Usefulness of self-administered questionnaires in screening for direct referral for polysomnography without sleep physician review

Abstract

Study Objectives:

To evaluate self-administered screening questionnaires (Epworth Sleepiness Scale [ESS], Berlin, OSA50, and STOP-Bang questionnaires) in patients considered for polysomnography for probable obstructive sleep apnea suitable for direct polysomnography without sleep specialist review and to evaluate the usefulness of combining questionnaires in this population.

Methods:

This was a retrospective review of tertiary sleep center referrals (November 2017 to April 2020) where ≥ 3 screening questionnaires were completed and type 1 polysomnography was performed. Sensitivity, specificity, positive and negative predictive values, and likelihood ratios to detect an apnea-hypopnea index (AHI) ≥ 15 or ≥ 30 events/h were calculated for each questionnaire (with or without ESS ≥ 8) or any positive questionnaire with ESS ≥ 8.

Results:

We included 2,152 patients. The questionnaires were completed in the majority (ESS 96%, Berlin 77%, OSA50 84%, and STOP-Bang 90%) of referrals. Berlin was most sensitive (82.5% and 85% to detect AHI ≥ 15 and ≥ 30 events/h, respectively) but least specific (23% both thresholds). STOP-Bang was least sensitive (66% and 42%, respectively) but most specific (68% and 60%, respectively). Sensitivity declined for the Berlin, OSA50 and STOP-Bang questionnaires when combined with ESS ≥ 8. Combining any questionnaire with ESS ≥ 8 returned an intermediate sensitivity of 61% and 73% and a specificity of 49% and 47% for AHI ≥ 15 and ≥ 30 events/h, respectively. STOP-Bang alone was predictive of obstructive sleep apnea on multivariate analysis but was only associated with a clinically nonsignificant positive likelihood ratio. However, STOP-Bang is associated with unacceptable false-positive and -negative rates, which did not support its use.

Conclusions:

Self-administered questionnaires are inadequate in patients under consideration for polysomnography and should not be used as clinical support for suitability of direct polysomnography without sleep specialist review. Combining questionnaires causes deteriorated performance.

Citation:

Hukins C, Duce B. Usefulness of self-administered questionnaires in screening for direct referral for polysomnography without sleep physician review. J Clin Sleep Med. 2022;18(5):1405–1412.

BRIEF SUMMARY

Current Knowledge/Study Rationale: There are numerous screening questionnaires for obstructive sleep apnea that have been validated in primary care or sleep clinic populations but their use in selecting patients for direct polysomnography without sleep specialist review is unclear and not validated.

Study Impact: Screening questionnaires do not have a role in selecting patients with suspected sleep-disordered breathing appropriate for direct polysomnography.

INTRODUCTION

Obstructive sleep apnea (OSA) is underdiagnosed in the community, partly due to its reliance on a model of care involving sleep specialist review followed by polysomnography (PSG). Limited access to a sleep specialist, whether because of excessive demand or distance to the nearest specialist, can often impede the prompt diagnosis and treatment of OSA. A proposed solution to overcome this block is direct access to PSG without prior sleep specialist review. The direct access model requires a health professional (often the primary care physician), with minimal to no specific training in sleep medicine, to recognize the condition and determine the best diagnostic investigation. The inherent risk for the individual is a false-negative test due to inappropriate test selection but there is also risk for the community. Inappropriate patient selection will result in a substantial number of negative tests, which increases the cost to government and/or health insurance. To reduce this risk, utilization of self-administered screening questionnaires has been suggested to help identify patients suitable for direct-access PSG.

