Title:Treatment of neonatal jaundice with filtered sunlight in Nigerian neonates: study protocol of a non-inferior, randomized controlled trial

Transcription

1 Author's response to reviews Title:Treatment of neonatal jaundice with filtered sunlight in Nigerian neonates: study protocol of a non-inferior, randomized controlled trial Authors: Tina M. Slusher (tslusher@umn.edu) Bolajoko O. Olusanya (boolusanya@aol.com) Hendrik J. Vreman (henk.vreman@stanford.edu) Ronald J. Wong (rjwong@stanford.edu) Ann M. Brearley (brea0022@umn.edu) Yvonne E. Vaucher (yvaucher@ucsd.edu) David K. Stevenson (dstevenson@stanford.edu) Version:3Date:20 November 2013 Author's response to reviews: see over

2 BOLAJOKO O. OLUSANYA Director Centre for Healthy Start Initiative 286A, Corporation Drive Dolphin Estate, Ikoyi Lagos, Nigeria Tel: th November 2013 Editor-in-Chief Trials c/o BioMed Central 236 Gray's Inn Road London WC1X 8HB, UK Dear Sir Submission of Manuscript titled: Treatment of neonatal jaundice with filtered sunlight in Nigerian neonates: study protocol of a randomized controlled trial Thank you for your interest in our subject-titled paper and for the valuable comments received from the reviewer(s) which have assisted us in clarifying the manuscript further. Attached is our point-by-point response to the issues raised during the review process. Changes made to the manuscript based on the reviewer s comments are underlined and marked in red. We look forward to your final decision in due course. With kind regards Bolajoko O. Olusanya FRCPCH, PhD

3 Response to Reviewer's report Title: Treatment of neonatal jaundice with filtered sunlight in Nigerian neonates: study protocol of a non-inferior, randomized controlled trial Major compulsory revisions 1. Throughout the study it is unclear as to what is the exact primary outcome. It is also not clear whether it is a binary or continuous variable. We have revised the primary outcome section of the manuscript for better clarity. The primary outcome for the randomized controlled trial is efficacy to be presented as a binary variable. The total serum bilirubin (TSB) level will be measured immediately before beginning therapy on a given day and immediately after therapy is completed that day, and the change in bilirubin level in mg/dl per hour of therapy will be calculated. For infants who are less than 72 hours old in the morning, the treatment will be judged efficacious if the bilirubin level rises more slowly than 0.2 mg/dl per hour of therapy that day. For infants who are more than 72 hours old, the treatment will be judged efficacious if the bilirubin level falls. Efficacy will be evaluated only for days in which the treatment was deemed safe. [see pages 8/9 under Methods]. 2. The authors state that they will record TSB in the morning and at the end of day time phototherapy. They have not mentioned what they are going to do with these values. The change in TSB level on a given day is used to determine whether that treatment day is judged efficacious, as described in Question 1 above. 3. Does the difference (in a negative direction) have to be above a certain cut-off to call it efficacious? Is the absolute difference going to be calculated and averaged over the number of days of phototherapy? Average suggests that the outcome will be expressed as a continuous variable. Does that mean that the mean of the averages in the FS-PT group is expected to be no worse than 10% of the mean in the CPT group? If the change in TSB is negative (i.e. it falls) then that treatment day is deemed efficacious as addressed in Question 1 above. The change in TSB will not be averaged. Instead, we will characterize the average efficacy of each type of phototherapy, defined as the proportion of safe days where a given therapy was used in which the therapy was efficacious. We will compare the average efficacy of SPT to that of CPT, with the goal of demonstrating that SPT is not appreciably worse than CPT (non-inferiority study), with a non-inferiority margin of 10%. We have revised the primary outcome section of the manuscript to clarify this. The average efficacy of each type of phototherapy, defined as the proportion of safe days where a given therapy was used in which the therapy was

