Study design
A parallel, cluster-randomized controlled trial will be conducted with mills as the unit of randomization. After a baseline survey, the intervention package will be rolled out in mills allocated to the intervention arm, and its impact will be assessed through three follow-up surveys carried out across all mills. Analyses of effectiveness will be supplemented by a limited economic evaluation of the intervention.
Recruitment and randomization
In Pakistan, no official sampling frame of textile mills is available but several business and textile associations represent the country’s textile mills and we will acquire a list of industrial plants in Karachi from them. To be eligible, textile mills will need to: (1) be from the formal sector and registered with business associations and the Government Labour Department; (2) have either spinning or weaving sections, or both; (3) employ at least 50 eligible workers in their spinning and weaving sections; (4) have a workforce in which at least 75% of staff have been employed for 12 months or longer; (5) have a management willing to participate in the planning and implementation of the study; and (6) have no plans to introduce major improvements over the next two years.
Eligible workers at each mill will be identified by the human resources department. We will include male textile workers aged ≥18 years who are employed in the bale opening, blowing, carding, spinning, twisting, winding, warping, weaving or waste recycling sections. Those who do not give consent will be excluded. We will also exclude workers from the wet-processing areas (dyeing or bleaching), stitching and packaging sections, as well as support or administrative staff. At each mill, personal air sampling will be done on five purposively selected workers (100-150 in all) representing the range of sections and jobs. For this purpose, the latter will be categorized into one of four groups: (1) helpers, cleaners and doffers; (2) machine operators; (3) jobbers and fitters; and (4) masters, in-charge or supervisors.
We will purposively select twenty-eight textile mills and assign them equally to the intervention or control arm by stratified randomization. Stratification will be according to the total number of employees at the mill (≤250 / >250) and occupational health and safety measures in place at baseline (as assessed by a walk-through survey and categorized as good or poor). Within strata, allocation to intervention or control will be through random numbers generated at Imperial College London and shared with the research team in Pakistan. Blinding of mill staff will not be possible but we will ‘blind’ the data collection team including the technicians measuring dust levels, and the laboratory personnel involved in weighing filters for gravimetric analysis.
Intervention
The study’s framework (Fig. 1) describes the links between the components of the intervention. The training component is partly adapted from Robson et al.’s theory of how training can be effective in leading to improved health outcomes (20). This considers worker and manager training as part of the wider workplace environment, where its effectiveness is modulated by other individual and organizational factors. In accordance with that framework, the intervention package will comprise four components:
- Initial training for all workers and managers on occupational health and safety
- Follow-up refresher sessions every three months
- Formation of workplace committees (including worker representatives) to draw up, agree and promote a health and safety plan that includes wet mopping, safe disposal of cotton dust, and the use of simple face masks, as well as further publicity about the risks from cotton dust.
- Provision of adequate supplies of facemasks to support the health and safety plan
An initial series of training sessions will be conducted in the intervention arm to explain the health hazards of cotton dust and appropriate preventive measures. These sessions will be based on ‘adult learning theory’ (18, 23), and will be delivered separately to the workers and managerial staff. To reinforce the content, regular refresher sessions will then be conducted every three months for the workers and every six months for the managers.
At each mill in the intervention arm, an Environmental and Occupational Health (EOH) plan will be developed with the management, covering steps to be taken in implementing the intervention package. As part of the plan, wet mopping and safe disposal of cotton dust (such as keeping it in closed containers and avoiding direct contact of workers with trash) will be promoted. Guidance will be provided to ensure that any available local exhaust ventilation system is functioning satisfactorily. In addition, better personal and workplace hygiene will be promoted among the workforce. Implementation of the EOH plan will be documented as part of the trial’s fidelity assessment.
Smooth delivery of the EOH plan will be assisted by formation of a ‘MultiTex’ committee at each mill in the intervention arm. This committee, which will include workers’ representatives, will be responsible for quality assurance and addressing day to day challenges.
In addition to training sessions, workers will be provided with facemasks for use during working hours and MultiTex committee members will be responsible for keeping a log of the number of facemasks distributed per shift and the number of workers who actually use them.
Measurements and their frequencies
To assess the impact of the intervention, surveys will be carried out across all mills at baseline (i.e. before the intervention is implemented), and then at follow-up 3, 12 and 18 months later (Fig. 2, 3 and 4, and additional file 1). The surveys will include detailed interviews using a questionnaire, and spirometry for each worker. In addition, personal dust exposures, measured with IOM samplers, will be recorded for selected subsamples of participants.
