Sequential respiratory, psychologic, and immunologic studies in relation to methyl isocyanate exposure over two years with model development.

Of 113 methyl isocyanate (MIC)-exposed subjects studied initially at Bhopal, India, 79, 56, 68, and 87 were followed with clinical, lung function, radiographic, and immunologic tests at 3, 6, 12, 18, and 24 months. Though our cohort consisted of subjects at all ages showing a varied severity of initial illness, fewer females and young subjects were seen. Initially all had eye problems, but dominant symptoms were exertional dyspnea, cough, chest pain, sputum, and muscle weakness. A large number showed persistent depression mixed with anxiety, with disturbances of personality parameters. The early radiographic changes were lung edema, overinflation, enlarged heart, pleural scars, and consolidation. The persistent changes seen were interstitial deposits. Lung functions showed mainly restrictive changes with small airway obstruction; there was impairment of oxygen exchange. Oxygen exchange improved at 3-6 months, and spirometry improved at 12 months, only to decline later. The expiratory flow rates pertaining to large and medium airway function improved, but those for small airways remained low. There were changes of alveolitis in bronchoalveolar lavage fluid on fiber optic bronchoscopy, and in 11 cases positive MIC-specific antibodies to IgM, IgG, and IgE were demonstrated. On follow up, only 48% of the subjects were clinically stable, while 50% showed fluctuations. Thirty-two percent of the subjects had lung function fluctuations. Detailed sequential behavior over 2-4 years was predicted for dyspnea, forced vital capacity, maximum expiratory flow rate (0.25-0.75), peak expiratory flow rate, VO2, and depression score. A model for clinical behavior explained a total variance of 52.4% by using the factors of cough, PCO2 and X-ray zones in addition to above five parameters. The behavior of the railway colony group (1640 patients) revealed a similar pattern of illness. When this observed pattern of changes was transferred to the affected Bhopal city sections (with an equitable age-sex distribution), our model results were again validated. Thus the picture of MIC-induced disease seems similar despite the differences for age-sex and initial severity of illness in our cohort and in the population of Bhopal city as predicted by our model. ImagesFIGURE 1.FIGURE 2.


Introduction
A leak ofmethyl isocyanate (MIC) from a 40-ton tank occurred in Bhopal, India, on December 2 and 3, 1984. Approximately 0.2 million residents were exposed to high concentrations (over 27 ppm) of MIC. As a result, it is believed that 2500 died within hours, about 3100 within a month, and 3289 over 4 years. In addition, in the first few days, many cattle, fowl, and other large animals also perished.
regularly for 2 years. We also developed a model on the basis of observed behavior for subjects in the railway colony and on Bhopal citizens ofaffected areas. This paper details our findings.

Materials and Methods
Persistently symptomatic residents ofBhopal (exposed to the MIC leak) who were seen at our hospital in Bombay within 7-90 days of the leak form the study cohort. Subjects were followed after 3, 6, 12, 18, and 24 months using a clinical respiratory questionnaire modeled on the BMRC (British Medical Research Council) questionaire, which was used earlier in our work on occupational diseases (6).
In addition to physical examination, the following tests were conducted: chest radiography, spirometery (Stead Wells Spirometer, W. E. Collins, Boston, MA); bronchodilator tests with nebulized ,3 stimulant aerosol; FV loop studies (Fleisch Pneumotachograph with Collin's bodybox and Hewlett-Packard X-Y recorder; minute ventilation, oxygen uptake at rest and during submaximal exercise (with clinical end point at 4-5 min); and (at 0 and 3 months) arterial blood gases at rest and with exercise with carboxyhemoglobin (COHb) and methemoglobin (MetHb) (CO-oximeter and blood gas analyzer; Instrumentation Laboratories, Boston, MA). At later times COHb and MetHb were done only in selected abnormal subjects. Clinical grading of respiratory illness was done on the basis ofthe number and severity of symptoms; a score of0-2 was considered mild, 3-4 moderate, and above 4 severe.