Self-administered screening questionnaires have been evaluated extensively in the past to estimate the likelihood of OSA in both general populations and sleep clinic populations. The most common questionnaires are the Berlin, STOP-Bang, and OSA50 questionnaires. The Berlin Questionnaire (BQ) ¹ was initially validated in primary care and comprises 12 self-administered questions centered on 3 clinical features of OSA (snoring, tiredness or fatigue, and hypertension or obesity). The BQ has a binary outcome: high risk if 2 or more categories are positive or low risk. The STOP-Bang ² was initially developed as a preoperative screening tool for OSA and is a composition of 8 yes/no questions related to snoring, tiredness, observed apneas, blood pressure, body mass index, age, neck circumference, and male sex. It has been evaluated in the general and sleep clinic populations. ³ Scores of 0–2 are regarded as low risk, 3–4 as intermediate risk, and 5–8 as high risk for moderate to severe OSA. ⁴ The optimal cutoff score to predict OSA is reported as either 3 ⁵ or 4. ⁶ The OSA50 ⁷ was developed for primary care as a 2-part screening tool of a self-administered questionnaire (weighted questions on obesity, snoring, apneas, and age of 50 years or over) followed by oximetry. The questionnaire alone has been used in sleep clinic ⁸ and other high-risk ⁹ populations. The reported cutoff score for this questionnaire is ≥ 5/11. ⁷ The sensitivity and specificity of these questionnaires are highly variable (BQ 24%–95% and 16%–83% and STOP-Bang 66%–97% and 5.5%–75%, respectively, for moderate to severe OSA), ¹⁰ reflecting the impact of study population selection on outcomes.

The use of screening questionnaires as a tool for selecting patients suitable for directly referred PSG is appealing. The rationale is to select patients with a high probability of OSA to justify the performance of expensive PSG. However, these questionnaires have not been evaluated for this purpose but are already utilized in the real world for this reason. For example, publicly funded PSG without sleep physician review is available in Australia where “the patient has been referred by a medical practitioner to a qualified sleep medicine practitioner or a consultant respiratory physician who has determined that the patient has a high probability for symptomatic, moderate to severe obstructive sleep apnea based on a STOP-Bang score of 4 or more, an OSA50 score of 5 or more or a high risk score on the BQ, and an Epworth Sleepiness Scale (ESS) score of 8 or more.” ¹¹ This represents not only the unvalidated use of self-administered questionnaires to determine probability of OSA in a population already likely to require PSG but also the unvalidated combination of different questionnaires to determine eligibility.

There are a number of aims to this study. The first aim was to evaluate the utility of self-administered screening questionnaires in assessing suitability for direct PSG without sleep physician review (based on their ability to predict moderate-to-severe or severe sleep apnea) in a population suspected of sleep-disordered breathing. The second aim was to assess the impact of combining separately developed and validated questionnaires using the Australian funding system as a model. These aims were approached by addressing 3 questions:

• Question 1.

  How well do screening questionnaires identify moderate-to-severe or severe OSA in patients with a high suspicion of OSA and under consideration for PSG?

• Question 2.

  Do the screening questionnaires predict likely severity of OSA in this population?

• Question 3.

  What are the optimal thresholds for screening questionnaires in this population?

METHODS

This was a retrospective cohort study. Patients referred to our tertiary sleep disorders service between November 1, 2017, and April 30, 2020, and who subsequently underwent an attended diagnostic sleep study were identified from our in-house database (Filemaker Pro, Santa Clara, California). The service has used electronic medical records since 1997 with customized databases. All data are collected prospectively. Our routine clinical practice is to screen all referrals by a self-completed questionnaire that includes the ESS, the BQ, the STOP-Bang questionnaire, and the OSA50 questionnaire. Patients were included in the study if they successfully completed at least 3 of the ESS, BQ, STOP-Bang, or OSA50 questionnaires. All patients suspected of OSA undergo in-laboratory PSG. PSG data were recorded with the Compumedics Grael acquisition system (Abbotsford, Australia). The recording montage comprised electroencephalogram (F4-M1, C4-M1, O2-M1), left and right electrooculogram (recommended derivation: E1-M2, E2-M2), chin electromyogram (mental/submental positioning), modified lead II electrocardiogram, nasal pressure (Direct Current amplified), oronasal thermocouple, body position, thoracic and abdominal effort (inductive plethysmography), pulse oximetry, left and right leg movement (anterior tibialis electromyogram), and sound pressure (dBA meter: Model DSL-332, Tecpel Co. Ltd, New Taipei City, Taiwan). PSGs were scored according to the 2012 version of The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications ¹² with Compumedics Profusion 4.0 (Compumedics Ltd, Melbourne, Australia, Build 468) software. Hypopneas were scored using the recommended rules. The protocol was approved by the institution’s human research ethics committee.

Question 1. How well do screening questionnaires identify moderate-to-severe or severe OSA in patients with a high suspicion of OSA and under consideration for PSG?

Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+), and negative LR (LR−) were calculated for the following:

• 1. Individual questionnaires

  • a. ESS (< 8 compared to ≥ 8),

  • b. BQ (low risk compared to high risk)

  • c. OSA50 (< 5 compared to ≥ 5)

  • d. STOP-Bang (< 4 compared to ≥ 4)

• 2. Individual questionnaire and an ESS ≥ 8

  • a. BQ + ESS (positive if high risk and ESS ≥ 8)

  • b. OSA50 + ESS (positive if ≥ 5 and ESS ≥ 8)

  • c. STOP-Bang + ESS (positive if ≥ 4 and ESS ≥ 8)

• 3. Australian model, a composite requirement modeled on the Australian Medical Benefits Scheme (positive if ESS ≥ 8 and high risk on Berlin or OSA50 ≥ 5 or STOP-Bang ≥ 4).

Multivariate analysis was performed by multiple logistic regression to determine the optimal combination of questionnaires to predict apnea-hypopnea index (AHI) ≥ 15 and ≥ 30 events/h.

Question 2. Do the screening questionnaires predict likely severity of OSA in this population?

The individual and combined questionnaires were compared with the AHI by chi-square analysis if outcomes were binary (Berlin and combined questionnaires) or by analysis of variance (ANOVA) on ranks for numerical outcomes.

Question 3. What are the optimal thresholds for screening questionnaires in this population?

The optimal thresholds of the Berlin, OSA50, and STOP-Bang questionnaires were assessed in this population by plotting the receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC).

Statistical analysis

Statistical analysis was performed with SigmaPlot 14.0 (Systat Software, Inc, San Jose, California). Data are presented as mean (95% confidence interval [CI]) where the distribution was parametric and as median (interquartile range) where the distribution was nonparametric. A P value < .05 was considered statistically significant.

RESULTS

There were 2,152 referrals within the study period that met the inclusion criteria. Patient characteristics are shown in Figure 1. A total of 2,075 referrals had acceptable ESS data; 1,655 referrals had acceptable BQ data; 1,938 referrals had acceptable STOP-Bang data; and 1,802 had acceptable OSA50 data.

Figure 1. Characteristics of study participants.

AHI = apnea-hypopnea index, BMI = body mass index, ESS = Epworth Sleepiness Scale, IQR = interquartile range.

Question 1. How well do screening questionnaires identify moderate-to-severe or severe OSA in patients with a high suspicion of OSA and under consideration for PSG?

The performance of the screening questionnaires to detect at least moderate or severe OSA is shown in Figure 2. The BQ without ESS criteria was the most sensitive but the least specific. STOP-Bang had the highest specificity for both AHI ≥ 15 and AHI ≥ 30 events/h and performed best in terms of PPV and LR+. Adding the requirement for an ESS ≥ 8 (BQ + ESS, OSA50 + ESS, and STOP-Bang + ESS) decreased the sensitivity of identifying moderate-to-severe OSA or severe OSA. The Australian model (a combination of any positive screening questionnaire and an ESS ≥ 8) had lower sensitivity than any individual questionnaire (without the ESS requirement). PPV was consistently lower and NPV higher for an AHI ≥ 30 events/h compared with an AHI ≥ 15 events/h.

Figure 2. Sensitivity, specificity, predictive values (positive and negative), and likelihood ratios (positive and negative) for detection of AHI ≥ 15 events/h (shown in black) and AHI ≥ 30 events/h (shown in red).

Results are shown for each questionnaire alone (square symbol) or with the additional requirement of an ESS score of ≥ 8 (triangular symbol). The criteria for the Australian Model are a combination of any positive screening questionnaire and an ESS ≥ 8. AHI = apnea-hypopnea index, ESS = Epworth Sleepiness Scale.

By multiple logistic regression analysis, only the STOP-Bang was significantly associated with both AHI ≥ 15 and AHI ≥ 30 events/h, showing odds ratios of 1.66 (95% CI, 1.51–1.82) and 1.53 (95% CI, 1.40–1.68), respectively (P < .001).

Question 2. Do the screening questionnaires predict likely severity of OSA in this population?

