SD 5.8) years (range 0–28). The sex and average age were similar in both the lupus and the control groups (table 1). The UTI prevalence was significantly higher in the lupus group: 36% (29/81) in comparison with 9% (8/86). The total number of UTI episodes during the one year follow up was also much higher in lupus patients (56 in comparison with 9), as was the number of patients with repeated UTI. Lupus patients presented a similar percentage of uncomplicated UTIs and asymptomatic bacteriuria, whereas the UTI type that was most commonly observed in the control group was asymptomatic bacteriuria (67%). Eleven cases of complicated UTI were recorded among lupus patients (which resulted in two admissions to hospital) and none among control subjects. Table 2 shows that no significant differences were seen between lupus and control groups in relation to UTI predisposing factors, except for urinary incontinence and low urinary frequency. These two factors were more common in lupus patients. It should be pointed out that despite the fact that the overall prevalence of urinary incontinence (including all levels of severity) was 24%, moderate-severe incontinence was only present in seven (9%) lupus patients. Only one of our 70 patients with SLE (1.4%) had secondary Sjögren’s syndrome (table 2). This patient had one episode of cystitis, which is 3.4% of the total number of UTIs in our lupus cohort.

Micro-organisms involved in UTIs, treatment and resistance pattern

E coli was the uropathogen most frequently isolated in both groups. Table 1 shows a detailed list of the micro-organisms isolated in the urine culture. Only three controls required outpatient treatment with antibiotics, and they recovered subsequently. In the lupus group, 47 patients recovered fully from UTI, 7 relapsed, and 1 was re-infected. Two lupus patients were admitted to hospital to obtain parenteral antibiotic treatment for urinary sepsis. Nightly antibiotic prophylaxis was only given in 13 of the 405 visits (3.2%) of the 81 lupus patients. Postcoital prophylaxis was given to only one lupus patient. The E coli resistance pattern of lupus patients was as follows: E coli was susceptible to amoxicillin-ampicillin (8/26, 31%), amoxicillin-clavulanate (21/26, 81%), cefalothin/cefazolin (20/26, 77%), cefotaxime/ceftriaxone (25/26, 96%), fosfomycin (26/26, 100%), gentamicin (25/26, 96%), pipemidic acid (15/26, 58%), norfloxacin (18/26, 69%), floxacin/ciprofloxacin (18/26, 69%), trimethroprim-sulfamethoxazole (13/26, 50%), and nitrofurantoin (26/26, 100%). We have not included the E coli resistance pattern in the controls because it was similar to that in lupus patients.

Lupus activity and organ damage

For lupus activity and organ damage over the five assessments, the 70 patients with SLE presented a mean SLEDAI score of 2.3 (SD 2.6) (range 0–12) and a mean SLICC score of 0.7 (SD1.1) (range 0–5).

Laboratory test results

Table 3 shows the laboratory test results of lupus patients with and without UTI.

Treatment received

Table 4 shows both the accumulated drug dosage, and the weekly drug dosage given to lupus patients with and without UTI throughout the study.

Final results after applying generalised estimating equation (GEE) models

The joint analysis of subjects participating in the study through an MI bivariate analysis (taking the “odds ratio (OR) cross-product ratio” as measurement) enabled us to observe a significantly different UTI distribution in both groups (lupus and control) on all five visits. Thus, controls had fewer episodes of UTI than lupus patients on all five visits (OR = 0.139; 95% confidence interval (CI) 0.06 to 0.33; p<0.001). This fact was confirmed by the use of the MII multivariate analysis (OR = 0.22; 95% CI 0.11 to 0.45; p = 0.001). The study of the lupus patients’ epidemiological variables related to the onset of UTI through this analysis showed that the only variables associated with the possibility of having UTI were age (p = 0.002), previous cases of UTI (p = 0.0001), and previous admissions to hospital due to UTI (p = 0.002). These results were confirmed by the MII multivariate analysis (table 5). The rest of the variables (sexual activity p = 0.81; use of barrier methods p = 0.30; menopause p = 0.39; constipation p = 0.14; genitourinary alterations p = 0.94; Sjögreńs syndrome p = 0.36), which were analysed jointly through the M0 analysis, did not seem to be significantly related to the onset of UTI. For the autoantibodies analysed (ANA, nDNA, SSA, SSB, RPO, Sm, aCL, and LA), only ANA levels >1/80 IU/ml were significantly related to the onset of UTI (p = 0.03), which was later confirmed by the MII multivariate analysis (table 5). Likewise, when using the MI bivariate model to analyse the probable relation between the onset of UTI, leucopenia <3×10⁹/l, thrombocytopenia <100×10⁹/l, and hypocomplementaemia (C3 <79 IU/ml and C4<16 IU/ml), only leucopenia (p = 0.08) and C3 levels <79 IU/ml seemed to be associated with UTI (p = 0.08). When the MII bivariate model was applied, it was only thrombopenia that showed a significant relation (p = 0.02); we also found that leucopenia and UTI tended to be associated (p = 0.093) (table 5).

