Short-term air pollution exposure decreases lung function: a repeated measures study in healthy adults.
Study subjects
Cohort of 2449 healthy adult employees in rural northern Belgium
Employees of SCK-CEN, VITO, BELGOPROCESS, BELGONUCLEAIRE and the European School, all attending the annual SCK-CEN check-up. Age 16-70 years, mean 37 (IQR 18) at entry; 1756 (72%) male; 94% Caucasian; 1739 (71%) never-smokers, 368 (15%) former and 342 (14%) current smokers, with smoking status unchanged over the study for 98% of the sample; on average 37% received free influenza immunisation. Mostly office workers with a college degree and similar socio-economic background, with a few blue-collar workers; no further profiling was possible because the dataset was anonymised beforehand. BMI is reported inconsistently in the source - 25.4 kg/m2 with IQR 4.8 in the Results text, 25.4 +/- 4.0 (SD) in Table 1 - and is not recorded as a numeric value here since the schema has no BMI slot for a subject. Asthmatic status could not be included in the analysis because it existed only in non-digitised written medical files.
Routine annual occupational health surveillance; participants were unaware of the study objectives, which the authors note makes participation bias very unlikely. Not a clinical population: these are working adults 'normally not considered to be vulnerable to current levels of air pollution in Belgium'.
Lung function assays
Repeated spirometry in 2449 Belgian adult workers under ambient air pollution
Annual occupational-surveillance spirometry (FVC, FEV1, FEV1/FVC, PEF) in 2449 healthy adult employees over 5404 clinical visits between 26/01/2011 and 30/01/2015, analysed against daily ambient PM10, NO2 and ozone at three lags. Most participants (35%) had 3 valid measurements during the study period; 28% had 2, 25% a single one, and 12% more than 3. Measured concentrations were well below the European limit value of 40 ug/m3 but exceeded 50 ug PM10/m3 on 26, 18, 14 and 13 days in 2011-2014, above the WHO recommended maximum of 3 days/year — the setting the authors stress, since effects appear at concentrations that comply with EU regulation. Mixed models separating within- from between-subject exposure showed the reported associations are driven by within-subject effects, while between-subject effects were not significant. Secondary analyses found no differential effect of sex, age (above or below 30) or smoking status, and sensitivity analyses excluding participants who changed smoking status or measurements taken during the influenza season did not change the conclusions. Limitations the authors state: asthmatic status could not be included (it was only in non-digitised written medical files), residential addresses and travel history were unavailable, no personal monitoring was possible, and the spirometry protocol changed from three to two manoeuvres mid-study. No numeric value is recorded for the exposure-stratified or sensitivity analyses because the paper reports them as 'data not shown' or only in figures. The 24-hour exposure_duration on every exposure condition encodes the daily averaging window of the model's exposure metric (validated daily concentrations), not a duration the paper reports; the paper's daily maximum 8-h average ozone metric is used only for the health-target comparison, not in these models.
Exposure conditions
| Condition | Agent | Concentration | Duration |
|---|---|---|---|
| Ambient rural air at the study site, 2011-2015 (the whole cohort; no unexposed comparison group exists) COND:intpanis-ambient-air |
air ENVO:00002005 |
— | — |
| Ambient PM10 on the day of the lung function test (lag 0) EXPOSURE:intpanis-pm10-lag0 |
respirable suspended particulate matter ENVO:01000405 |
24.9 microgram per cubic meter (mean; standard deviation ±15.5 microgram per cubic meter) | 24 hour |
| Ambient PM10 one day before the lung function test (lag 1) EXPOSURE:intpanis-pm10-lag1 |
respirable suspended particulate matter ENVO:01000405 |
24.3 microgram per cubic meter (mean; standard deviation ±15.6 microgram per cubic meter) | 24 hour |
| Ambient PM10 two days before the lung function test (lag 2) EXPOSURE:intpanis-pm10-lag2 |