From the FV loop, we manually calculated gas flow (V) at 25, 50, and 75 %, peak expiratory flow rate (PEFR), and PIFR (inspiratory and expiratory flow) rates at respective phases and peak values). The transducers were calibrated before and after each reading, and an integrator was used for deriving the volume. Oxygen uptakes were calculated from volumes measured in a Douglas bag from the collected expired air. The chest radiographs were read together by two observers and classified as normal, emphysema, cardiovascular changes, or parenchymal and interstitial deposits. The latter were categorized as punctate, linear, micronodular, or reticular, along with their zonal distributions.
Serial observations at 3 months and later included psychiatric, psychosocial, and personality dysfunction in the cooperative group. In 12 subjects seriously ill at 1-3 months, we performed flexible fiber optic bronchoscopy with bronchoalveolar lavage. In eight subjects, lung biopsy was performed (five by needle and three open). From the moderately ill group, 52 ECGs and 32 RA tests and antinuclear factors were measured at the initial assessment. In a few cases with large broncho-reversibilty, graded histamine airway hyperreactivity, lung volumes, and compliance were measured at 3 and 6 months. Immunologic studies (in collaboration with University of Pittsburgh, Pittsburgh, PA) with radioallergic (RAST) and MIC-specific antibodies were done in 99 subjects. These tests were done by preparing an antigen tagged with MIC tohuman serum albumin (MIC-HAS). Its efficacy and specificity were tested in guinea pigs, and antibody titers were measured by enzyme-linked immunoassay (ELISA) inhibition for IgG, IgM, and IgE. We also measured total IgE and RAST IgE with MIC-HSA (4).
Statistical analyses were done by Student's t-tests, proportional chi-square tests, and, for antibody titers, by linear regression.

Model Development
After the initial analyses on the original cohort at 0 (n = 113), 3 (n = 79), 6 (n = 56), 12 (n = 68), 18 (n = 68), and 24 (n = 87) months by paired t-tests, we did intercorrelations by Pearson's correlation coefficients at all phases separately and together. This helped us to identify significant variables to be used for evaluating sequential changes. To define changes in a more regular cohort (n=81; defined as those who attended at 0 and 24 months and at least once more), we substituted missing values with the average behavior and compared the individual behavior. The sequential behaviors in the original cohort, as actual patterns of 24 selected tests, were categorized as stable, improving, worsening, worsening after early improvement, and fluctuating. From these categories, we selected nine parameters for multifactorial regression (cough, dyspnea, forced vital capacity [FVC], PEFR, maximum expiratory flow rate [MEFR 0.25-0.75], Pco2, X-ray zones, Hamilton depression score, and oxygen uptake [Vo2]).
To account for differences ofage-sex distribution between the original and regular cohort and between the railway colony patients and Bhopal city population, the following method was used: for both sexes and three age groups (0-19, 20-44, and 45+ years) numbers were calculated separately for actual proportions in each of the populations. From mean values of parameters in the cohort, a new mean value was derived. On the basis of initial known distribution in each group (as modified by age-sex), we assumed for later periods a worse transition probability for sequential changes. This was tansfrred to the new population; the progress was found to be similar in all age-sex groups. The originally selected (nine) variables were then subjected to multifactorial regression for all six phases, giving six Eigen values. From these two-factor composite equations we used the SPSS (Berkeley, CA) program to derive two-factor scores and severity scores for close-factor analyses. We obtained Eigen values and derived variance to explain the cohort behavior by this method. Table 1 lists the sex-age distribution of the original cohort studied at our institute. Ifthe distance of exposure from the gas leak was greater, there was a greater chance of initial clinical illness being mild. A slighdy greater proportion of severe c.ases in our cohort were seen as compared to the cases treated in the railway colony. (See below; two-thirds ofour cohort came from this group.) In comparison with the initial clinical severity of illness, when we assessed the cases at our institute, 69% had improved, 4% had worsened, and the rest were unchanged. Table 2 lists the sequential patterns in clinical synptoms under three categories. With six assessments done over 2 years on this cohort, we restudied 68 at 12 months and 87 at 24 months. The frequency (percent) ofchest symptoms is listed as neurologic or psychosomatic as assessed by qualified psychiatrists. Psychiatric testing was not done at the first stage as the subjects were very ill and it was perhaps cruel to do prolonged testing immediately after the tragedy; the tests were also not possible due to not obtaining cooperation in a proportion of cases.