The differences in AHI between outcomes of screening questionnaires are shown in Figure 3. The AHI was significantly different at varying questionnaire results (P < .001 in all cases), with evidence of a dose-response relationship between the AHI and increasing values of numerical screening questionnaires. Nevertheless, there is substantial overlap in the lower range of AHI between all screening questionnaire values, which limits the practicality of estimating the likely severity of OSA.

Figure 3. AHI at the different results for questionnaires.

(A) Australian criteria: combination of any positive screening questionnaire and an ESS ≥ 8. (B) Berlin Questionnaire. (C) OSA50 questionnaire. (D) STOP-Bang Questionnaire. The box shows the upper quartile, median, and lower quartile. The whiskers show the 95% CI and the data points show the range of the data. AHI = apnea-hypopnea index, CI = confidence interval, ESS = Epworth Sleepiness Scale.

Question 3. What are the optimal thresholds for screening questionnaires in this population?

The sensitivity and specificity for detecting an AHI ≥ 15 or an AHI ≥ 30 events/h at different cutoff values for ESS, OSA50, and STOP-Bang are represented by the ROC curves in Figure 4. This assessment could not be performed for the BQ, Australian Model, OSA50 + ESS, or STOP-Bang + ESS where the outcomes are binary (positive or negative). The AUCs of the ROC between AHI ≥ 30 and AHI ≥ 15 events/h were similar for each. The ESS ROC curve followed the line of identity with an AUC of 0.50–052 and therefore no threshold was predictive of either severity of OSA. The STOP-Bang had the highest AUC of 0.72 for both OSA severities. The optimal cutoff in this population for STOP-Bang was ≥ 4 and for OSA50 was ≥ 5, which are the thresholds commonly recommended from other populations and used in this study.

Figure 4. ROC of the questionnaires in detecting AHI ≥ 15 events/h.

(A) ESS score. (B) OSA50. (C) STOP-Bang Questionnaires and AHI ≥ 15 events/h. (D) ESS score. (E) OSA50. (F) STOP-Bang Questionnaires and AHI ≥ 30 events/h. “A” refers to area under the curve. AHI = apnea-hypopnea index, ESS = Epworth Sleepiness Scale, ROC = receiver operating characteristic.

Figure 5 shows the relationship between false-positive and false-negative rates to detect an AHI ≥ 15 events/h at different STOP-Bang thresholds. A low threshold was associated with high false-positive rates (35% for a threshold of ≥ 2), but thresholds of just ≥ 4 were associated with false-negative rates of greater than 40%, with false-positive rates still up to 24%. However, when STOP-Bang thresholds were considered in association with an ESS of ≥ 8, there was a far more flattened relationship with false-negative rates of  > 50% at all thresholds.

Figure 5. Plot of false-positive and false-negative rates for the STOP-Bang questionnaire alone (closed circles) and STOP-Bang and an ESS score ≥ 8 (open circles) at different thresholds of STOP-Bang (thresholds are shown as annotations).

ESS = Epworth Sleepiness Scale.

DISCUSSION

This study assessed (1) whether commonly used screening questionnaires (ESS, BQ, STOP-Bang, and OSA50 questionnaires alone or in combination) can be used to assess suitability for direct PSG without sleep physician review in a population highly selected for probable OSA, (2) whether screening questionnaires predict likely severity of OSA in this population, and (3) the optimal thresholds for screening questionnaires in this population. Our data showed moderate differences in sensitivity between questionnaires to detect moderate-to-severe OSA (65.5%–82.5%) but smaller differences to detect severe OSA (73.5%–85%). We also assessed the effect of combining questionnaires, using the Australian model of funding for direct PSG requiring ESS of ≥ 8 and a positive screening questionnaire (high risk on the BQ or OSA50 ≥ 5 or STOP-Bang ≥ 4). Our data demonstrated that the combination of different questionnaires (either the ESS with 1 of the questionnaires or the choice of ≥ 1 questionnaires) did not improve and in many cases weakened the performance of the questionnaire. Last, we showed that the cutoffs of ≥ 5 for OSA50 and ≥ 4 for STOP-Bang were the most appropriate in this population, albeit with low to moderate accuracy. There was no appropriate cutoff value for the ESS.