respirable suspended particulate matter ENVO:01000405 |
23.9 microgram per cubic meter (mean; standard deviation ±15.4 microgram per cubic meter) | 24 hour |
| Ambient NO2 on the day of the lung function test (lag 0) EXPOSURE:intpanis-no2-lag0 |
nitrogen dioxide CHEBI:33101 |
23.1 microgram per cubic meter (mean; standard deviation ±9.6 microgram per cubic meter) | 24 hour |
| Ambient NO2 one day before the lung function test (lag 1) EXPOSURE:intpanis-no2-lag1 |
nitrogen dioxide CHEBI:33101 |
21.2 microgram per cubic meter (mean; standard deviation ±10.0 microgram per cubic meter) | 24 hour |
| Ambient NO2 two days before the lung function test (lag 2) EXPOSURE:intpanis-no2-lag2 |
nitrogen dioxide CHEBI:33101 |
20.0 microgram per cubic meter (mean; standard deviation ±10.1 microgram per cubic meter) | 24 hour |
| Ambient ozone on the day of the lung function test (lag 0), warm season May-September EXPOSURE:intpanis-o3-lag0 |
ozone CHEBI:25812 |
44.3 microgram per cubic meter (mean; standard deviation ±19.0 microgram per cubic meter) | 24 hour |
| Ambient ozone one day before the lung function test (lag 1), warm season May-September EXPOSURE:intpanis-o3-lag1 |
ozone CHEBI:25812 |
47.0 microgram per cubic meter (mean; standard deviation ±18.7 microgram per cubic meter) | 24 hour |
| Ambient ozone two days before the lung function test (lag 2), warm season May-September EXPOSURE:intpanis-o3-lag2 |
ozone CHEBI:25812 |
46.6 microgram per cubic meter (mean; standard deviation ±19.4 microgram per cubic meter) | 24 hour |
Outputs
| Group | Output | Description |
|---|---|---|
| treated | Observed lung function pooled over all 5404 clinical visits (Table 1) LF:intpanis-spirometry-cohort-output-ambient · under COND:intpanis-ambient-air |
Arithmetic mean +/- SD of each spirometry parameter across all 5404 clinical visits of the 2449 participants, as reported in Table 1. These are absolute observed values under ambient exposure, not a treated-group response: the cohort was continuously exposed and the study has no unexposed arm, so experimental_group is treated and measured_under points at the ambient-air condition rather than at any single pollutant/lag metric. The paper also reports an FVC range of 1.89 to 10.84 L (mean 4.71 L, IQR 1.49) and a FEV1 range of 0.9 to 7.35 L (mean 3.79 L, IQR 1.18), and sex-specific mean FVC of 5.06 L in males and 3.67 L in females; the SD values recorded below are the Table 1 figures.
fvc: 4.71 liter (mean; standard deviation ±1.03 liter; n=5404)
fev1: 3.79 liter (mean; standard deviation ±0.83 liter; n=5404)
fev1 fvc ratio: 80.9 percent (mean; standard deviation ±7.5 percent; n=5404)
peak expiratory flow: 8.78 liter per second (mean; standard deviation ±2.20 liter per second; n=5404)
FVC, L 4.71 ± 1.03 Table 1 pulmonary outcome row: the pooled mean and SD FVC (4.71 +/- 1.03 L) recorded on this output record. FEV1, L 3.79 ± 0.83 Table 1 pulmonary outcome row: the pooled mean and SD FEV1 (3.79 +/- 0.83 L) recorded on this output record. FEV1/FVC, % 80.9 ± 7.5 Table 1 pulmonary outcome row: the pooled mean and SD FEV1/FVC ratio (80.9 +/- 7.5 %) recorded on this output record. PEF, L/s 8.78 ± 2.20 Table 1 pulmonary outcome row: the pooled mean and SD PEF (8.78 +/- 2.20 L/s) recorded on this output record. |
Evidence
To evaluate the association between daily changes in air pollution and lung function in healthy adults we obtained annual lung function measurements from a routine worker health surveillance program not designed for research purposes. Forced Vital Capacity (FVC), Forced Expiratory Volume in the first second (FEV1), FEV1/FVC and Peak Expiratory flow (PEF) from a cohort of 2449 employees were associated with daily measurements of PM10, NO2 and ozone at a nearby monitoring station in the North of Belgium. Repeated measures were available for the period 2011-2015.
States the study design, the cohort size, the four spirometry outcomes and the three pollutants this assay records, and the 2011-2015 repeated-measures period.
A total of 5404 clinical visits with lung function test were obtained for the 2449 participants.
