Histopathologic and Microbiologic Aspects of Ventilator-associated Pneumonia

Over a period of 1 yr, 25 patients whose lungs had been mechanically ventilated for more than 72 h and who died in our respiratory intensive care unit were included in the study. Patients with immunosuppression or hematologic neoplasia were excluded because of their inherent particularities regarding etiology, immune response, and presentation of pulmonary infection.

In 20 of the 25 patients, pulmonary infection on the day of death was clinically suspected. Twelve of the 25 patients were admitted to the intensive care unit because of acute respiratory failure due to pneumonia (n = 6), congestive heart failure (n = 2), alveolar hemorrhage (n = 2), or pulmonary neoplasia (n = 2). Twenty-one of the 25 patients (84%) had either diffuse (7 cases) or localized (14 cases) chest radiograph infiltrates 24 h before death. In four patients, there were no infiltrates. Chest x-ray results were reviewed on the day of the study. The clinical diagnoses of the pulmonary infiltrates were pneumonia in 14 patients, pneumonia and acute respiratory distress syndrome in 5, congestive heart failure in 1, and alveolar hemorrhage and pneumonia in 1. The mean interval between appearance of infiltrates in chest x-ray and death was 6+/-6.1 days for patients whose lungs were ventilated < 10 days and 16.6+/-14.2 days for patients whose lungs were ventilated+/-10 days (Mann-Whitney test, P = 0.04). General characteristics of the study population and causes of respiratory intensive care unit admission and death are described in Table 1. Seventeen patients received prior antibiotic therapy (group A), whereas the remaining 8 subjects (group B) did not receive antibiotics within the 48 h before death. The indications (n = 17) for antibiotic therapy in group A patients were pneumonia in 8, aspiration pneumonia in 2, multiple trauma in 2, postoperative prophylaxis in 1, peritonitis in 1, sepsis in 1, purulent tracheobronchitis in 1, and organ donor in 1. The mean duration of antibiotic treatment was 9.5+/-7.9 days. Family members in each case gave informed written consent for the study to be performed, and permission from the Ethical Committee of our center was granted.

Immediately (within 45+/-45 min, range 0-90 min) after death, patients' lungs remained mechanically ventilated with FI^O21.0. Bilateral fiberoptic bronchoscopies were performed using different bronchoscopes for each lung following standard techniques. [10,11]One guided biopsy from each lung, through thoracotomy incisions, was obtained with the aid of the light of the bronchoscope. The bronchoscope was guided to the area of visual maximal inflammation corresponding to chest x-ray infiltrates. In the absence of infiltrates, the bronchoscope was guided to the lower lobes. Furthermore, bedside blinded (not guided by fiberoptic bronchoscope) bilateral pulmonary biopsies, following strict aseptic techniques, were obtained through the same thoracic incisions. The thoracotomy incision was done at the 5th intercostal space from the midclavicular to the midaxillary lines. In organ donors, this procedure was performed in the operating theater; in the remaining patients, the procedure was done in our intensive care unit. Lungs were not removed from the thorax, and sampling of posterior segments was possible because the lungs were partially collapsed as mechanical ventilation was maintained throughout the study. Three fragments were obtained from both superior and inferior lobes from each lung and 2 from the middle lobe, such that the overall number of biopsies per patient was 16 (2 guided and 14 blinded). The total number of biopsies was 375 (in three patients, one lung only was sampled). In 14 patients (217 samples), samples were obtained within 30 min after death; in 5, within 60 min (62 samples); and in 6, within 90 min (96 samples). Of the 14 patients in whom samples were obtained within 30 min after death, 8 were organ donors, and the period between brain death and obtaining the sample was less than 12-24 h. One organ donor had the sample obtained between 30 and 60 min. The average size of the bronchoscopically guided biopsies and that of the blind biopsy samples was 2 x 2 x 2 cm each. Biopsy samples were obtained from peripheral zones of the lung. Each fragment was sectioned into two pieces, one for quantitative microbiology culture and the other for histopathology processing.