4 efficacious, will be compared. FS-PT will be considered at least as efficacious as CPT if the average efficacy on days where FS-PT was used is no worse than 10% less than the average efficacy on days where CPT was used. 4. Under analysis, it is mentioned that the primary outcome will be analysed by a 95% exact binomial confidence interval for the true difference in efficacy. This suggests that efficacy is a binary outcome. The calculated sample size also suggests that it is a binary outcome. But efficacy has not been defined anywhere in the study. Efficacy is a binary outcome, as addressed in Question 1 above. 5. The authors must state what is the assumed delta value for exchange transfusion to call FS-PT non-inferior? The primary goal of this study is to demonstrate that the efficacy of SPT is no worse than that of CPT, with a non-inferiority margin, delta, of 10%. A secondary goal of the study is to demonstrate that the proportion of infants who require exchange transfusion under FS- SPT is no worse than that under CPT with a non-inferiority margin, delta, of 5%. This clarification has been reflected in the secondary outcome section of the revised manuscript [see pages 6/7: Aims & Hypotheses under Methods]. 6. The authors must state why have they have not chosen 90% power. In a non-inferiority trial, it is important to minimise the beta error, i.e. the probability of missing out on a true difference by chance because this is almost as bad as an Alpha error (unlike the paradigm of a superiority trial). By convention, many researchers opt for a 90% power in a non-inferiority trial or equivalence study. The level of 80% power was chosen because it is an established standard of practice for clinical trial design. We are aware that many statisticians are advocating for changing the standard practice to 90% power, and we appreciate the reviewer s suggestion that this is particularly important in non-inferiority trials. Preliminary results from our trial suggest that the assumptions we used in the sample size estimation were quite conservative: the safety of both SPT and CPT appears to be above 99%, rather than the assumed 90%, the typical course of treatment appears to average about 1.3 days instead of the assumed 1.0 days, and the rate of loss due to parent delisting or missing data is much less than the assumed 10%. We therefore feel it is unlikely that our study is underpowered. 7. Under sample size calculation, the sentence assuming that the safety of both FS-PT and CPT is 90%, a sample size of 560 total days is required is not clear. The authors should explain why

5 safety has entered into the sample size calculation. Has a separate sample size calculation been done to test for sale safety, and have the authors taken the higher of the two sample sizes? If so, the assumptions for the safety part have not been stated. The sample size calculation gives the required number of treatment days that are evaluable for efficacy. Since we will evaluate efficacy only for days in which the treatment was deemed safe, we increased the sample size to account for the possibility that the proportion of safe days might be as low as 90%. The sample size section of the manuscript has been revised to clarify this point [see page 9, paragraph 3]. We did not size this study to assess safety, since safety is not a primary outcome of this study. The safety of SPT has already been demonstrated in the prior observational study. 8. One of the inclusion criteria is if parents choose to come for screening up to 14 days of life. For such cases, the prior TSB will not be known. Therefore, there would be no way of knowing whether the TSB is already in a downward trend or whether it is on the upslope of the TSB curve. If it is the former, the TSB would decline irrespective of the intervention. How are the authors going to handle this situation? Infants whose TSB levels are already on a downward trend will be included in the study, but this is not a potential confounding factor, since those infants will be randomly assigned to SPT or CPT. The number of such infants will therefore be roughly evenly balanced between the treatment and control groups, and thus will not bias the comparison. 9. Some clarity is required regarding the unit of enrollment- whether it is the infant or the day. The authors calculated the sample size according to number of days; then stated that a typical course of treatment lasts one day; and then equated the number of days with the number of subjects. The averaging of values from several days per subject seems to suggest that values from several days will be telescoped into one day. The unit of enrollment is the infant. The unit of statistical analysis is the treatment day as described in Question 3 above. In the sample size calculation, we determined the number of treatment days that would be required, and then used estimates of the number of treatment days typically needed per infant in order to determine the number of infants to enroll. 10. The authors should explain why they have left the criteria for doing an exchange transfusion to the discretion of the treating physician who may either follow AAP guidelines or a standard cut-off of 20 mg/dl. Why have they not specified the exact criteria for the purpose of this research study, despite the fact that this is an open-label trial and prone to performance and ascertainment bias.