Gives the number of clinical visits with a lung function test (5404) behind the pooled Table 1 spirometry values and the PM10 and NO2 comparisons in this file, and the 2449 participants they came from. The ozone comparisons are restricted to the warm season (May to September), a subset of these visits whose n the paper does not report, so they carry no sample_size.
A measuring session was considered usable when at least two manoeuvres, free from artefacts, passed the test acceptability criteria (quality grade C).
States the acceptability criterion for a usable measuring session that the spirometry protocol on this assay records.
Participants performed three respiratory manoeu- vres (maximum exhalation) during the clinical visits in the years 2011, 2012 and 2013. Because of the high reproduci- bility of the measurements, most of the participants per- formed only two manoeuvres in the years 2014 and 2015.
Documents the mid-study protocol change recorded on the spirometry protocol: three respiratory manoeuvres in 2011-2013, two in most participants in 2014-2015.
The same device was used throughout the study.
Supports the protocol claim that a single spirometer was used for the whole study, which is why no between-instrument variation is modelled.
Ambient air pollution levels were measured at an official air quality monitoring station of the Flanders Environment Agency (VMM) in Dessel (station number 42 N016)
Identifies the fixed ambient monitoring station that produced every exposure concentration recorded on this assay, and its 6 km distance from the medical centre.
the day before (lag 1) or 2 days before (lag 2) the clinical visit.
Defines the lag 0, lag 1 and lag 2 exposure windows that the three exposure conditions per pollutant correspond to.
The average concentration of PM10,N O2 and O3 on the day of the clinical visit (lag 0) were 24.9 ± 15.5 μg/m3, 23.1 ± 9.6 μg/m3 and 44.3 ± 19.0 μg/m3, respectively
States the same-day (lag 0) mean and SD concentrations recorded on the three lag-0 exposure conditions: PM10 24.9 +/- 15.5, NO2 23.1 +/- 9.6 and O3 44.3 +/- 19.0 ug/m3.
One day before (lag 1) 24.3 ± 15.6 5 14 29 129 15
Table 2 PM10 row: the lag-1 mean and SD (24.3 +/- 15.6 ug/m3) recorded on the PM10 one-day-before exposure condition.
Two days before (lag 2) 23.9 ± 15.4 5 14 27 129 13
Table 2 PM10 row: the lag-2 mean and SD (23.9 +/- 15.4 ug/m3) recorded on the PM10 two-days-before exposure condition.
One day before (lag 1) 21.2 ± 10.0 3 14 27 65 13
Table 2 NO2 row: the lag-1 mean and SD (21.2 +/- 10.0 ug/m3) recorded on the NO2 one-day-before exposure condition.
Two days before (lag 2) 20.0 ± 10.1 3 12 25 65 13
Table 2 NO2 row: the lag-2 mean and SD (20.0 +/- 10.1 ug/m3) recorded on the NO2 two-days-before exposure condition.
One day before (lag 1) 47.0 ± 18.7 3 34 61 104 27
Table 2 ozone row: the lag-1 mean and SD (47.0 +/- 18.7 ug/m3) recorded on the ozone one-day-before exposure condition.
Two days before (lag 2) 46.6 ± 19.4 1 33 60 106 27
Table 2 ozone row: the lag-2 mean and SD (46.6 +/- 19.4 ug/m3) recorded on the ozone two-days-before exposure condition.
Analyses for ozone were restricted to the warm season (May to September) when impacts were considered most likely.
Justifies restricting the ozone exposure conditions to the warm season (May-September), as their names record.
Average annual concentrations of PM10 and NO2 were well below the European limit values of 40 μg/m3 during the entire study period
Supports the description's statement that the measured concentrations were below the European limit values, which is what makes the observed associations notable.
The mean (SD) PM10 concentration on the day of the lung function test was 24.9 (15.5) μg/m3. A 10 μg PM10/m3 increase on the day of the clinical examination was associated with a 18.9 ml lower FVC (95% CI: -27.5 to -10.3, p < 0.0001), 12.8 ml lower FEV1 (-19.1 to -6.5; p < 0.0001), and a 51.4 ml/s lower PEF (-75.0 to -27.0; p < 0.0001). The FEV1/FVC-ratio showed no associations. An increase of 10 μgNO2/m3 was associated with a reduction in PEF (-66.1 ml/s (-106.6 to -25.6; p < 0.001)) on the day of the examination.