All lung tissue specimens were embedded in paraffin and stained with hematoxylin and eosin, periodic acid Schiff, May-Grunwald Giemsa, and Gram stains. The microscopic evaluation was performed after a classification of pneumonia evolution based on the most widespread guidelines for this type of lesion. [14]In the current study, pneumonia was divided into four phases related to the following sequence of events in lung tissue: early, intermediate, advanced, and resolution. Early-phase pneumonia was defined as the presence of capillary congestion with increased number of polymorphonuclear leukocytes at this level; the alveolar spaces usually showed a fibrinous exudate. Intermediate-phase pneumonia was considered when alveoli had presence of fibrin, few erythrocytes, and several polymorphonuclear leukocytes. Advanced-phase was assessed when polymorphonuclear leukocytes filled most of the alveoli, and macrophages were incorporating cellular debris in the cytoplasm. In the resolution-phase, the inflammatory exudate was eliminated because of macrophagic activity of mononuclear cells. In Figure 1, the different evolutionary phases are illustrated. Bronchiolitis was defined as an intense proliferation of polymorphonuclear leukocytes localized within the lumen of bronchioles and associated with purulent mucus plugs and bronchiolar wall alterations [11](Figure 2). Focal, confluent bronchopneumonia and lung abscesses were examined in lung biopsy samples, according to the definition of Rouby et al. [11].

Other pulmonary findings, such as diffuse alveolar damage, [15]were assessed. Diffuse alveolar damage was classified into two phases: The early exudative phase was defined as the presence of interstitial and alveolar edema, with varying degrees of intraalveolar hemorrhage and fibrin deposition and initial development of hyaline membranes. The late stage of diffuse alveolar damage was defined by the presence of fibroblast proliferation within the interstitium, interstitial inflammation and alveolar lining cell hyperplasia remaining prominent, but residual edema or hyaline membrane formation being minimal. [15]For clarity, results regarding diffuse alveolar damage were presented as a dichotomic variable (presence or absence).

Biopsy samples were placed, with sterile sand and 3 ml of sterile saline, in a mortar (Vidrafoc, Barcelona, Spain) and homogenized. Serial dilutions (10 sup -1, 10 sup -2, 10 sup -3) of each sample were prepared in sterile normal saline. One hundred microliters of each dilution of biopsy samples was inoculated into the following agar media: 5% sheep blood, chocolate, Centers for Disease Control and Prevention blood, McConkey, blood charcoal yeast extract, and Sabouraud-dextrose. All cultures were incubated at 37 degrees C under aerobic and anaerobic conditions and in carbon dioxide-enriched atmosphere. Cultures were evaluated for growth 24 and 48 h later and discarded, if negative, 5 days after, except for Centers for Disease Control and Prevention blood and Wilkins-Chalgren, evaluated at 7 days, and Sabouraud, evaluated at 4 weeks. All microorganisms isolated were identified by standard laboratory methods. [16]Results are expressed as colony-forming units per gram of tissue (cfu/g = number of colonies x dilution factor x inoculation factor). Results of biopsy cultures were analyzed first considering individual pulmonary biopsy samples and second considering lung lobes. In the latter, a lobe was considered as compatible with pneumonia if at least one biopsy sample obtained from the same lobe was positive. For the analysis of individual lobes, when bacterial isolates coincided in more than one sample from the same lobe, the species with the highest bacterial count was chosen. Positive lung cultures were considered when one or several microorganisms were isolated in the culture of a lung sample. A false-negative result was considered if no growth was obtained from a lung biopsy culture and the histologic examination demonstrated pneumonia. A false-positive result was defined as a positive bacterial growth in absence of histologic pneumonia. Results also were expressed quantitatively to evaluate the bacterial concentration of pulmonary samples. [9,11-13]The threshold of 10³cfu/g of tissue was used, as previously recommended. [11]The reference test used to evaluate the sensitivity and specificity of pulmonary biopsy cultures was the histology of these biopsies. Bacteriologic results were analyzed by dividing the microorganisms into those with high pathogenicity and those without or with a lesser degree of pathogenicity (Staphylococci other than Staphylococcus aureus, Streptococci other than Streptococcus pneumoniae, and yeasts).

EPIinfo (version 5.01) [17]software was used for the statistical analysis. Results are expressed as mean+/-SD. Analysis of variance for repeated measures was performed for multiple comparisons. When significant F statistic was observed, the Scheffe test was applied. The t-test was employed for the comparison of quantitative variables (Mann-Whitney test if the distribution was not normal). The chi-square test (Fisher's exact test when needed) was used to compare proportions. The significance levels were set to 0.05. Results of microbial cultures of biopsy samples were categorized into true-positives, false-positives, true-negatives, and false-negatives. These were used to determine the sensitivity and the specificity according to Bayes' rules.