For the association between the onset of UTI and drug administration, the MI bivariate model only showed a significant relation with the weekly administration of methotrexate (p = 0.03), which was not totally confirmed as MII only showed a trend (table 5). Finally, no significant relation was found between the onset of UTI, and disease activity (according to the SLEDAI) (p = 0.32) and organ damage (according to the SLICC/ACR) (p = 0.36).

DISCUSSION

In previous studies,^17,18 as well as in the present research, lupus itself has been considered as an infection predisposing factor, regardless of the rest of the variables. The average age of our lupus patients was 41.4 (SD 11.9) years and was significantly associated with the onset of UTI on all five visits. This age was within the range in which UTIs and lupus have a greater occurrence.^29,30 Previous episodes of multiple urinary infection constitute another UTI predisposing factor,²⁹ a variable that was also considered a risk factor in our study.

On the other hand, we found that sexual activity and barrier methods were not related to the onset of UTI in lupus patients. From our perspective, this may be due to the fact that sexual relations on their own cause a minimal modification of the vaginal flora,³¹ an essential mechanism in the pathogenesis of UTI which is directly related to the use,³² and the frequency of use,³³ of condoms. We must acknowledge that we did not measure this variable in this manner. Thus, we did not determine the number of times that patients had made use of it before a UTI episode.

As for constipation, this did not seem to constitute a UTI risk factor in the subjects of our study or in adults within the same age range.³⁴

Anatomical and functional abnormalities in the urinary tract are known to be risk factors for the onset of complicated UTIs.³⁵ In our cohort of lupus patients, a very reduced number of cases presented these abnormalities, which is why it was necessary to group them in order to adapt to the model. Even so, they did not appear to influence the onset of UTI. Moreover, stress incontinence did not seem to have any influence on the onset of UTI, as only two out of the 19 women had severe incontinence, and incontinence is related to the onset of UTI when it is severe.³⁶ We found no association between UTI and low urinary frequency, a result that differs from those described by other authors, according to whom a reduction in the number of UTIs observed in premenopausal women was due to the adequate consumption of water and an appropriate number of daily urinary voidings.³⁷

We did not find any statistically significant associations between the disease length and the onset of UTI. This result coincides with the results of other authors,^6–11 who found that lupus infections can appear at any time of the disease development without a predominance during the first years, as has been sometimes suggested.^38,39

To date, no research has studied the incidence of UTI in patients with lupus and Sjögren’s syndrome. None the less, there exists research devoted to patients with rheumatoid arthritis (RA) which concludes that whenever RA and Sjögren’s syndrome are associated, the incidence of UTI is more frequent in those patients.⁴⁰ In our research, we could not find this specific association, which may be because only a small number of lupus patients also had Sjögren’s syndrome at the same time.

As for the clinical type of lupus, most studies of infections included patients with SLE, with the exclusion of patients with CL.^{5,6,9–11,41–46} Our study, which included 70 patients with SLE (86%) and 11 with CL (14%), showed that the onset of UTI during the course of the study was fully independent of the clinical type of lupus.

In the analysis of autoantibodies and complement levels, only the ANA were significantly related to UTI, but not the nDNA. This finding was comparable to that obtained by Ter Borg et al in 1991⁴⁷ and by other authors,¹⁷ who did not find any relation between nDNA levels and the onset of disease. As already observed by other authors,⁴⁸ the association between UTIs and ANA may be due to the fact that the disease is still active in these patients and that they do not have chronic renal insufficiency, both of which are causes of the disappearance of ANA.