6 The study physicians were not the primary treating physicians and were only responsible for the phototherapy. The study physicians do not have privileges to admit to the nursery or to do an exchange transfusion in this hospital. Therefore it was imperative that we leave the final decision to do an exchange transfusion to the primary physicians caring for the infant and to use their criteria for an exchange transfusion. We did confer with these physicians and reviewed our criteria for exchange transfusions and gave them recommendations. But again as it is ultimately the decision of the primary treating physician to determine care that was beyond the scope of our study. Although there is no uniform criterion across institutions in Nigeria, as a rule-thumb the threshold of 20mg/dL for exchange transfusion is the most widely used. 11. There are a large number of criteria for withdrawal. While some of these are understandable and are necessary for ensuring the safety of the infant, there is a concern that too many postrandomisation exclusions could vitiate the process of randomisation. The authors must justify the high number of post-randomisation withdrawal criteria. Some of the post-randomisation withdrawal criteria are such that they would push the direction of research towards the null value, something that is undesirable in a non-inferiority trial. The exclusion criteria (see page 12) are applied at the time of enrollment, before randomization, and thus would not bias the treatment comparison. The withdrawal criteria (see page 12) are primarily concerned with the safety of the infant. The evidence from our previous studies indicates that the safety of SPT is very high and that withdrawal from the study for any reason other than completion of treatment is quite rare. We have no reason to expect that infants assigned to SPT will need to be withdrawn from the study for any of the listed reasons more often than those assigned to CPT, so the treatment comparison should not be biased. 12. It is not clear whether a per protocol analysis or an intention to treat analysis will be performed or both. Unlike a superiority trial, in a non-inferiority trial the per protocol analysis gives a more conservative estimate. Conventionally, both are reported and the per protocol analysis is considered the primary analysis. However the authors do not dwell on this issue at all. The authors originally planned to compare the therapies on an intent-to-treat basis. We appreciate the reviewer s reminder that intent-to-treat is not necessarily conservative for non-inferiority trials and have modified our plan accordingly [see page 20, paragraph 2]. We have revised the analysis plan to indicate that we will conduct three distinct analyses: 1) an intent-to-treat analysis based on the assigned treatment for each enrolled infant 2) an as-treated analysis based on the actual treatment received on a given day

7 3) a per-protocol analysis including only enrolled infants who received the assigned treatment on all of their treatment days. Preliminary results from our trial suggest that the three analyses will likely differ very little: at six months, the actual treatment differed from the assigned treatment on only two of the 370 treatment days, and the number of enrolled infants who had not received the assigned treatment on all of their treatment days was only two out of 264. Minor essential revisions 13. The intervention (FS-PT) is designed not for a place which has no access to CPT, electricity et cetera, but for a place which has partial access to these facilities. Every infant will require CPT at night, when it is raining or very cloudy. The sense- that the intervention is only a partial replacement for the standard of care- does not come out very well in the introduction or discussion. Our concept of FS-PT is not intended exclusively for places with access to CPT. However, we are ethically bound to carry out this RCT study in a centre with CPT facilities to provide basis for evaluating its efficacy compared to CPT, and by extension, compared to no treatment. Based on our findings, we shall be in a position to recommend its usage in places where there is no CPT or any form of treatment. 14. The standard deviation of the average irradiance of FS-PT from the observational study is missing. The standard deviation has been added [see page 6, paragraph 1]. 15. The authors must explain why they have adopted a fairly liberal definition of normothermia (35.5 to 38). A newborn infant cannot be considered safe at temperatures outside I am uncomfortable with this definition of normothermia and the authors must explain why they chose such wide limits. There is also no mention of cold stress. We wished to have safe but broad enough criteria to make continued FS-PT possible but narrow enough criteria to make it safe and we felt these ranges did so. We were unable to find data supporting/suggesting harm from mild hypothermia (well above that allowed in neonatal hypothermia protocols) or mild hyperthermia. One member of our team is an experienced neonatologist (VYE) and the other a pediatric intensivist (TMS) with extensive experience and training in neonatology in low-resource settings; the criteria were also reviewed and approved by others. In addition, during year one of the study prior to the RCT infants were monitored closely and no untoward effects of temperatures within the chosen range were noted. This point has been added to the text [see page 17, paragraph 3].