Interviews:
To assess respiratory symptoms we will use a translated version of the Medical Research Council, UK respiratory questionnaire (24). Questions concerning byssinosis will be taken from the WHO respiratory questionnaire from technical report series 684 (25). To assess knowledge, attitudes and practices (KAP) among textile workers and managers we will use a structured questionnaire which was developed by the authors as part of previous research on textile workers. The study questionnaires will be pretested and delivered by trained data collectors in the field. The total time required for the questionnaires will be approximately 20 minutes.
Spirometry and anthropometric measurements:
We will use the ndd (EasyOne) spirometer to measure lung function, following established guidelines (26). Forced expiratory volume in one second (FEV1), forced vital capacity (FVC) and FEV1/FVC ratio will be recorded. Pre- and post-bronchodilator spirometry will be conducted at least one hour after the worker has started a shift (i.e. after he has been exposed to cotton dust that day). We will note the time of the day when spirometry is carried out, as well as the interval since start of the shift, and the number of consecutive days that have been worked leading up to the measurement. Every effort will be made to undertake spirometry at each follow-up on the same working day and during the same shift as at baseline. For a sub-sample of 100 workers, pre- and post-shift spirometry will additionally be performed, to gauge cross-shift changes in lung function. We will record height in cm using a stadiometer, and weight in kg.
Cotton dust exposures:
We will use IOM sampling heads with Casella Apex2 pumps for personal monitoring of dust exposures, measuring the inhalable fraction of particulate matter (PM <100 μm) over an 8-hour shift. Pre- and post-sampling airflow and sampling time will be recorded, as well as temperature and humidity readings. Each individual’s average PM exposure concentration will be calculated from the difference in filter weights and sampled air volume. Individual data will then be combined to derive mean exposures for combinations of job and section at each mill. Where possible, the same workers will be monitored at follow-up as at baseline, but those who cannot be re-contacted will be replaced by others with similar job tasks. Dust sampling will be done at all the textile mills across the intervention and control arms at baseline and will be repeated at 3, 12 and 18 months after intervention. In addition to personal sampling, area sampling will be carried out in the spinning and weaving sections of each mill using a Dylos DC 1700 particle counter to measure the average concentration of dust (particle size >0.5 or >0.25 µm) over an 8-hour shift. The particle counters will be placed at fixed sites in the centres of work stations in the spinning and weaving sections, at approximately 1.5 m height, and away from the ventilation outlets or fans. Side by side gravimetric and Dylos measurements will be made for a small number of sites to provide a calibration factor for the Dylos particle concentration in relation to cotton dust.
At the end of the study, walk-through surveys will provide direct qualitative assessment of any changes in safety provisions during the trial.
During the course of the study, we will also collect information from mills which receive the intervention on costs of its implementation and any direct or indirect savings that may have resulted. We will also document changes in staff turnover, sickness absenteeism and associated costs.
Study outcomes
The primary outcomes of the trial at 18 months, post-intervention, will be: (1) personal dust exposure levels (in mg/m3), and their change from baseline, in the intervention as compared with the control arm; (2) dust levels in different sections of the mill (in mg/m3), and their changes from baseline, in the intervention as compared with the control arm; (3) changes from baseline in the prevalence of respiratory symptoms, in the intervention as compared with the control arm (a composite variable will be used for presence of one or more respiratory symptoms including; cough, sputum, wheeze, shortness of breath, and chest tightness); and (4) changes in FEV1 from baseline to follow-up, in the intervention as compared with the control arm.
Process evaluation
Process evaluation will be performed according to the logic model of the trial (Fig. 5) using the process measures below:
- For the workers’ training component: number of training sessions held; number of participants in each training session.
- For improvement in KAP of workers: improvements in the proportions of workers reporting good knowledge, appropriate attitude and appropriate protective practices in the intervention compared to the control arm.
- For the managers’ workshops: number and proportion of managers trained at each mill; proportion of mill managers using personal protective equipment (PPE) during visits to high exposure areas.
- Changes in KAP of managers: documented through a questionnaire and compared between the intervention and the control arm.
- For administrative measures: number of meetings held by the ‘MultiTex’ committees; plans and policies developed at each mill as a consequence of the study; incorporation of training program in overall OHS plans of the mills.
- For PPE component: number of facemasks distributed on a daily basis and proportion of workers using facemasks at random visits conducted by the MultiTex committee members.
We will use a checklist for evaluating the fidelity of our trial in line with the MRC guidelines for complex interventions (27). This checklist considers aspects of fidelity such as adaptation of the intervention, modifications of intervention and dose (strength) of the intervention.