Pneumonia. The incidence of histologic pneumonia in the study population was 92% (23 of 25 patients). Two subjects had no histologic evidence of pneumonia, 3 had unilateral pneumonia, 17 others (68%) had bilateral pneumonia, and the remaining 3, in whom the study was restricted to one lung, had pneumonia also. A total of 47 lungs (24 left and 23 right) from the 25 patients were examined. Overall, 168 biopsies were obtained from the 24 left lungs, 79 of which showed histopathologic evidence of pneumonia (47%). A similar percentage of histopathologic features of pneumonia (46%, 95 of 207 biopsies) was obtained from the 23 right lungs. Thus, the cumulative percentage of histologic pneumonia was 46% (174 of 375). There were no differences in the percentage of histologic pneumonia when comparing guided (53%) versus blinded biopsies (45%). Histologically proven pneumonia involved predominantly lower lobes (57%) compared to upper (30%) or middle lobes ((13%) P < 0.0001 (chi²test)).

In 14 (30%) of 47 histologic lung examinations, there was a single histologic pattern of evolution, predominantly the intermediate-phase (64%). Early phases of pneumonia alone were observed in two patients (five biopsy samples), and resolution phase alone was not observed. Different histologic evolution phases coexisted in 26 (55%) lungs. In seven lungs (15%), histologic pneumonia was not shown. The incidence of bronchiolitis was 52% (13 of 25 patients), corresponding to 33 biopsy samples (9%). Regarding positive biopsy cultures without histologic presence of pneumonia, bronchiolitis was observed in 11 of 136 samples (8%). Large bronchi (with bronchial smooth muscle and cartilage) were never observed in our biopsies. Focal bronchopneumonia was observed in 128 (34%) samples, corresponding to 15 patients. Confluent bronchopneumonia was observed in 46 (12%) samples, corresponding to 2 patients. Six patients had focal bronchopneumonia in one lung and confluent bronchopneumonia in the other. Two patients had no pneumonia.

Among the four patients with normal chest x-ray results, three had histologic pneumonia. Among the 21 patients with chest x-ray infiltrates, the incidence of histologic pneumonia was 95% (20 of 21). The percentage of samples showing histologic pneumonia was similar in the 14 patients with infiltrates and in the 7 patients with diffuse infiltrates (50% and 62%, respectively).

The percentage of lung biopsies consistent with the diagnosis of pneumonia was similar in the 17 patients of group A receiving antibiotics and in the 8 patients of the group not receiving antibiotics: 49% (133 among 272 lung biopsy samples) versus 40% (41 among 103 lung biopsy samples). However, as shown in Table 2, the early-phase process was more frequent in group A than group B. In patients receiving antibiotics, pneumonia was more frequent among those with documented microbial antibiotic resistance: In the 10 patients in whom microbial antibiotic resistance was demonstrated, 56% of pulmonary biopsy samples (81 of 144) demonstrated histologic evidence of pneumonia, whereas in the 7 patients in whom no microbial antibiotic resistance could be evidenced, pneumonia was present only in 41% (52 of 128) of the samples examined (P = 0.01, chi squared test).

Diffuse Alveolar Damage. Eight patients (32%) had at least one pulmonary biopsy sample showing histologic signs consistent with diffuse alveolar damage. Six of these patients demonstrated bilateral and two unilateral chest x-ray infiltrates. The total number of biopsy samples showing diffuse alveolar damage was 44 of 375 (12%). Diffuse alveolar damage coexisted always with the presence of histologic pneumonia in each individual. Among the 44 samples with diffuse alveolar damage, 32 (73%) showed histologic signs of pneumonia in the same specimen.

Relationship between Mechanical Ventilation and Histopathology. When analyzing patients whose lungs were mechanically ventilated for < 10 days (n = 14), 12 (86%) showed histologic signs of pneumonia (79 of 199 (40%) biopsy samples), whereas all patients' lungs (n = 11) with > 10 days of mechanical ventilation showed pneumonia (95 of 176 samples (54%); P = 0.005). Four patients whose lungs were mechanically ventilated < 10 days and another four whose lungs were mechanically ventilated > 10 days showed diffuse alveolar damage in 20 of 199 (10%) and 24 of 176 (14%) biopsy samples, respectively (P = 0.28).

