A simple protocol for tuberculin skin test reading certification Um protocolo simples para a certificação da leitura da prova tuberculínica Un protocolo simple para la certificación de la lectura de la prueba cutánea de tuberculina

Although tuberculosis preventive therapy is one of the cornerstones for eliminating the disease, many barriers exist in the cascade of care for latent tuberculosis infection, including the need to certify healthcare professionals for reading tuberculin skin tests (TST). This paper proposes and evaluates a simple protocol for TST reading training. Primary care workers from different backgrounds received a 2-hour theoretical course, followed by a practical course on bleb reading. Blebs were obtained by injecting saline into sausages and then in volunteers. A certified trainer then evaluated the effectiveness of this protocol by analyzing the trainees’ ability to read TST induration in clinical routine, blinded to each other’s readings. Interobserver agreement was analyzed using the Bland-Altman test. The trainees’ reading accuracy was calculated using two cut-off points – 5 and 10mm – and the effect of the number of readings was analyzed using a linear mixed model. Eleven healthcare workers read 53 saline blebs and 88 TST indurations, with high agreement for TST reading (0.07mm average bias). Sensitivity was 100% (94.6; 100.0) at 5mm cut-off and 87.3% (75.5; 94.7) at 10mm cut-off. The regression model found no effect of the number of readings [coefficient: -0.007 (-0.055; 0.040)]. A simple training protocol for reading TST with saline blebs simulations in sausages and volunteers was sufficient to achieve accurate TST induration readings, with no effect observed for the number of readings. Training with saline blebs injected into voluntary individuals is safer and easier than the traditional method. Training Courses; Observer Variation; Tuberculin Test Correspondence A. Trajman Rua Macedo Sobrinho 74, apto. 203, Rio de Janeiro, RJ 22271-080, Brasil. atrajman@gmail.com 1 Mestrado Profissional em Atenção Primária à Saúde, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brasil. 2 McGill University, Montréal, Canada. 3 Instituto de Medicina Social, Universidade do Estado do Rio de Janeiro, Rio de Janeiro, Brasil. 4 Instituto Nacional de Cardiologia, Rio de Janeiro, Brasil. 5 Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brasil. doi: 10.1590/0102-311X00027321 Cad. Saúde Pública 2021; 37(8):e00027321 QUESTÕES METODOLÓGICAS METHODOLOGICAL ISSUES This article is published in Open Access under the Creative Commons Attribution license, which allows use, distribution, and reproduction in any medium, without restrictions, as long as the original work is correctly cited.


Introduction
Tuberculosis (TB) remains the first cause of death from infectious disease worldwide 1 . Tuberculosis preventive therapy (TPT) is highly effective in individuals with latent tuberculosis infection (LTBI) 2 , being the cornerstone for eliminating TB by 2050 3,4 . The many barriers at different steps of the cascade of care for LTBI, however, result in less than 30% of those eligible for TPT receiving a prescription and less than 19% completing treatment 5 . The effect of TPT on public health remains thus minimal 1 , despite the inclusion of new shorter and safer TPT regimens by the World Health Organization (WHO) guidelines 2 .
The risk of progression from LTBI to active disease is higher in those with a positive LTBI test, either the tuberculin skin test (TST) or interferon-gamma release assays (IGRA) 6 . Not surprisingly, TPT has been shown to be more beneficial for those with a positive LTBI test 7,8,9 . Thus, TST (or IGRA, where available) is a mandatory test in the LTBI cascade of care, except for patients living with HIV (PLHIV) or for patients younger than five years old, who have a higher risk of progressing to active disease 2 . However, of the 30 million HIV-negative individuals five years of age or older that should receive TPT by 2022 4 , 20 million require a positive LTBI test. TST is simple to perform, dispense with laboratory infrastructure and has good accuracy 10 . Newer and more specific TST using recombinant tuberculin based on ESAT-6 and CP-10 proteins, have already been adopted in some BRICS countries 11,12,13 .
Despite being a simple test, reproducibility of induration readings has been reported to be low 14,15,16,17 . Thus, in some health systems, certification of healthcare workers for TST reading is needed, but specific requirements for certification vary widely 15 . While ten TST applications and readings under the supervision of a certified professional are required for certification in Georgia, United States (https://dph.georgia.gov/sites/dph.georgia.gov/files/TB-TSTCertclass.pdf, accessed on 11/ Sep/2020), Brazil has established 80 to 100 readings with 80% agreement with the trainer as a requirement 18 . Although protocols to train healthcare professionals for TST application and reading also vary widely, a control quality of TST performance based on a remote m-health method (mTST) has been reported as accurate, especially for readings under 5mm and over 14mm 19 . Based on this experience 19 , we posited that a training protocol based on saline blebs simulation would be sufficient to certify healthcare professionals and tested this hypothesis.