States the headline lag-0 estimates recorded as response comparisons: -18.9 ml FVC, -12.8 ml FEV1 and -51.4 ml/s PEF per 10 ug PM10/m3, no FEV1/FVC association, and -66.1 ml/s PEF per 10 ug NO2/m3.
FVC and FEV1/FVC were not significantly associated with NO 2 exposure. None of the lung function outcomes were significantly associated with O 3.
States the null findings recorded as no-change comparisons: FVC and FEV1/FVC were not associated with NO2, and no lung function outcome was associated with ozone.
We found negative associations between daily variations in ambient air pollution and FVC, FEV1 and PEF in healthy adults.
States the authors' overall conclusion, which is what links this assay to the decreased-lung-function adverse outcome.
the reported associations are driven by within-subject effects
Supports the description's note that the associations are driven by within-subject variation, which is what the repeated-measures design buys.
we have not looked into possible physiological pathways for the effects observed in this study.
Supports the file's omission of any key event relationship: the authors state they did not investigate physiological pathways, so no mechanistic KE-to-KE link is recorded.
FVC, ml −18.9 (−27.5 to − 10.3) <0.0001 −9.8 (−24.1 to 4.6) 0.18 −7.0 (−16.6 to 15.2) 0.93
Table 3 FVC row for lag 0: the source of the three same-day FVC comparisons (PM10 -18.9, NO2 -9.8, O3 -7.0 ml per 10 ug/m3) with their 95% CIs and p-values.
FEV1, ml −12.8 (−19.1 to − 6.5) <0.0001 −13.8 (−24.2 to − 3.5) 0.009 4.1 (−8.0 to 16.2) 0.51
Table 3 FEV1 row for lag 0: the source of the three same-day FEV1 comparisons (PM10 -12.8, NO2 -13.8, O3 +4.1 ml per 10 ug/m3).
FEV1/FVC, % 0.072 (0.02 to 0.17) 0.14 −0.14 (−0.31 to 0.27) 0.10 0.09 ( −0.07 to 0.26) 0.26
Table 3 FEV1/FVC row for lag 0: the source of the three same-day FEV1/FVC comparisons (PM10 0.072, NO2 -0.14, O3 0.09 percent per 10 ug/m3).
PEF, ml/s −51.4 (−75.0 to − 27.7) <0.0001 −66.1 (−106.6 to − 25.6) 0.001 41.0 (−1.0 to 83.0) 0.06
Table 3 PEF row for lag 0: the source of the three same-day PEF comparisons (PM10 -51.4, NO2 -66.1, O3 +41.0 ml/s per 10 ug/m3), including the -27.7 CI bound that the abstract prints as -27.0.
FVC, ml −23.0 (−31.4 to − 14.6) <0.0001 −5.6 (−19.4 to 8.3) 0.43 −14.2 (−29.1 to 0.67) 0.06
Table 3 FVC row for lag 1: the source of the three one-day-before FVC comparisons (PM10 -23.0, NO2 -5.6, O3 -14.2 ml per 10 ug/m3).
FEV1, ml −15.8 (−21.9 to − 9.6) <0.0001 −8.8 (−18.7 to 1.2) 0.08 −9.0 (−20.2 to 2.2) 0.12
Table 3 FEV1 row for lag 1: the source of the three one-day-before FEV1 comparisons (PM10 -15.8, NO2 -8.8, O3 -9.0 ml per 10 ug/m3).
FEV1/FVC, % 0.089 ( −0.00439 to 0.18) 0.06 −0.07 (−0.23 to 0.094) 0.42 0.073 ( −0.08 to 0.23) 0.36
Table 3 FEV1/FVC row for lag 1: the source of the three one-day-before FEV1/FVC comparisons (PM10 0.089, NO2 -0.07, O3 0.073 percent per 10 ug/m3).
PEF, ml/s −56.6 (−79.9 to − 33.4) <0.0001 −66.0 (−104.4 to − 27.6) 0.0008 −9.9 (−48.9 to 29.0) 0.62
Table 3 PEF row for lag 1: the source of the three one-day-before PEF comparisons (PM10 -56.6, NO2 -66.0, O3 -9.9 ml/s per 10 ug/m3).