Results
Respiratory symptioms scores decreased at 3, 6, and 12 months but increased significantly later. The commonest symptom was  testing was not done at the first stage as the subjects were very ill and it was perhaps cruel to do prolonged, testing immediately after the tragedy; the tests were also not possible due to not obtaining cooperation in a proportion of cases. Respiratory symptoms scores decreased at 3, 6, and 12 months but increased significantly later. The commonest symptom was dyspnea on exertion; a paroxymal component was unusual. These subjects were intermittently given antibiotics and bronchodilators by local doctors, but despite this, the symptoms persisted. From neurologic symptoms spontaneously elicted, muscle weakness and poor memory became more frequent (0.05). The eye problems (which were subsequently present in 2-10%) and abdominal symptoms (seem unrelated) are not listed. Severl more symptoms were diagnosed by a psychiatrist. Perhaps some ofthese were related to procedures in the protracted legal compensation case. Table 3 lists the psychiatric abnormalities as assessed by psychiatrists and social workers. Table 3 shows that only 19-27% were considered normal. The proportions with pure anxiety or depression increased over 2 years but those with mixed lesions decreased (p < 0.05). Hamilton scoring revealed that the proportions with normal scores for both anxiety and depression declined over 2 years (p<0.05). Those with mild abnormalities simultaneously increased. These assessments also covered memory count, I.Q. assessment, and Bender-Gestalt scores. For the latter two scores, the proportions with normal scores increased significantly after 12 months (p< 0.05), while for memory count there was no improvement. Table 4 reports results ofpersonality dysfunction and psychosocial studies that were obtained after filling a detailed proforma and checking. The results show a large prevalence of post-MIC exposure abnormalities. In six categories there was significant partial improvement, best seen at 18 months. Table 4 lists increased prevalences ofabnormal psychosocial function after Table 4. Sequential patterns in personality and psychosocial function in MIC-exposed subjects. ficant partial improvement, best seen at 18 months. Table 4 lists increased prevalences ofabnormal psychosocial function after MIC exposure. Some assessments were omitted in a few subjects, where considered not applicable (e.g., job, marital). The total abnormality score at 3 and 12 months correlated significantly to FVC and FEVy (forced expiratory volume in 1 sec), suggesting that these abnormalities were a consequence oflung functional disability. These persistentpsychologic abnormalities may be secondary to organic lung disease.
Radiographic and Lung Function Changes Table 5 lists chest radiographic patterns. Only 2-4% offilms over 2 years were read as normal by our criteria. The proportions having overinflation, pleural scars, heart enlargement, and Table 5. Serial radiographic patterns in the MIC-exposed (n = 113) cohort.
Month (n) Pattern 0(113) 3 (79) 6(56) 12 (68) 18(68)  consolidation declined by 6 months and later (p< 0.05). At the early stage, only in a few, consolidations responded to antibiotics and in one case distinct calcified scars developed at the right base. However, the main changes seen were interstitial shadows, which were distributed in fewer zones after 12 months; but generally 3-4 zones were involved with linear or punctate deposits. The punctate deposits were seen less often at 18 and 24 months, and micronodular infiltrates decreased at 12 months or later (p<0.05 for both). Table 6 lists the mean trends in lung function and blood gases. While at 0 and 3 months, all subjects cooperated with blood gas studies, at later periods (not done at 6 months) we restricted these to more abnormal subjects. Oxygen uptake at rest did not show further change after 3 months (Table 6).