Methods
From October 2019 to Jnauary 2020, we conducted an operational study in a primary care health unit in a high TB incidence favela in Rio de Janeiro, Brazil, setting chosen as to reduce the risk of multiple patients without any induration jeopardizing the evaluation step. Any healthcare professional or healthcare student was eligible to voluntarily participate in the training. We included all individuals who were available and agreed to participate by signing the informed consent form, and excluded those who had been previously trained. Based on reading variance reported in protocols from the 1960's 15 , we established a sample of 40 trainees.
Trainees attended a 2-hour seminar on the importance of TPT and recommendations on how to identify contacts, apply, read, and interpret TST (adapted from Dick Menzies & Linette McElroy; Supplementary Material, http://cadernos.ensp.fiocruz.br/static//arquivo/suppl1-e00027321_5854. pdf), and watched a 7-minute demonstration video from the Brazilian Ministry of Health (available at https://www.youtube.com/watch?v=fRbg7vsa44o&t=4s). Following this theoretical introduction, the trainees participated in a reading training, where temporary blebs were produced by injecting 0.1mL of saline in sausages (simulating the intradermal technique; Supplementary Figure S1, http://cadernos.ensp.fiocruz.br/static//arquivo/suppl2-e00027321_9122.pdf). Finally, to simulate the Mantoux technique, we intradermally injected 0.1mL of saline in voluntary healthcare worker trainees, who signed a separate informed consent form. All blebs should have a minimum 6mm diameter and last 10 to 15 minutes, like the tuberculin bleb 20 . Readings of both the sausages and the arm blebs were individual, each student blinded to the other readings. All training steps were provided by an experienced LTBI physician (A.T.) with no TST certification or experience with TST reading, who Cad. Saúde Pública 2021; 37(8):e00027321 underwent training with sausages and voluntary saline blebs in the previous year as preparation for two research projects 19,21 . The practical course lasted less than two hours, depending on the number of trainees attending. A schematic of the training process can be found in the Supplementary Figure  S2 (http://cadernos.ensp.fiocruz.br/static//arquivo/suppl2-e00027321_9122.pdf).
To evaluate the effectiveness of this training, trainees underwent an evaluation step by reading TST indurations in service, which were compared to the reference standard of "true" induration readings by a trainer certified by the Brazilian Ministry of Health (90% concordance in 80-100 readings with a certified trainer). Both trainees and trainer were blinded to each other's readings, and all readings (saline simulation and tuberculin induration) were performed using a caliper ruler (Supplementary Figure S3, http://cadernos.ensp.fiocruz.br/static//arquivo/suppl2-e00027321_9122.pdf).
Interobserver agreement was set at a difference up to 2mm between readings and analyzed using the Bland-Altman test 22 . The trainee's reading sensitivity and specificity and their 95% confidence intervals were calculated using 5mm and 10mm as cut-off points (measures used in Brazil, depending on the risk of progression to active TB and time since BCG vaccination) 23 . To verify if the number of readings or the type of product injected (saline vs. tuberculin) had any significant effect on the agreement, we developed a linear mixed model with two hierarchical levels (per patient and per reader), including all readings (blebs and indurations), having as dependent variable the absolute difference between the trainees' and trainer's results in mm and as independent variables: the trainer's readings, the number of readings and the type of product injected (saline or tuberculin). The model was considered robust if the mean residuals were close to zero. All analyses were performed in Stata (https:// www.stata.com).
The study was approved by the Ethics Research Committee of the Anna Nery Nursing School/São Francisco de Assis Health Care Institute, Federal University of Rio de Janeiro (n. 3.256.347 on April 10, 2019) and by the Rio de Janeiro Health Secretariat (n. 3.470.316 on July 25, 2019). All patients and healthcare workers signed an informed consent form before participating in the study.

