Labib 2019.

Study characteristics
Methods	Randomised clinical trial, parallel groups Follow‐up: 12 months Setting: Misr International University in Cairo, Egypt
Participants	115 participants, 132 permanent teeth Age: 18–47 years; mean 29.25, SD 6.8 years Inclusion criteria: men and women (aged 18–50 years), able to tolerate necessary restorative procedures, willing to sign the informed consent, accepted the follow‐up period, posterior permanent tooth with occlusal/occlusal/proximal deep carious lesion, radiographically (bitewing radiograph) extending to the inner 1/3 of dentine (D3) with a radio‐opaque layer between the carious lesion and the pulp chamber. Sensible teeth according to cold pulp test Exclusion criteria: allergy to any of the restorative materials, people undergoing orthodontic treatment with fixed appliances, pregnant women, people with debilitating systemic diseases, teeth with previous restorations, spontaneous pain or prolonged pain (> 15 seconds) after sensitivity test (cold test), which would indicate irreversible pulpitis, negative sensibility tests, periapical radiolucencies and sensitivity to axial or lateral percussion; mobile teeth, indicating periodontal disease or trauma; external or internal resorption
Interventions	2 treatment arms Both arms followed the same baseline treatment protocol: under rubber dam isolation and local anaesthesia, carious tissue removal to hard dentine performed peripherally, while soft dentine close to pulp (pulpal floor/axial wall) was left and restored with glass ionomer material. Group 1 (66 teeth): SE: at a second appointment, the glass ionomer was cut back pulpally or axially (or both), to act as a base leaving a sufficient bulk of 2 mm for the final restoration. For proximal preparations, a sectional matrix system (Palodent Plus, Dentsply) was used. Selective etching of enamel was performed using 35% phosphoric acid gel (Scotchbond Universal Etchant, 3M ESPE, St Paul, NM, USA) for 15 seconds, rinsed for 15 seconds with water, gently air dried with water‐free/oil‐free air for 5 seconds and excess moisture blot‐dried using absorbent tissue. A single layer of universal adhesive (Single Bond Universal, 3M ESPE, St Paul, NM, USA) was applied to the entire cavity preparation using an applicator brush, rubbed in for 10 seconds, air dried for 5 seconds and light‐cured for 20 seconds as described above. The overlaying resin composite (Filtek Z350 XT, 3M ESPE, St Paul, NM, USA) was applied in 2‐mm thick increments, which were light‐cured for 20 seconds as described. The restoration was contoured and finished while polishing was achieved using 1‐step polishing system (Dimanto, Voco, Germany). Group 2 (66 teeth): SW: at a second appointment, re‐entry for the carious lesion was performed by total removal of the glass ionomer and the residual carious lesion until firm dentine remained. The cavity was partially filled again with the highly viscous glass ionomer to serve as a base. The final composite resin restoration was provided using the same restorative procedures as described for group 1.
Outcomes	Primary outcome Treatment success: absence of endodontic/restorative (pulp necrosis, irreversible pulpitis, postoperative pain, need of tooth extraction)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The randomisation unit was the tooth. Sequence generation performed using Microsoft Excel (Microsoft, Washington, USA) by an independent contributor. Block randomisation was performed with 1:1 allocation ratio using random block sizes of 34, 30, 30, 30 and 8. Randomisation sequence and block sizes were concealed from the primary investigator and other operators.
Allocation concealment (selection bias)	Low risk	Quote: "The randomisation unit was the tooth. Sequence generation was performed using Microsoft Excel (Microsoft, Washington, USA) by an independent contributor. Block randomisation was performed with 1:1 allocation ratio using random block sizes of 34, 30, 30, 30 and 8. Randomisation sequence and block sizes were concealed from the primary investigator and other operators. Participants who came back for their second visit (T2) with vital teeth and showing no signs of failure had each tooth allocated remotely via phone prior to further treatment (n = 113)."
Blinding of participants and personnel (performance bias) All outcomes	High risk	The restorative procedure used 2 steps for all groups; in the first visit, a glass ionomer restoration was placed; in the second step, this was either only cut back to serve as a liner for a composite resin restoration (in SE) or completely removed to allow the second excavation step (SW), followed by reapplication of a glass ionomer base and a resin composite restoration. This was done to prevent operator bias during the first visit and to keep the participant blinded to trial design. However, there was still potential for bias at the second stage as the operator knew whether they were removing all glass ionomer base (for SW) or only partial (for SE).
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quotes: "Randomisation sequence and block sizes were concealed from the primary investigator and other operators." "Assessment of clinical and radiographic criteria for success was performed by two blinded, calibrated outcome assessors for each tooth. In addition, the following data were collected by a dentist who had not been involved in the treatment at all visits, including possible re‐treatment visits in case failures occurred: (1) Materials used. (2) Travel times for the patients. (3) The time needed for each visit. Data collection used standardised forms for each visit."
Incomplete outcome data (attrition bias) All outcomes	Low risk	8% censoring per annum was within pre‐established margins. Similar numbers lost to follow‐up, 7 to 8 T2 and 12 to 14 at T3. Methodologically consistent, but could not rule out that lost to follow‐up was not related to the treatment provided in violation of independent censoring. In contrast, the accounting of the extraction as a failure instead of censoring is a satisfactory management of this risk.
Selective reporting (reporting bias)	Low risk	Quote: "Reporting of this trial follows the Consolidated Standards of Reporting Trials (CONSORT) guidelines to ensure transparent and complete reporting." All expected outcomes reported.
Other bias	Low risk	No other biases detected.