Despite its increasing recognition in the community, OSA remains underdiagnosed. The ability to bypass sleep specialist review and go straight to PSG may improve access to diagnosis and treatment. However, improving access to sleep medicine services must be moderated by the costs borne by the community. It has been suggested that screening questionnaires may aid with selection of suitable patients, but there are currently no screening questionnaires specifically developed for this purpose. A number of screening questionnaires that have been developed for primary and surgical care, such as the BQ, the STOP-Bang, and the OSA50, could ostensibly be used to select direct-to-PSG patients. In this study, we found that there were statistical differences in AHI between positive and negative values of binary outcome questionnaires (BQ and the composite of ESS and a positive questionnaire) and at different cutoff values for numerical scores (OSA50 and STOP-Bang). However, there was significant overlap in the range of AHI at every level. This would indicate that these questionnaires do not provide any useful guidance on the likely severity of OSA, which could be useful in determining those suitable for direct PSG. Interestingly, adding the requirement for an ESS ≥ 8 significantly reduced the sensitivity of STOP-Bang for both an AHI ≥ 15 and AHI ≥ 30 events/h. There was also wider variability in specificity between questionnaires and a significantly lower sensitivity of the STOP-Bang for an AHI ≥ 15 events/h if an ESS ≥ 8 was also needed.

It is important to consider which calculation is most important in this scenario. Sensitivity and specificity most commonly assess the detection of disease in the community, which is arguably of lesser importance in a prescreened population with high pretest probability for OSA. The PPVs and NPVs are of more importance where these questionnaires are used by government for funding of direct PSG, as these values are more closely related to the inherent accuracy of the test (although also affected by disease prevalence in the population). Predictive values indicate the absolute probability that disease is present with a positive result (PPV) or absent with a negative test result (NPV). ¹³ From a funding perspective, a high PPV (and low false-positive result) is desirable to reduce “unnecessary polysomnography” and make this expensive investigation more cost-effective. A high NPV is also desirable to reduce the number of patients who ultimately require PSG needing sleep specialist review (and the added public cost of that review). Our data showed a PPV of only 37%–53% for severe OSA and 45%–73% for at least moderate OSA and an NPV of 63%–79% for severe OSA and 45%–73% for at least moderate OSA. Therefore, if the decision to proceed with PSG was based on the results of the questionnaire(s) alone, then 25%–63% of directly referred PSG would fail to show at least moderate OSA. The likelihood ratios are arguably of most importance from an individual patient perspective as these change the pretest probability of disease. However, in a high-pretest-probability population, an LR+ of 2 (which doubles the pretest probability) or LR− of 0.5 (which halves pretest probability), as seen with our data, is unlikely to affect subsequent clinical management. In fact, when considering a high-probability population such as this, positive likelihood ratios of 5–10 and negative likelihood ratios of 0.1–0.2 are required to alter clinical decisions. ¹³

The ability to specifically screen for and expedite investigation of severe OSA has public health significance. Severe OSA has greater associations with incident hypertension, ¹⁴ incidence of and outcomes from stroke, ^{15, 16} and all-cause mortality. ¹⁷ Thus, the ability to predict severity would allow for appropriate prioritization of severe OSA when resource allocation needs to be rationalized. All questionnaires had significantly lower PPV and higher NPV for AHI ≥ 30 events/h compared with AHI ≥ 15 events/h, as expected, because these values are influenced by prevalence (moderate-to-severe OSA is more prevalent than severe OSA). The STOP-Bang (with or without the requirement for ESS) had the highest positive likelihood ratio and lowest negative likelihood ratios; however, these were not clinically significant. Only STOP-Bang was significantly associated with an AHI ≥ 15 and AHI ≥ 30 events/h on multivariate analysis, albeit with modest odds ratios (1.66 and 1.53, respectively). There was significant overlap in the range of AHI at every threshold of these questionnaires and therefore they do not provide any useful guidance on the likely severity of OSA. With the data supporting the STOP-Bang as the best overall tool, we assessed its screening performance in terms of false-positive and false-negative rates at cutoffs between ≥ 2 and ≥ 8. Our data did not support its use. At a low threshold of ≥ 2, the false-positive rate was 35%, which results in a significant waste of resources. On the other hand, an intermediate cutoff of 4 resulted in a false-positive rate of 27% with a false-negative rate of over 40%. Higher cutoff values with a STOP-Bang of ≥ 8 were associated with a false-positive rate of 10% but a false-negative rate of over 55%. The cost implication of a high false-positive rate is a high rate of direct-referral PSG in people without OSA. The impact of a high false-negative rate is a larger number of people requiring sleep physician review before PSG. Either outcome will increase health costs. Overall, our findings showed that, although STOP-Bang demonstrated the best associations with OSA and severity, it is not clinically useful in determining those suitable for direct PSG.