Exercise Vo2 increased adequately at 3 months, and the improvement was maintained (paired comparison in 44 subjects: 1 123 ± 280 mL atO, 1157 ± 403 mL at 3 months, p < 0.05). Initially Po2 dropped significantly with exercise in 39% subjects. Po2 at rest did not change significantly, but Pco2 rose at 24 months (p<0.05), and pH declined between 12 and 24 months (p< 0.05); COHb and MetHb decreased to near normal levels by 3 months (p< 0.05). There was no relationship ofthese values to the initial clinical severity. Table 6 lists lung functions as true values. Ifthese are evaluated as standardized (to age, height, and sex) and compared statistically, the conclusions remain similar. Also, ifone restricts the analyses to a common group, the trends seen are unchanged (it seemed that more improved cases defaulted easily). Thus, FVC and FEV1 improved by 12 months, then declined later. The PEFR values increased by 12 months and later maintained improvement. MEFR (0.25-0.75) values did not improve, nor did MEFR (200-1200). There was a significant reduction of FEV1/FVC % by 12 months that persisted. Significantbronchodilator response was seen in a small proportion, which did not seem to accentuate later. Initially only 11.6% showed an improvement in FEVy (between 11 and20%) and in 8.9% itwasabove20%. At 3 months, the respective proportions were 7.7 and 8.8%. Thus, there seems Tible 6. Sequential lung function trends in MIC-eposed (n = 113) cohort.a Month (n) Measurement 0 (113) 3 (79) 6 (56) 12    to be no increase in "asthmatic" tendency after MIC exposure. In 10 cases tested for graded histamine airway reactivity (0.03-16 mg/mL concentration) at 3 and 6 months using stand-ard techniques, 5 reacted at 1 mg/mL concentration (PC20 [provocative concentration: 20% decrease in FEVIJ) while 3 showed PC20 at 5 mg/mL. The changes in DMBC (direct maximal breathing capacity), RR (respiratory rate), or MV (minute ventilation) were insignificant. In four cases with severe disability, there were no abnormalities of response to graded hypoxia. When the initial FVC and FEVy were correlated to respiratory symptom scores, there was a very significant relationship (FVC: r = -0.76; FEV,: r = -0.79; postbronchodilator FEV1: r = -0.85). Thus we concluded that our clinical symptomatology was reflected in the functional values measured.
Immunologic studies done in 99 subjects collected at intervals are shown in Table 7. Total IgE levels showed insignificantly high levels initially, which later declined. Preliminary studies proved that antibodies specific to MIC (as GSA or HSA) were detected by ELISA inhibition assays, and these did not crossreact to other isocyanates or conjugates. Thus, MIC-KLH was shown to inhibit antibody specific to MIC. Upon storage, the sera, due to lag from delay in sending samples to United States, did not decay (which may have affected their evaluation). Mean RAST IgE binding revealed levels from 2.31 ± 2.95% ((0 months) to 3.68 ± 4.95 % (12months). While these were related closely to total IgE levels values, both were unrelated to clinical or functional changes seen. In 11 cases, MIC-specific antibodies were detected (IgM, 7; IgE, 4; IgG, 6) on several occasions. in one case antibodies were detected 1 year after exposure, but in most subjects, these were found in a sample taken within 3 months. In 10 subjects with clinically adequate data, 4 had severe and 6 had moderate initial illness. Radiographically, one to three zones showed interstitial deposits in three subjects and four to six zones in seven cases. It appeared that those ill initially but who improved later developed positive antibodies. This finding of MIC-specific positive antibodies, albeit with low titers, has crucial medical and legal significance for a cause-effect relationship. Table 7. Immunologic results in MIC-exposed (n =99) subjects.

Defaults and Relationship to Clinical and Functional Sequence
Though we had planned a full follow-up for 2 years in six phases, we did not succeed fully. At 1 and 2 years, we followed 68 and 87 subjects. Of these, those who were assessed functionally and clinically on at least two subsequent occasions (including the last) were included in the regular (81) cohort. When these were compared, it was seen that defaulters (32) were slightly less abnormal initially and improved more. On preliminary analysis, it was seen that while many showed improvement up to 12 months and deteriorated later, there were other patterns, e.g., aSDs omitted as they were similar to those listed in Table 7. aSDs similar to values for the cohort in Table 6. bn = 48 at 0 months.