8 16. Since the air blue canopy is superior, what is the harm in using it at all times; be it overcast or sunny? A single device would make it operationally (and cost wise) much easier. The Air Blue canopy is superior only with respect to transmission of therapeutic blue light. Our laboratory studies (ref 25) have demonstrated that the safety and efficacy of FS-PT depend primarily on the following three radiological characteristics of sunlight: We found that nearly all tested films prevent transmission of harmful UV radiation by 99%, thus exposing subjects under the film canopy to insignificant levels of UV, supporting our conclusion that FS-PT is safe in this respect. Furthermore, blocking the transmission of IR radiation (radiant heat) is engineered, and found to be an important characteristic that varies between films. When optimized, exclusion of IR contributes to thermal safety and comfort of newborn and caregiver during FS-PT. Finally, in order to make PT effective AND safe, the level of therapeutic blue light needs to be modulated and optimized for this characteristic. Minimally effective to adequate PT occurs with 8-10 µw/cm 2 /nm and a level of 30 µw/cm 2 /nm is considered to constitute intensive PT (ref 26). The most powerful PT device on the market delivers ~59 µw/cm 2 /nm (see Vreman et al. 2008). Thus, the irradiance for clinically-effective PT ranges from 8-30 µw/cm 2 /nm and possibly higher. However, direct sunlight can deliver up to 120±20 µw/cm 2 /nm of blue light of nm (as measured with a BiliBlanket Meter II). In contrast, an overcast sky only can deliver ~20±5 µw/cm 2 /nm. Thus, for safe PT with direct sun, the irradiance needs to be reduced by at least 50%, hence the need for the Titanium film canopy, which was found to transmit 33% of blue light, yielding a maximal irradiance of 33% x 140 µw/cm 2 /nm= 46 µw/cm 2 /nm. However, the Air Blue 80 film has a transmission of 79% and this could yield an irradiance of 79% x 120 µw/cm 2 /nm= 95 µw/cm 2 /nm. This is clearly well beyond that considered safe at present. However, if it can be demonstrated in a future study that this level of irradiance is safe, only one film, i.e. Air Blue, could be used. In contrast, PT with the two films under an overcast sky, would generate irradiances of: 33% x 20 µw/cm 2 /nm= 6.6 µw/cm 2 /nm and 79% x 20 µw/cm 2 /nm= 15.8 µw/cm 2 /nm, respectively. The former value is clearly too low for efficacy. Indeed, we have observed that moving mothers and children between two different canopies rarely occurs in practice. In fact, we have been debating if a canopy is needed during overcast sky when the irradiance is already suboptimal. After all, both UV and IR irradiation are also massively absorbed by a cloud cover. However, under this circumstance, safety could be compromised if the mothers would not move into the shade of the canopy when the sun reappears. However, the Reviewer s comment is thought-provoking and appreciated. It clearly indicates that more research is needed towards making the method of using FS-PT simpler, more cost effective, and more practical.

9 17. The authors have not mentioned where the infants will be placed. In the hospital courtyard, balcony, terrace, near an open window? To get sunlight from 10 am to 5 pm, it would have to be completely open area. In that case, how are issues such as exposure to wind, dust etc going to be handled? We have added a paragraph (see page 16, paragraph 3) describing the optimal location where canopies should be placed: It is imperative that the treatment canopies be placed in a location with proven day-long sun exposure throughout the year, preferably one also protected from potentially violent wind currents and surface dust generated by traffic. In fact, the single most important reason for replacing films is wind damage. A roof top terrace, presently in use, has proven to be a very satisfactory location. A courtyard of adequate size surrounded by low buildings could be even more suitable. Finally, perhaps most ideal and practical, will be the construction of a suitable FS-PT room or a small permanent building with a glass or plexiglass roof, in an appropriately sunny part of a hospital. This would be a most ideal solution to prevent issues with wind, dust, security, etc. However, such a treatment facility may overtax local resources. 18. The level of nursing care required for the FS-PT seems to be extraordinarily high and I have concerns whether this can be sustained outside a research environment in a resource-poor setting. It would require very frequent monitoring of temperature, pink skin and irradiance. The authors should comment on this. We agree with the reviewer that this level of nursing care may be impracticable in many resource-poor settings. However, these frequent measurements were intended to guide us appropriately in giving safe advice to health care providers using these devices after the study especially in areas where monitoring is limited. 19. Will the authors be measuring the sun s irradiance directly as well? Do the films permit unknown fraction of the irradiance in the desired wavelengths through? Or is it a fixed amount of irradiance? Irradiance will be measured every minutes. The % of blue light is constant for a given film and was determined earlier as reported in Ref 25. See reply to question 16 for more details. Discretionary revisions 20. Under normal circumstances, newborns are hardly ever get exposed directly to sunlight. I have yet to see a case of a sunburnt neonate and I'm not sure whether this is a common phenomenon in Nigeria. Do the doctors and nurses have sufficient expertise in diagnosing sunburns? Is onset of pink skin a sufficient marker of sunburn?

10 We consulted with a dermatologist regarding the best way to diagnose sunburn and this is what was recommended. Doctors and nurses in Nigeria are able to note new onset red skin with sufficient accuracy to determine if the infant is likely sunburned. Of note, the film characteristics are such (removing 99% of sunburn-causing UV radiation) that it would be almost impossible for an infant to become sunburned while under the film. However, since this is a real concern with direct sunlight, we wanted to include sunburn as a potential problem and then demonstrate that it is not an issue with the use of FS-PT. Editorial requests: Remove the ethical and funding documents from the additional files section? We do not publish such confidential information. Done