Results
Of the 64 eligible healthcare workers, 15 were available for training in three different sessions. Each trainee read an average of 6.8 (± 2.2) blebs in sausages and 3.5 (± 1.3) blebs in volunteers. Of the 15 participants, 11 were evaluated in service with an average of 14.7 (± 5.4) TST readings each. These included one pharmacist, two physicians, two registered nurses, two nursing assistants, and four community health agents with nursing assistant training. Overall, they performed 141 readings -53 blebs and 88 indurations -before the project's interruption due to the SARS-CoV-2 pandemic.
As for the mixed regression model coefficients (Table 1), the number of readings had no significant effect on the agreement. Its coefficient was -0.0007, meaning that 131 readings (estimated by 1/0.0007) would be necessary to result in a 1mm difference between the trainees' and trainer's readings. Tuberculin resulted in a 0.9mm (estimated by 1/1.065) larger difference than saline. Finally, the larger the trainer's reading, the greater the difference with the trainee's (result consistent with the lower sensitivity for the 10mm cut-off). Since the mean residuals of the model were close to zero, it was considered robust.

Figure 1
Bland-Altman agreement of tuberculin skin tests reading between trainees and trainer (N = 88, using tuberculin).
Note: the Y-axis shows the difference between trainees' and trainers' tuberculin induration readings and the X-axis, the mean value of these readings.
Mean error was 0.07mm (-0.51; 0.66). Bland & Altman 22 recommended that 95% of the data points should be within ± 2s of the mean difference for measurements to be considered concordant. In this graphic, 94.3% of the points were within ± 2s.

Discussion
The simple protocol for TST reading training with saline simulations in an average of 3.5 volunteers presented here was sufficient to obtain a high agreement for TST induration readings, and a very high sensitivity for classifying patients into positive or negative results according to two different cut-off points (5mm and 10mm). The number of readings had no effect on the agreement. Overall, these findings suggest that training with very few saline blebs may be effective in certifying healthcare workers to read TST in service. This proposed protocol simplifies the logistics of training, is more ethically acceptable, and dispense with the need for highly experienced trainers. Another of its advantages was training TST application, as incorrect administration of the technique was cited as a potential cause for false negative TST results 20 . An improvement to this protocol would be to add a step to evaluate the application and verify if the trainer can read the bleb size within 10 to 15 minutes, the usual duration 20 .
The literature also reports a high intra-and interobserver variability in TST reading 24,25,26 , with greater differences in readings, especially when these approach positivity cut-off points 16 . Since a 2mm difference, for example, represents 50% on a 4mm result and only 20% on a 10mm result, more errors are expected with larger readings. Larger readings could also be expected with tuberculin as the caliper ruler squeezes the bleb, which is not observed with induration. A well-conducted study comparing intra-and interobserver readings using a pair of calipers and a transparent ruler, however, showed that the variability between TST induration readings is not influenced by changing the reading instrument 27 .
TST is the most widely LTBI test available and despite its limitations, the test has good accuracy, does not require laboratory facilities, and is simple to perform and interpret 19 ; any healthcare professional skilled in performing intradermal vaccines, as BCG, is able to inject the tuberculin. TST reading, however, has been subject to complex protocols that require a certified trainer and the availability of several voluntary arms, thus hampering its wide availability in public healthcare systems. Our findings have relevant pragmatic implications. Our protocol is also safer and more ethically acceptable. Tuberculin injection, although generally well tolerated, can cause local adverse reactions such as itching, pain, blistering, a necrotic reaction and vagal malaise 28 . Saline injection is safe despite mild temporary local discomfort, usually a burning sensation. An alternative method to avoid training on human arms is to use artificial arm models, but evaluating the effectiveness of such method was beyond the scope of the present study.
Our study has a few limitations. First, generalizability is difficult from the small sample of a single healthcare unit. Second, this small sample size does not allow for stratified analyses by professional background as other studies have done with the ability of paramedics 29 , pharmacists and other healthcare workers 30,31 to read TST. However, the overall performance of our participants was very good, suggesting that being able to read TST does not depend on professional background. Finally, we did not evaluate different reading techniques 32 .
Our study also has strengths, such as the in-service evaluation carried out by a certified trainer and both trainees and trainers being blinded for the readings. Also, different analyses attest to the robustness of the findings.
In summary, the simple and realistic training protocol successfully evaluated in this study may reverse the scenario of unavailability of certified personnel in LTBI cascade of care, situation created by complex certifications 33 . As new and more specific TST are developed 11,12,13 , TST reading will remain a necessity.
Cad. Saúde Pública 2021; 37(8):e00027321 Contributors L. L. Gloria contributed to the data collection and writing, and approved the final version. M. L. Bastos contributed to the study design and writing, and approved the final version. B. Santos Júnior contributed to the statistical analyses, data interpretation and writing, and approved the final version. A. Trajman contributed to the study design, data collection, analysis and interpretation and writing.

Conflicts of interests
The authors declare no conflicts of interest.