As for the number of leucocytes, we found some relation to the onset of UTI when they were lower than 3×10⁹/l, which had already been described.³² This finding can be explained through the capacity of leucocytes to limit the uropathogen invasion to tissues, clearing them from urine and intervening in their dispersion control.³² Besides, the probability of patients with lupus and thrombocytopenia <100×10⁹/l (26/405; 6%) having UTI on all five visits was 4.33, which seems to be merely casual.

As for drugs administration, no association was found between the onset of UTI and accumulated and daily administration of corticosteroids in our lupus patients. This result is due to the fact that most our patients needed a 4.75 mg daily dosage of prednisone, which is lower than the risk dosage for infection recorded by most authors.^17–46 For the cytostatic drugs, we found a relation between the weekly dosage of methotrexate and the onset of UTI. In 1994, in a prospective study of patients with RA, Van der Veen et al analysed the probable influence of methotrexate on the onset of infection, with more infections being found in the group that had been given this drug.⁴⁹ Recently, septic arthritis due to L monocytogenes was recorded in a patient with SLE who was receiving low dose methotrexate treatment.⁵⁰

We found no relationship between UTI and lupus activity. This may be explained by the fact that most patients had a SLEDAI score of under 4, which is lower than that recorded in some studies as a risk indicator for SLE infection.¹³ It should also be pointed out that other authors, besides us, have found no association between disease activity and the onset of infection in patients with SLE.^6–46

The uropathogen most frequently isolated was E coli, (28/47; 60%), a result that is similar to that described previously.^9,10,12–14 It should be borne in mind, however, that UTIs in patients with SLE do not always have a good prognosis and are likely to develop into a severe episode requiring admission to a hospital intensive care unit.⁵¹ As for the study of E coli antibiotic sensitivity, the trimethoprim-sulfamethoxazole resistance percentage was around 50%, which was also related to resistance to ampicillin (69%) and to first generation cefalosporins (23%). In Spain there are very few data on E coli antibiotic resistance patterns, one source of which is the multicentre study including Canada and 16 European countries; in this research, a TMP-SMX resistance percentage of around 35% was detected in Barcelona, Spain.⁵² On the other hand, these authors found that the E coli resistance phenotype to TMP-SMX was usually associated with other antibiotic resistance, mainly to ampicillin and cefalothin,⁵² as is the case in our study. As for quinolones, the resistance percentage detected was around 30%, a result comparable to that obtained in previous studies.⁵³ The E coli resistance to nitrofurantoin and to fosfomycin detected in Europe was very low,⁵² below or equal to 2%. This result is similar to the present study. Therefore, on the basis of the recommendations by the Infectious Diseases Society of America (IDSA),⁵⁴ nitrofurantoin or fosfomycin would be the empirical treatment preferred for acute cystitis in our lupus patients. With regard to nitrofurantoin, it would be reasonable to be cautious when using it for the treatment of cystitis as it is a lupus-inducing drug, despite the fact that it has not been related to the idiopathic SLE activity.⁵⁵ In our study, the E coli resistance percentage to third generation cefalosporin and to gentamicin was 3.8%. Thus, taking into account the IDSA recommendations,⁵⁴ these two antibiotic families would be the first choice for our patients with lupus and pyelonephritis.

In short, lupus is a disease which, itself, predisposes patients to urinary infections. Patients with UTI who have had previous UTI episodes more frequently have leucopenia, thrombopenia, and ANA >1/80 IU/ml (regardless of nDNA levels, organ damage, and disease activity and development). UTIs are usually lower tract infections, acquired through the community, and appear repeatedly. In most cases, outpatient treatment is given and E coli is the most common uropathogen. On the other hand, the E coli antibiotic spectrum was similar to that described in previous studies with healthy subjects. Therefore, we suggest an empirical treatment of cystitis for our lupus patients, based on nitrofurantoin, or fosfomycin, as first choice; and, in the case of pyelonephritis, third generation cefalosporin and/or gentamicin.