Results in Regular Cohort
stable, improving, worsening, fluctuating, and improvment The results in the regular cohort (n = 81) are presented in followed by deterioration. The general behavior was that Tables 9 and 10. Table 9 shows that despite the data being 25-30% were constantly improving and one-half belonged to the restricted to 81 cases, the trends for various parameters are last category. similar; the significance of differences are also the same. Table  We decided to look at this problem in detail for 15 parameters. 10 gives results for depression, anxiety scores, and IgE. Again In those where follow-up data showed a clear trend, we graphthe changes are similar to those in the original cohort. ically derived their patterns individually; the results are shown When distribution of clinical chest symptoms and signs are in Table 8. It is seen that all parameters do not change identically, compared in six phases along with severity of initial clinical lTble 11. Correlation coefficients in the total cohort at the initial stage.   illness (not tabulated), the trends are similar to the total cohort. More older females suffered greater illness initially, but later the behavior in both sexes was similar. Therefore, we worked out trends in correlation coefficients (as a guide to regression analyses) on the total cohort. These were worked out on all possible parameters in six phases individually and together, but results on more significant values covering all phases are listed in Tables  11 and 12. Intercorrelation over Two Years Functional parameters were closely interrelated. At the initial stage (Table 11), cough, sputum, and dyspnea are moderately interrelated, along with abnormal lung signs (but not overinflation). Sputum is significantly related to FEVI, MEFR(0.25-0.75), and MEFR (200-1200). Dyspnea, however, is not significantly correlated to lung functions, but cough is correlated to FEVy and MEFR (200-1200). Lung signs correlated to FVC and FEVy, while MetHb correlated to MEFR (0.25-0.75) and PEFR and Po2 to MEFR (0.25-0.75). Table 12 correlates the coefficients covering all six phases. Due to larger volumes of data, several more correlates now seem significant. Thus, chest symptoms and signs are interrelated, as are lung functions, blood gases, DMBC, Vo2, scores for anxiety and depression. IgE levels correlated to MetHb values.
From these results we concentrated on the following parameters in modeling for a sequential behaviour in the regular cohort: a) clinical: cough, dyspnea; b) function: FVC, MEFR (0.25-0.75), PEFR, Vo2, Pco2; c) X-ray zones and depression score. parameters. For dyspnea, there were four patterns with 45.7 % as stable (Fig. 3). For FVC, the major pattern (of four) was fluctuating (35.8%), for MEFR (0.25-0.75), the major pattern (of five) seen was worsening (40.7%), followed by worsening after early improvement (24.7%) (VE75 [expiratory flow at 75% of VC] from FV loop tends to behave similarly). For PEFR, the main pattern was improvement (53.1%) (as also for Vo2, 39.5 %). For depression score, the main pattern was fluctuating. Thus, on the regular cohort we predicted changes in six parameters fairly accurately. Table 14 shows the further development ofthis concept by doing mutifactorial analyses. Analyses were conducted on seven parameters listed for factor 1 at all six phases and later only for the first and sixth phases. Thus, as step 1, six (Eigen value) equations were derived and as step 2 factor scores as listed in Table  14 were derived. Corresponding factor scores for six variables with large values in the first equation and smaller ones in the second equation were worked out. Then the factorial score for two factors and severity scores were obtained, accounting for variance. Eigen values for two sets of variables were derived. When FVC, MEFR (0.25-0.75), and PEFR, which are closely related, are accounted for in the first effort, correlation with cough, dyspnea, PCO2, and X-ray zone become significant (Table 14). By variance analyses in these two attempts we could explain a variance of 36.5 and 15.9%, respectively (Table 14). Thus, using seven parameters, our model could explain behavior of52.4% (a model with moderate accuracy). Ifone correlated the clinical severity, the model prediction is significantly correlated (p < 0.01). We validated our model by working on the other four phases. As the behavior over 2 years ofMIC-exposed cases was not different by age/sex, we thought this multifactorial model could be applied to other cohorts or Bhopal city population also. Table 15 describes these attempts to transfer this behavior to the railway colony cohort. It describes the age-sex distribution according to the severity of initial clinical illness seen from the actual records oftreated cases scrutinized and classified by the senior author. A total of 1640 subjects were treated, and in 1617 cases, data were adequate for such classification. Table 15   . Similar patterns of forced vital capacity shown n milliliters on vertical axis for the month ofmeasurement. Note that the behavior is different from that in Figure 1. that in the railway colony cohort, there are a greater number of younger subjects and more females, along with an excess ofmildly affected cases. On comparison for these behaviors, in this population and our regular cohort, we found no statistical differences. The differences we looked for were the clinical, functional, and psychologic parameters. Table 16 gives additional data on age, sex, types ofclinical initial symptoms, and severity of illness. The admitted group consisted of a larger proportion of moderate and seriously ill cases. In a large majority, the initial illness lasted up to 2 weeks and in 30.3 % it persisted to become chronic. In those admitted, the illness, perhaps due to treatment, seemed to have remitted more often. Table 17 details results by applying our factorial model to predict in three severity categories the clinical behavior for 2 FIGURE 8. Patterns of Hamilton depression scores. Each of these parameters seem to have peculiar variations.