Histologic Findings and Respiratory Failure In the 12 patients who were initially admitted because of acute respiratory failure, 102 of 167 biopsies (61%) showed histologic evidence of pneumonia and 22 of 167 (13%) had diffuse alveolar damage (5 patients). In the remaining 13 patients who were not admitted because of acute respiratory failure, these percentages were 72 of 208 (35%) for pneumonia and 22 of 208 (10%; 3 patients) for diffuse alveolar damage (P = 0.0001 and 0.32, for pneumonia and diffuse alveolar damage, respectively).

Qualitative Aspects. All the serologies yielded negative results. Two hundred eighty of 375 (77%) biopsies yielded positive cultures, for a total of 470 microorganisms. Among them, 149 (40%) cultures were polymicrobial, obtained from 20 patients (80%). Forty percent (190 of 470) of the isolates were Gram-negative bacilli, of which 72% were Pseudomonas species. Thirty-eight percent of the isolates were Gram-positive cocci (38% Staphylococcus aureus), and the remaining 22% included yeasts, fungi (45% were Candida albicans), and other nonpathogenic microorganisms. In Table 3, the different specific microorganisms are listed in relation to the presence or absence of histologic pneumonia. Samples without pneumonia from group B (without antibiotics) grew a large number of Staphylococcus aureus compared to the remaining samples (31% vs. 6%, P = 0.0001). On the other hand, Staphylococcus epidermidis grew more frequently in samples from patients receiving antibiotics (16% vs. 3%, P = 0.001).

Microbial cultures of bronchoscopically guided biopsies were coincident in both lungs in 12 patients and did not coincide in 10. In three patients, this comparison was not possible, because they were pneumonectomized. When comparing cultures from both lungs in the same patient, cultures coincided in 34 microbial species but did not in 19 microbial species. The noncoincident microorganisms were Candida species in five, Aspergillus species in three, Pseudomonas species in three, Staphylococcus epidermidis in three, Corynebacterium species in two, Acinetobacter species in one, Serratia marcescens in one, and methicillin-resistant S. aureus in one case. Twelve microorganisms retrieved from pulmonary samples from group A (with antibiotics) were resistant to the given antibiotics. The most frequent microorganisms resistant to antibiotics were methicillin-resistant S. aureus, Pseudomonas species, Xantomonas maltophilia, Enterococcus faecalis, and Candida Krusei and Aspergillus fumigatus.

The sensitivity of positive biopsy samples, considering all isolated microorganisms, was 80% in patients receiving antibiotics and 100% in whom antibiotics were withdrawn (false-negative rate of 20% and 0%, respectively). Specificity was 30% and 4%, respectively, in both groups. Thus, the rate of false-positive results was 70% and 96%, respectively. When nonpathogenic microorganisms were excluded from the analysis, the sensitivity was 93% and 72%, respectively, and specificity was 15% and 40%. Regarding the delay between death and the extraction of pulmonary biopsies, we found 20% of false-positive results in patients studied within 30 min, 21% in patients studied within 30 to 60 min, and 6% in those studied between 60 and 90 min.

Quantitative Aspects. In Figure 3, the relationships between microbiology and histopathology in patients with and without antibiotic treatment are illustrated and analyzed according to the recommended threshold of 10³cfu/g. As shown in Figure 3(A), biopsy samples obtained from patients of group A receiving antibiotics were characterized by a reduction of the number of species with counts greater or equal to 10 sup 3 cfu/g (P < 0.0001 for biopsy samples demonstrating pneumonia and P = 0.001 for biopsy samples without pneumonia) when compared to group B patients without antibiotics. Interestingly, in patients not receiving antibiotics, 49% of species isolated from biopsy samples without histologic pneumonia were found in greater or equal to 10³cfu/g. As shown in Figure 3(B), when considering whole pulmonary lobes, the results were similar to those obtained when analyzing individual samples.

In Figure 3(C and D), the number of bacterial species (< or greater or equal to 10³cfu/g), excluding nonpathogenic microorganisms, is expressed in relation to the presence or absence of histologic pneumonia in patients from groups A and B, taking into account individual pulmonary biopsies (Figure 3(C)) or pulmonary lobes (Figure 3(D)). The percentage of species with counts greater or equal to 10³cfu/g in samples without histologic pneumonia in patients without antibiotics (group B) was 62% and 67% when considering individual biopsies and lung lobes, respectively. These percentages were high when compared to those obtained when considering all microorganisms (49 and 44%, respectively).