FVC, ml −17.8 (−26.6 to − 9.0) <0.0001 −3.7 (−16.9 to 9.4) 0.58 −10.9 (−24.8 to 3.0) 0.12
Table 3 FVC row for lag 2: the source of the three two-days-before FVC comparisons (PM10 -17.8, NO2 -3.7, O3 -10.9 ml per 10 ug/m3).
FEV1, ml −10.7 (−17.1 to − 4.3) 0.001 −9.7 (−19.0 to −4.0) 0.04 −5.2 (−15.6 to 5.3) 0.33
Table 3 FEV1 row for lag 2: the source of the three two-days-before FEV1 comparisons (PM10 -10.7, NO2 -9.7, O3 -5.2 ml per 10 ug/m3).
FEV1/FVC, % 0.12 (0.02 to 0.22) 0.016 −0.10 (−0.26 to 0.05) 0.19 0.10 ( −0.05 to 0.25) 0.20
Table 3 FEV1/FVC row for lag 2: the source of the three two-days-before FEV1/FVC comparisons (PM10 0.12, NO2 -0.10, O3 0.10 percent per 10 ug/m3).
PEF, ml/s −36.3 (−60.5 to − 12.1) 0.003 −99.0 (−135.8 to − 22.1) <0.0001 −10.9 (−46.9 to 25.1) 0.55
Table 3 PEF row for lag 2: the source of the three two-days-before PEF comparisons (PM10 -36.3, NO2 -99.0, O3 -10.9 ml/s per 10 ug/m3).
P-values < 0.01 are considered significant and are indicated in bold
States the significance threshold the paper applies to Table 3 (p < 0.01), which every response comparison's description is judged against rather than the conventional 0.05.
The age of the 2449 participants ranged between 16 and 70 years, with an average of 37 (Inter Quartile Range (IQR) = 18) years when entering the study.
States the cohort's age range (16-70 years) and average age at entry recorded on this subject.
A total of 1739 (71%) participants were non-smokers, 342 (14%) were current smokers, and 368 (15%) reported to be former smokers.
States the smoking distribution recorded in subject_characteristics: 1739 (71%) non-smokers, 342 (14%) current and 368 (15%) former smokers.
The study included 1756 (72%) male participants. The majority (94%) was Caucasian.
States the proportion of male participants (1756, 72%) and the 94% Caucasian majority recorded in subject_characteristics.
The results for this apparently healthy group of adults are novel because few panel studies have looked at healthy adults.
Supports describing this cohort as apparently healthy adults, the population whose novelty the authors claim and which places this extraction in the healthy-adult rather than the vulnerable-subgroup arm of the AOP.
Similar FEV1 reductions were seen at lag 1 and lag 2.
Records what the authors themselves claim about NO2 and FEV1 at lags 1 and 2, next to the threshold-based coding of those two comparisons: neither estimate meets the paper's stated p < 0.01 criterion, so both are recorded as no_change, but the authors read them as reductions similar to the significant lag-0 one.
Response comparisons
FVC change per 10 ug/m3 PM10 at lag 0
Estimated change in FVC per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. The paper reports only the threshold p < 0.0001, not an exact p-value, so p_value is omitted rather than invented. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 PM10 at lag 0
Estimated change in FEV1 per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. The paper reports only the threshold p < 0.0001, not an exact p-value, so p_value is omitted rather than invented. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 PM10 at lag 0
Estimated change in FEV1/FVC per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (0.072). Note that the printed 95% CI (0.02 to 0.17) excludes zero although the printed p-value is 0.14; recorded as printed. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 PM10 at lag 0
Estimated change in PEF per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. The paper reports only the threshold p < 0.0001, not an exact p-value, so p_value is omitted rather than invented. Note a discrepancy in the source: Table 3 prints the 95% CI upper bound as -27.7 while the abstract prints -27.0 for the same estimate; the Table 3 value is recorded here. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FVC change per 10 ug/m3 NO2 at lag 0
Estimated change in FVC per 10 ug/m3 increase in nitrogen dioxide on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-9.8). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 NO2 at lag 0
Estimated change in FEV1 per 10 ug/m3 increase in nitrogen dioxide on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 NO2 at lag 0
Estimated change in FEV1/FVC per 10 ug/m3 increase in nitrogen dioxide on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-0.14). Note that the printed 95% CI (-0.31 to 0.27) is not centred on the -0.14 point estimate; recorded as printed. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 NO2 at lag 0
Estimated change in PEF per 10 ug/m3 increase in nitrogen dioxide on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FVC change per 10 ug/m3 O3 at lag 0
Estimated change in FVC per 10 ug/m3 increase in ozone (warm season, May-September) on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-7.0). Note an internal inconsistency in the source: the printed 95% CI (-16.6 to 15.2) is not centred on the -7.0 point estimate; the values are recorded exactly as printed, not corrected. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 O3 at lag 0