This study also explored whether the recommended cutoff values were appropriate in the selected population (likely OSA under consideration for the type of sleep study). Our data demonstrated that the cutoffs for OSA50 (≥ 5) and STOP-Bang (≥ 4) were appropriate for this population, albeit with low to moderate accuracy as shown by ROC AUCs of 0.65–0.72. ¹³ There was no appropriate cutoff value for the ESS (as demonstrated by ROC AUCs of 0.50–0.52, suggesting a chance result ¹³).

There are differences between our results and those of another recent Australian study. Senaratna and colleagues ¹⁸ assessed the ESS, BQ, OSA50, and STOP-Bang in a prospective Australian general-population cohort and compared the screening questionnaires with type 4 sleep studies (oximetry and nasal pressure/flow). These authors reported similar sensitivity of the individual questionnaires (65%–86%) but, if combined with ESS, sensitivity was significantly lower than our results at 36%–51% if coupled with an ESS of ≥ 8. However, specificity was higher than our results (81%–86% increasing to 94%–96% if coupled with an ESS of ≥ 8). Positive likelihood ratios of the questionnaires linked to an ESS of ≥ 8 were higher at 6.4–7.3 with negative likelihood ratios comparable to our study. The differences between the 2 studies are most likely due to the different study populations. The participants in the study of Senaratna et al were randomly selected from a prospective general-population cohort. The population in our study was patients with suspected sleep-disordered breathing and therefore preselected for high probability of OSA. In addition, our study compared questionnaires to type 1 attended PSG, while Senaratna et al compared questionnaires to unattended type 4 studies.

Our study has a number of strengths. There were over 2,100 patients in this cohort with reasonable data completeness. Of these included referrals, 98% had acceptable completion of the ESS, 77% of the BQ, 90% of the STOP-Bang, and 84% of the OSA50 questionnaires. This study was a “real-life” assessment, where the population was selected based on the suspicion of OSA, and addresses the usefulness of common questionnaires to determine the appropriate mode of PSG. Therefore, the study is externally valid to assess the specific study questions. Finally, our gold standard for the diagnosis of OSA was type 1 attended PSG in all cases. There are inherent weaknesses in a retrospective study, including risk of selection bias and incomplete data. However, the data in our study were entered prospectively into a comprehensive electronic medical record, which reduces the potential for error. Our study also assumed that the value of screening patients was to identify those with more severe OSA for direct PSG. This assumption has validity as clinical decision-making in mild-to-moderate sleep apnea is more complex and requires more clinical experience in the field. A better target group to identify would be patients with uncomplicated OSA and treatment requirements who did not require specialist input. However, defining such “uncomplicated” patients is too difficult and subjective to use as an outcome measure. The exclusive use of type 1 PSG may affect the external applicability of our findings in populations where unattended sleep studies are frequently performed. However, we do not believe that this is likely to be significant when screening for moderate-to-severe OSA as in this study.

In summary, although the concept of using self-administered questionnaires to select patients with a high probability of more severe OSA and who may be suitable for direct PSG is appealing, we have demonstrated that the questionnaires in this study do not achieve this aim and are associated with unacceptable false-positive and false-negative rates. In addition, combining separately developed and validated questionnaires, such as that used in Australia for the public funding of PSG (such as requiring both a “positive” ESS and “positive” questionnaire or accepting any one of multiple questionnaires), does not improve their predictive ability or, in some cases, reduces their ability.

ABBREVIATIONS

CI

confidence interval

AHI

apnea-hypopnea index

AUC

area under the curve

BQ

Berlin Questionnaire

ESS

Epworth Sleepiness Scale

LR

likelihood ratio

NPV

negative predictive value

OSA

obstructive sleep apnea

PPV

positive predictive value

ROC

receiver operating characteristic

DISCLOSURE STATEMENT

Both authors have contributed to and approved the preparation of this manuscript. The authors report no conflicts of interest.