(Figure 4) shows the mean concentrations of microorganisms obtained from lung biopsies demonstrating pneumonia according to the different evolution phases. Biopsy cultures from patients without antibiotic treatment yielded higher microbial concentrations in all evolution phases (P < 0.001). However, when considering groups A and B separately, there were no differences among evolution phases.

(Figure 5) shows the mean concentrations of microorganisms obtained in lung biopsies according to the histologic grade. There were no differences among the mean bacterial concentrations obtained from lung biopsies without pneumonia, focal bronchopneumonia, or confluent bronchopneumonia, independent of prior antibiotic treatment. The only significant difference was observed comparing samples with bronchiolitis with all other three categories from patients with antibiotics (P < 0.05).

The main findings of this study are: (1) Histologic pneumonia is frequently observed in critically ill patients who die and is characterized by a multifocal process showing different evolution stages at the same time. (2) Even in patients who do not receive antibiotics, it is difficult to show a relationship between the bacterial burden and the evolution stage or the histologic grade of the lung infection. (3) The threshold of 10³cfu/g cannot be considered as an appropriate reference test for ventilator-associated pneumonia because of the lack of correlation between histology and microbiology. (4) Treatment with antimicrobial agents results in a significant decrease in lung bacterial burden. (5) Isolated bronchiolitis is a finding that partly explains some positive cultures of lung biopsies without histologic evidence of pneumonia.

In this study, 23 of 25 (92%) patients showed histologic findings compatible with pneumonia. This high incidence of histologic pneumonia may be due to the particular subset of the ultimately fatal study population. The percentage of samples with histologic pneumonia increased significantly in relation to the duration of mechanical ventilation (more than 10 days). This observation agrees with clinical studies that demonstrated that the risk of ventilator-associated pneumonia increases with the length of mechanical ventilation. [18]Moreover, histologic pneumonia was more common in patients who were admitted with acute respiratory failure. Adult respiratory distress syndrome [2]and severe respiratory failure [18]have been shown to be predisposing factors for ventilator-associated pneumonia.

In the current study, pneumonia showed several important histologic features. First, ventilator-associated pneumonia was found to be a nonhomogeneous multifocal process frequently involving both lungs, specially lower lobes. This histologic pattern may be due to the flow and volume patterns generated by mechanical ventilation that probably favors the distribution of bacteria from central to distal airways. Rouby et al., [11]who sampled only one lung, described a similar observation. Rouby et al.'s and our results support the concept that a single lung biopsy cannot exclude pneumonia. Second, ventilator-associated pneumonia is a heterogeneous process characterized by the different phases of evolution at the same time. We are unable to explain this finding, but we believe that studying these aspects may be worthwhile in future postmortem research. As far as we know, our group is the first to report on the different evolution phases of pneumonia. Previous human [2,12,19,20]and animal studies [21]only evaluated the degree of pneumonia extension. Both the widespread distribution and the heterogeneity of ventilator-associated pneumonia may explain why blind endobronchial sampling has similar accuracy compared with visually guided samples. [22,23]It is important to note that, when using a histologic classification of extension and/or severity, only two categories (lack of lung infection and focal bronchopneumonia) were well represented in our study. Confluent bronchopneumonia and bronchiolitis were less frequent and lung abscesses were never observed. In other words, we have only found in our biopsies mild degrees of ventilator-associated pneumonia, and this must be considered during the interpretation of the relationship between histology and microbiology.

Histologic signs of diffuse alveolar damage and pneumonia may be confusing, particularly in the initial and resolution phases. Early stages of diffuse alveolar damage show severe abnormalities of capillary walls and pneumocytes, with thickening of the alveolar septa and accumulation of cell debris in the alveoli, which precedes the hyaline membrane formation. This morphologic picture is quite different from the initial phases of pneumonia in which there is no destruction of any lung structure but capillary dilatation, fibrinous exudate, and accumulation of polymorphonuclear leukocytes. Resolution stage of diffuse alveolar damage is a regenerative process in the interstitium with fibrosis, whereas in pneumonia, there is an intraalveolar lesion that returns to the normal anatomic morphology after macrophage activity. [14,15]The hypothetical overlap between initial and resolution phases of pneumonia and diffuse alveolar damage could induce an overestimation in the presence of histologic pneumonia. This does not apply to our results, because early phases of pneumonia alone (not coexisting with other phases) were found only in 2 of the 25 patients, and resolution phases were not found alone in any of the patients studied.