Estimated change in FEV1 per 10 ug/m3 increase in ozone (warm season, May-September) on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (4.1). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 O3 at lag 0
Estimated change in FEV1/FVC per 10 ug/m3 increase in ozone (warm season, May-September) on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (0.09). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 O3 at lag 0
Estimated change in PEF per 10 ug/m3 increase in ozone (warm season, May-September) on the day of the lung function test (lag 0). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (41.0). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FVC change per 10 ug/m3 PM10 at lag 1
Estimated change in FVC per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. The paper reports only the threshold p < 0.0001, not an exact p-value, so p_value is omitted rather than invented. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 PM10 at lag 1
Estimated change in FEV1 per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. The paper reports only the threshold p < 0.0001, not an exact p-value, so p_value is omitted rather than invented. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 PM10 at lag 1
Estimated change in FEV1/FVC per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (0.089). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 PM10 at lag 1
Estimated change in PEF per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. The paper reports only the threshold p < 0.0001, not an exact p-value, so p_value is omitted rather than invented. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FVC change per 10 ug/m3 NO2 at lag 1
Estimated change in FVC per 10 ug/m3 increase in nitrogen dioxide one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-5.6). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 NO2 at lag 1
Estimated change in FEV1 per 10 ug/m3 increase in nitrogen dioxide one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-8.8). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. The authors themselves describe this as an FEV1 reduction similar to the significant lag-0 one ("Similar FEV1 reductions were seen at lag 1 and lag 2."), so the no_change coding reflects the paper's stated p < 0.01 criterion rather than the authors' own reading of the result.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 NO2 at lag 1
Estimated change in FEV1/FVC per 10 ug/m3 increase in nitrogen dioxide one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-0.07). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 NO2 at lag 1
Estimated change in PEF per 10 ug/m3 increase in nitrogen dioxide one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FVC change per 10 ug/m3 O3 at lag 1
Estimated change in FVC per 10 ug/m3 increase in ozone (warm season, May-September) one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-14.2). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 O3 at lag 1
Estimated change in FEV1 per 10 ug/m3 increase in ozone (warm season, May-September) one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-9.0). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 O3 at lag 1
Estimated change in FEV1/FVC per 10 ug/m3 increase in ozone (warm season, May-September) one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (0.073). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 O3 at lag 1
Estimated change in PEF per 10 ug/m3 increase in ozone (warm season, May-September) one day before the lung function test (lag 1). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-9.9). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FVC change per 10 ug/m3 PM10 at lag 2
Estimated change in FVC per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. The paper reports only the threshold p < 0.0001, not an exact p-value, so p_value is omitted rather than invented. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 PM10 at lag 2
Estimated change in FEV1 per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 PM10 at lag 2
Estimated change in FEV1/FVC per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (0.12). Note that p = 0.016 is below 0.05 but above the paper's own 0.01 threshold, and that the point estimate is an INCREASE in the ratio, not a decrease. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 PM10 at lag 2
Estimated change in PEF per 10 ug/m3 increase in particulate matter with an aerodynamic diameter < 10 um two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FVC change per 10 ug/m3 NO2 at lag 2
Estimated change in FVC per 10 ug/m3 increase in nitrogen dioxide two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-3.7). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 NO2 at lag 2
Estimated change in FEV1 per 10 ug/m3 increase in nitrogen dioxide two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-9.7). Note that the printed 95% CI (-19.0 to -4.0) is not centred on the -9.7 estimate; recorded as printed. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. The authors themselves describe this as an FEV1 reduction similar to the significant lag-0 one ("Similar FEV1 reductions were seen at lag 1 and lag 2."), so the no_change coding reflects the paper's stated p < 0.01 criterion rather than the authors' own reading of the result.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 NO2 at lag 2