Statistical analysis:
Data will be entered through Epi Data and analyzed with STATA. Frequency counts will be derived for baseline socio-demographic variables and respiratory symptoms (categorical variables). Measures of central tendency and dispersion will be reported for continuous variables (cotton dust levels and lung function indices). Correlations between the cotton dust levels and lung function (FEV1, FVC and FEV1/FVC ratio) will be assessed through Pearson’s correlation coefficient.
Models based on generalized estimating equations (GEEs) will be developed to determine changes in the cotton dust levels, frequency of respiratory symptoms, FEV1 (lung function indices) and respiratory sickness absenteeism. The changes in these outcome variables from baseline will be compared between the intervention and control groups. The GEE models will take into account the clustering by mills, as well as the correlation between repeated measures. Multicollinearity will be assessed between covariates and will be taken into account in the multivariable models.
We will analyse by intention to treat, and workers will be classified according to their assigned treatment group at the time of randomization. Sensitivity analyses will be performed to explore the potential impact of losses to follow-up.
Statistical power
Based on previous occupational trials, we assumed that the intra-class correlation coefficients for the outcome measures were 0.02.
For cotton dust levels: assuming the level of significance at 5%, power 80% and a 15% or 0.65 mg/m3 difference in mean PM levels in the intervention arm compared to control arm (expected mean dust level in control arm: 0.6 with a standard deviation of 2.5); the sample size required without clustering is 233/arm. The variance inflation factor is [1+(50-1)*0.02)]=1.98; so the sample size is inflated to 461, which rounds up to 10 groups of 50 in each arm – which increases to 10 groups of 61 in each arm after inflating for losses (*1.05*1.05*1.1 = *1.212).
For respiratory symptoms: the trial is powered to detect a difference of 10% in the proportion of workers reporting respiratory symptoms between intervention and control arms. Assuming level of significance at 5%, power 80%, proportion of workers having respiratory symptoms at the baseline as 40%; without clustering we would require 354/arm. With clustering, the new sample size is 701/arm which – in groups of 50 – means 14 groups of 50 in each arm of the study. Inflating these numbers for losses gives 14 groups of 61.
For lung function: assuming level of significance at 5%, power 80%, 5% or 150 ml difference in mean FEV1 (standard deviation 31) in intervention arm compared to control arm (expected mean FEV1 in control arm: 3279 ml, standard deviation 690); the requirement without clustering would be 333 per arm, which inflates to 660 with clustering, which again leads to 14 groups of 50 in each arm and this again inflates to 14 groups of 61 after inflating for losses.
Ethical considerations:
The study has received conditional ethical approval from the Ethics Review Committee at Aga Khan University, Pakistan (5290-CHS-ERC-18) and the Research Ethics Committee at Imperial College London, UK (19IC4968). Written informed consent will be obtained from each participant, and they will be free to leave the study whenever they desire. The workers’ workshops will be brief and conducted in small groups so that mill productivity is not affected. A small financial compensation will be provided to each participant at each time of data collection. Workers will be provided with light refreshments after spirometry and during workshops.
All the participants will be informed verbally about their spirometry results and guided accordingly. If a study participant wishes to receive the spirometry report this will be provided. Workers who are found to have significant, previously unrecognized lung disease will be counseled, and if they wish, they will be referred to a suitable government hospital for further evaluation and management. A letter of referral will be provided.
All data will be kept confidential and the questionnaire reports and spirometry results of individual workers will not be shared with their employers. However, employers will be provided with anonymized summary results for their mills at the end of the study to help them plan appropriate control and preventive measures.
The study will help the textile workers improve their knowledge about the health hazards of cotton dust exposure and guide researchers further towards developing strategies for the implementation of protective measures. In addition, the textile workers will be tested for their respiratory function through spirometry, which will be free of cost. Those in the intervention arm will also be provided with free disposable face masks on a daily basis during the study period.
Incentives for the workplaces will include free monitoring of dust levels, employee health assessments and training. Such measures will benefit their audit and compliance processes, and in this way they may be able to get better business by fulfilling the requirements of major textile buyers globally (28). After the conclusion of this study, by agreement before the trial, expert advice on OHS improvements will also be provided to the mills in the control arm.
Dissemination
In Pakistan, dissemination seminars will be organized to share findings with various stakeholders including representatives from textile mill associations, concerned NGOs, national and provincial Environmental Protection Agencies, Labour Department, Ministry of Textile and Ministry of Health. Textile mill and business associations are key stakeholders for this project. They will help in getting access to the industry and will also facilitate the implementation of policy recommendations that emerge from the research.