Diffuse alveolar damage was found in 32% of the study patients. This figure is low compared with other work [11]and may be due to the differences in population studied. All patients having diffuse alveolar damage also showed foci of pneumonia, and in 73% of the samples, diffuse alveolar damage was combined with histologic pneumonia. Our results confirm data reported by Rouby et al. [11]These observations support the concept that ventilator-associated pneumonia is often a complication of diffuse alveolar damage and vice versa. Damaged lungs are at high risk of bacterial infection, as shown in experimental and clinical circumstances, probably due to the impairment of intrapulmonary antibacterial defenses. Intratracheal inoculation of Pseudomonas aeruginosa was followed by superinfection of the lower airways in an animal model of paraquat lung injury. [24].

The specificity of lung cultures in this study was invariably low, regardless of the prior administration of antibiotics. Furthermore, when using the recommended threshold of 10³cfu/g, this resulted in poor value for differentiating colonization from true alveolar infection. In a previous study [9]from our group using a single biopsy per lung and in another by Rouby et al., [11]this finding was also confirmed. Our results have important clinical implications, because quantitative bacterial thresholds are used by most Intensivists for the diagnosis of ventilator-associated pneumonia. [1].

In this study, there was a poor relationship between histology and microbiology of lung cultures, even in absence of prior antibiotic administration. Moreover, when analyzing microbiologic data, excluding microorganisms with low virulence, results did not vary markedly. These findings are consistent with the view that postmortem biopsy cultures do not represent a valid reference for the evaluation of quantitative cultures of bronchoscopic and nonbronchoscopic techniques. This probably happens because, in the complex setting of the critically ill patient, many factors may interfere and influence bacterial concentration of the lung. We believe, based on the available data, that the only valid reference test for the diagnosis of ventilator-associated pneumonia is the histologic confirmation of the disease. [13].

The rate of false-positive biopsy cultures was high in the current study. In postmortem studies false-positive lung cultures may be due to lung contamination after death or poor aseptic techniques during thoracotomy, bronchial colonization, and bronchiolitis. As regards contamination after death, some authors have reported a high incidence of positivity of lung cultures despite the absence of any relationship with in vivo processes of pulmonary infection. [25]The delay between death and pulmonary sampling could explain some false-positive results. However, the percentage of false-positive results in this study was less in samples obtained within 90 min, compared with those obtained within 30 and 60 min. This could be explained because the majority of patients in whom the extraction procedure was performed within 30 min after death were organ donors. In those patients, brain death occurred 12-24 h before the study protocol, and perhaps this fact favored the presence of pulmonary colonization. Concerning the technique of thoracotomy, this was performed under strict aseptic methods, minimizing contamination during the procedure. Bronchial colonization does not explain the false-positive results, because the pulmonary biopsy samples were taken from the periphery of the lung, making the presence of bronchial fragments into the biopsy unlikely. Finally, bronchiolitis is thought to contribute to some of the false-positive results. [11]Bronchiolitis accounted for 8% of the false-positive cultures in our study.

A possible association between antibiotic treatment and bacterial growth was evaluated. A relationship was observed between the high bacterial counts and the absence of prior antibiotic treatment in all pneumonia stages. A significant reduction in distal airway bacterial concentration was shown after 3 days of antibiotic treatment. [26]These data strengthen the importance of distal airway sampling before starting antibiotic therapy. However, the impact of prior use of antibiotics on lung cultures is difficult to assess with our study design due to the lack of control patients.

In summary, ventilator-associated pneumonia was found to be a polymicrobial, multifocal, and dynamic histologic process with little relationship to the bacterial culture results. These histologic and microbiologic features of ventilator-associated pneumonia may explain the difficulties in validating both invasive and noninvasive methods for the diagnosis of this common pulmonary complication related to mechanical ventilation.

The authors thank Dr. H. K. F. van Saene, for suggestions during the preparation of this manuscript.