Estimated change in FEV1/FVC per 10 ug/m3 increase in nitrogen dioxide two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-0.10). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 NO2 at lag 2
Estimated change in PEF per 10 ug/m3 increase in nitrogen dioxide two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is significant by the paper's own criterion. The paper reports only the threshold p < 0.0001, not an exact p-value, so p_value is omitted rather than invented. Note that the Results text rounds this estimate to '-99 ml/s'. The printed 95% CI (-135.8 to -22.1) is not centred on the -99.0 estimate; recorded as printed. This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FVC change per 10 ug/m3 O3 at lag 2
Estimated change in FVC per 10 ug/m3 increase in ozone (warm season, May-September) two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-10.9). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1 change per 10 ug/m3 O3 at lag 2
Estimated change in FEV1 per 10 ug/m3 increase in ozone (warm season, May-September) two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-5.2). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
FEV1/FVC change per 10 ug/m3 O3 at lag 2
Estimated change in FEV1/FVC per 10 ug/m3 increase in ozone (warm season, May-September) two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (0.10). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
PEF change per 10 ug/m3 O3 at lag 2
Estimated change in PEF per 10 ug/m3 increase in ozone (warm season, May-September) two days before the lung function test (lag 2). Linear mixed model with a random intercept per participant (REML, unstructured autocorrelation), adjusted for sex, age, age squared, body mass index, smoking status, quality grade of the lung function test, apparent temperature, hour of the day, day of the week, month and year of the clinical visit. The paper treats p < 0.01 as significant, so this estimate is NOT significant by the paper's own criterion, which is why change_direction is no_change even though the point estimate has a sign (-10.9). This is a regression slope across a continuously exposed cohort, not a contrast between a control group and a treated group: the study has no unexposed arm, so control_output and treated_output are deliberately omitted. sample_size is omitted deliberately: the paper restricts the ozone analyses to the warm season (May to September), so this estimate comes from a subset of the 5404 clinical visits, and the paper never reports that subset's n.
Statistical test: linear mixed model (SAS MIXED procedure), random intercept per participant, REML with unstructured autocorrelation
Protocols
Vitalograph Pneumotrac spirometry in an occupational surveillance programme
Lung function measured with a Vitalograph Pneumotrac (Ennis, U.K.) in a standing posture, otherwise following the most recent European Respiratory Society and American Thoracic Society guidelines for lung function testing in occupational settings. The same device was used throughout the study, and participants were supervised and coached by one of three trained nurses. Examinations were performed on weekdays in the morning between 8 am and 12 am at the SCK-CEN medical centre. FVC, FEV1, FEV1/FVC ratio and PEF were obtained with the Vitalograph Pneumotrac software; the manoeuvre with the largest FEV1 + FVC sum was recorded and used for all other parameters. Three respiratory manoeuvres (maximum exhalation) were performed per visit in 2011, 2012 and 2013; because of the high reproducibility of the measurements, most participants performed only two manoeuvres in 2014 and 2015. The authors flag this mid-study change as a deviation from ERS/ATS guidelines and a limitation. No replicate_requirements value is recorded because the schema's slot is a single integer and the requirement changed during the study.
Equipment: Vitalograph Pneumotrac spirometer (Ennis, U.K.)
Daily ambient PM10, NO2 and ozone from a fixed regulatory monitoring station
Validated daily concentrations of PM10, NO2 and O3 obtained from the official air quality monitoring station of the Flanders Environment Agency (VMM) in Dessel (station number 42N016), a rural background station 6 km east of the SCK-CEN medical centre, with no major local pollution sources near either site. Exposure was calculated as the average concentration on the day of each clinical visit (lag 0), the day before (lag 1) or two days before (lag 2). The 24-h mean outdoor temperature and humidity came from the meteorological station Antwerpen-Luchtbal (T2M802, about 50 km from the air quality station) and were used to compute apparent temperature with the August-Roche-Magnus approximation. The authors note that NO2 exhibits large spatial variation, so exposure misclassification is likely greater for NO2 than for PM10, biasing the NO2 effect estimates towards the null.
Equipment: Thermo Andersen ESM FH 62 I-R (PM10); ThermoFisher Model 42i (NO2); Teledyne API Model T400 (ozone)