years on the railway colony cohort. At 0 months, older subjects had more severe illnesses. The proportion in the mild category increased later, particularly at younger ages. But for all ages, the moderate category predominates in later periods, particularly at older ages. This means that after some improvement (which was greater in younger subjects), there is enhancement of symptoms and their severity over 2 years. Table 18 projects age-standardized (to account for any differences) sequential behavior of four functional parameters and depression as per our model on the railway colony cohort. Behavior is similar to our regular cohort and broadly shows a fluctuating course; for FVC, maximal improvement in all three groups is seen at 18 months (irrespective of initial severity). For MEFR (0.25-0.75), there is a gradual reduction over 24 months to a level that is slightly lower in the severe group at 24 months (though differences are large initially). For PEFR, the trends are similar to FVC, but with no differences between three groups of clinical severity of illness. For V02, similar trends are seen. For  depression scores, it is seen that in all three groups the scores worsen up to 18 months and then stabilize (slightly lower in mild group only). As our model was derived on a small cohort based on 0and 24-month data, data were validated for all six phases; our attempt to project behavior ofthe railway colony cohort was successful. As there were no significant differences in sequential behavior by age or sex, we applied this further to the affected Bhopal city population. Table 19 gives the age-sex distribution and clinical severity of illness in the Bhopal city population. This was derived from a full census of the inner city and a random, stratified sample of three variably affected communities. As per design ofthe Indian Council of Medical Research (ICMR), the cohort was to be representative ofthe general city. Here the age-sex distribution is spread evenly, and distribution ofclinical illness grades more equitably. While there are several significant differences between the three populations (our regular, railway colony, and the ICMR cohorts), as the course of MIC-induced disease did not seem to differ by age or sex, we do not think these differences would invalidate application ofour predicted model to the general city population. Table 20 gives projected age-sex standardized (to account for significant differences) sequential behavior offive parameters for the Bhopal city population over 2 years. Thus, for three grades ofclinical illness, the behavior does not appear significantly different (therefore listed separately in three groups only for FVC). The patterns of MEFR, PEFR, Vo2, and depression are also essentially similar to the railway colony cohort. Table 20 also shows somerelevantdataon abnormal lung symptoms asobserved at 3 months and for deaths/1000 at 3 months and 2 years in the three zones ofclinically affected grades ofcity population. It confirms that three zones of differently affected degrees of illness show different morbidity.  We may conclude that the model developed on our cohort of 113 subjects could be applied to the Bhopal city population for judging the sequential behavior of MIC-induced disease.

Discussion
As our findings up to 6 months have already been published (1)(2)(3), it is fair to state that our earlier evidence of interstitial restrictive lung pathology with small airway disease is confirmed. The airway component seems to have progressed, leading to a fluctuating course with persisting disability. We found persistent flow rate abnormalities, particularly for V(E75) (E indicates expiratory flow rate; 75 indicates expiratory flow rate at 75% of VC). For V(IP) (IP, inspiratory peak flow rate), V(I50) (I indicates inspiratory flow rate; 50 indicates inspiratory flow rate at 50% ofVC), and V(I75) the improvement was maintained, and for V(EP) (expiratory peak flow rate), V(E75), and V(E50) there was a significant decline after 18 months. There were also persistent abnormalities of flow-volume loop (3).
Studies done by Alarie et al. (5) indicate that severe airway obstruction persists after a single MIC exposure (37 ppm for 3 hr) and recovery from pulmonary effects is very slow. Recovery from MIC was considered to be slower than from H2SO4, toluene diisocyanate, and wood smoke (5). Alarie et al. claimed that the recovery pattern was similar to that following exposure to smoke from polyvinyl chloride (7). They found VE (expired minute volume) particularly to be reduced, and VT (tidal volume) and respiratory rate did not increase after MIC exposure (5). The type of FV loop abnormalities we reported (3) were similar to those seen by Alarie et al. (5).
Alarie et al. also did not find a cyanidelike effect (5) after 37 ppm of MIC. In view ofthe controversy ofthe cyanidelike effects (e. g., a common complaint of persistent muscle weakness) in Bhopal subjects, this may be important. While we have found evidence of rise in COHb and MetHb initially, possibly attributable to effects of breakdown products of MIC, the psychosomatic symptoms of the type discussed earlier seem to be correlated to organic lung disability. As organic lung damage is not known to occur in chronic cyanice poisoning, we think our sequential results also provide evidence against such a therory. Isocyanates have been associated with occupational asthma (8). Our results show, along with chronic small airway disease, evidence of restrictive lung disease (1,2). These changes may be similar to those observed by Charles et al. (9). What was not known in earlier studies, for MIC-induced disease, was its chronic fluctuating course, for which our follow-up data and model provide evidence.
Schwetz et al. (10) found that 1-3 ppm MIC exposure 6 hr/day for 4 days in mice resulted in a significant number ofdead fetuses at birth and a significant decrease in neonatal survival. In the railway colony cohort, there were frequent spontaneous abortions (reported also in exposed areas of Bhopal), but there was no evidence of increased fetal abnormalities experimentally (10) or in the city population )11).
Other aspects of immunologic, mutagenic, and genotoxic effects due to MIC have been studied by Deo et al. (12). These stueies revealed (on a similar group of67 exposed and 15 control subjects) minor chromosomal aberrations, low responsiveness of lymphocytes to phytohemagglutinin delay in cell cycle, increase in T-lymphocytes, and lowering of sister chromatid ex-changes. However, such abnormalities were inconsistent and minor; this is confirmed by Tice et al. (13). They found that MIC exposure for 4 days resulted in a small increase in chromosomal aberrations and sister chromatid changes; but no genotoxicity or mutations were found (14). Similar to our findings, specific antibodies have been reported after TDI occupational exposure, which showed a fluctuating, increasing, or declining course (15).
Bucher et al. (16) observed granulomatous inflammation with persistent lung damage, intraluminal airway fibrosis, and bronchiolitis after a 2-hr exposure to MIC in rats (17k. Their studies on repeated exposure to 1-6 ppm MIC did not result in significant direct effects on nonrespiratory tissues (18). Our observations conform to this pattern and both corroborate pesistent pulmonary damage.
Andersson et al. (19) have presented some data from Bhopal community studies that indicated variations in morbidity according to wind direction. Thus, the railway colony, situated southeast ofthe factory, suffered a high mortality. Our follow-up data indicates that affected subjects with varying initial severities of illness have a similar course.
Fedde et al. (20) reported low Po2, increased metabolic acidosis, tissue hypoxia, and increased ventilation-perfusion imbalance due to bronchiolar obstruction after a high exposure (240-628 ppm) ofMIC for 15 min. Our findings ofearly hypoxia and later ofsmall airway obstruction confirm their data; however, within the first few days, there was respiratory alkalosis, but not hypercapnia and acidosis. Several advances have been made in understanding MIC-induced disease since the first article by Kimmerle and Eben (21). With exposure at 23 ppm MIC, they found heavy breathing, lung edema, and irritiation of mucosa (21). Because ofthe continued wide use ofisocynates, there may still be large interest in their systemic toxicity. In a study of783 cases, 2-4 months after exposure to MIC, Rastogi et al. found that females were more seriously affected (22); although major change in lung function was restrictive, obstruction was noticed in some. In another paper from this group on epidemiologic data (23) in a study of 1109 subjects, most showed respiratory, cardiovascular, abdominal, and skeletal symptoms. Their description ofacute cases (24) was similar to other observations (25). Their experimental studies after a single MIC exposure at 3.2 mg/L for 8 min. (26) showed thickening of alveoar septa, lymphoid hyperplasia, peripheral emphysema,peribronchial edema, and cellular infiltrates with exfoliatedbronchiolar epithelium. these findings are similar to those of Bucher et al. (16)(17)(18).
Our clinical and functional data on follow up amplify in detail some ofthe lung changes. Our development ofa predictive model is unique in an occupational setting, although this has been done by several researchers on experimental or clinical problems. This effort might help in delineating the future course ofhuman effects in relation to such an industrial disaster.