Participants preference was assessed based on the Question 1 of PPQ. Question 1 (All things considered, which route of administration did you prefer?) asks participants to report their preference for the route of administration (IV, SC, or no preference). A point estimate with associated 95% CI for the percentage of participants who preferred atezolizumab SC was calculated. Participants experiencing any of the following events: treatment withdrawal prior to eligibility for PPQ, or death without answering Question 1 of PPQ, or treatment not started; were excluded from the analysis set. Participants who answered Question 1 of the PPQ without having at least 2 consecutive administrations of treatment with each administration modality (SC and IV) were excluded from the analysis. Percentages have been rounded off. As planned, participant preference was summarized and presented by overall (all participants) and by randomized treatment sequence using FAS.

FAS included all randomized participants. Overall number analyzed is the number of participants who answered Question 1 of PPQ.

TASQ is a 12-item, participant-reported questionnaire measuring the impact of each mode of treatment administration (TASQ-IV for IV treatment and TASQ-SC for SC treatment) on five domains: Physical Impact, Psychological Impact, Impact on Activities of Daily Living, Convenience, and Satisfaction. TASQ-IV/-SC was administered at treatment Cycles 3 and 6 according to the order of treatment received per arm during the Crossover Period. Participants satisfaction was assessed based on the Question 1 of TASQ-IV/SC which asks participants about their satisfaction with respect to route of administration (very satisfied, satisfied, neither satisfied nor dissatisfied, dissatisfied, participant did not answer the question). Question 1 - TASQ- IV (How satisfied or dissatisfied were you with the IV infusion?) and TASQ- SC (How satisfied or dissatisfied were you with the SC injection?)

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants who answered Question 1 of TASQ IV/SC.

At Cycle 6, Day 1, participants were expected to select the route of study treatment administration (SC or IV) they would like to receive during the Treatment Continuation Period (starting at Cycle 7). Percentage of participants who chose SC administration have been reported here. Percentages have been rounded off.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis.

The HCPQ- Drug Preparation Area Question 1 was completed by the HCPs within the pharmacy/drug preparation area where atezolizumab IV reconstitution or atezolizumab SC was prepared before the actual drug administration took place. The HCPQs were completed for every participant at each treatment cycle (Cycles 1-6, i.e., 3 cycles of atezolizumab IV followed by 3 cycles of atezolizumab SC or vice versa) of the treatment cross-over period. HCPs responded to the following parts of Question 1 that sought to evaluate the amount of time it took to prepare the IV infusion/SC injection of atezolizumab: "How long (in minutes) did it take to prepare the treatment for use?"

Overall number analyzed included HCPs who completed Question 1 of the survey. The number analyzed included HCPs who completed Question 1 of the survey at the specified treatment cycles.

HCPs within pharmacy/drug preparation area responded to HCPQ-Drug Preparation Area Question 2 at Cycle 6 of Treatment Crossover Period for every participant: "If all IV infusions are switched to SC, please indicate how strongly you agree/disagree with each of following statements: a=Staff will have increased availability for other tasks in pharmacy; b=Administrative procedures around atezolizumab SC will require less time; c=Atezolizumab SC formulations will provide more flexibility for staff in managing their workload; d=Due to ready-to-use atezolizumab SC formulations, potential dosing errors will be avoided; e=Due to ready-to-use atezolizumab SC formulations, there will be less drug wastage; f=Without having to reconstitute the drug, less storage space for atezolizumab SC related supplies will be required in pharmacy; g=Preparation procedures & associated time staff time commitment will be reduced, h=It will ease drug administration for participants with difficult venous access."

Overall number analyzed included HCPs who completed Question 2 of the survey at treatment Cycle 6. Percentages have been rounded off.

HCPs who prepared study treatment within the pharmacy/drug preparation area responded at to the following HCPQ-Drug Preparation Area Questions 3 and 4, at Cycle 6 of the Treatment Crossover Period for every participant: "Looking back over the Atezolizumab treatment sessions, please indicate based on your opinion which administration method: Question 3. Was quickest from start to end of preparation to finish of administration (excluding observation period)?; Question 4. Required less resource use for preparation and administration, for example nursing time, facility costs, equipment etc?" The four available response options were: Atezolizumab IV, Atezolizumab SC, No Difference, and Missing. Percentages have been rounded off.

Overall number analyzed included HCPs who completed Questions 3 and 4 of the survey at treatment Cycle 6.

The HCPQ-Treatment Room Question 1 was completed for every participant at each treatment cycle of the Treatment Crossover Period (Cycles 1-6,i.e., 3 cycles of atezolizumab IV followed by 3 cycles of atezolizumab SC/vice versa) by HCPs who administered treatment to the study participants. HCPs responded to following parts of Question 1 that sought to evaluate the amount of time it took to complete activities related to treatment administration: "If new IV access was needed for this cycle of treatment, please indicate what type of IV access was provided (central venous catheter, peripherally inserted central catheter (PICC)/peripheral vein cannulation) & how long (in minutes) this took to set up (only for participants receiving IV treatment)? How long (in minutes) did it take to administer the treatment, i.e. total infusion duration? How long (in minutes) was the participant in the treatment room for in total?" Durations with non-zero HCP responders have been reported here.

Overall number analyzed included HCPs who completed Question 1 of the survey. For the questions related to IV access, the number analyzed only includes number of HCP responders for participants who required new IV access at a given treatment cycle. The number analyzed included HCPs who completed Question 1 of the survey at the specified treatment cycles. Percentages have been rounded off.

HCPs who administered treatment responded to Question 2: If all IV are switched to SC, please indicate how strongly you agree/disagree with each of following statements: a= Participants will be moved outside of infusion unit to receive SC injections (physician consultation room); b=Atezolizumab SC route will allow more flexible treatment scheduling; c=More participants will be treated in infusion unit; d=Waiting list for any IV treatment at infusion unit will be reduced; e=Staff resources will be re distributed to other departments of the hospital (less staffing required within infusion unit); f=There will still be sufficient interaction time between HCPs & participants (for participant education); g=Staff will spend more time for further professional education/development; h=Staff will dedicate more time to attending to administrative tasks for participants; i=Participants will spend less time in care unit; j=Administration by atezolizumab SC injection is preferred by participants.

Overall number analyzed included HCPs who completed Question 2 of the survey at treatment Cycle 6. Percentages have been rounded off.

HCPs who administered study treatment responded at Cycle 6 of the Treatment Crossover Period to the following HCPQ-treatment room Questions 3 to 7: "Looking back over the atezolizumab treatment sessions, please indicate based on your opinion which administration method: Question 3. Which method was most convenient for the participant? Question 4. Which method was best for optimizing participant care in your center? Question 5. Which method took the least time from start to finish of administration? Question 6. Which method required the least resource use for administration? Question 7. Which method was preferred by participants?" The five available response options were: Atezolizumab SC, Atezolizumab IV, No Difference, Unsure and Missing. Percentages have been rounded off.

Overall number analyzed included HCPs who completed Questions 3 to 7 of the survey at treatment Cycle 6.

HCPs who administered study treatment responded at Cycle 6 of the treatment Cross-over Period to the following HCPQ-treatment room Question 8: How frequently would you offer or recommend atezolizumab SC administration to your participants in the future? The four available response options were Always, Sometimes, Never and Missing. Percentages have been rounded off.

Overall number analyzed included HCPs who completed Question 8 of the survey at treatment Cycle 6.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), global health status (GHS) and quality of life (QoL), and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The functioning items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score indicating better functioning. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The functioning items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score indicating better functioning. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The functioning items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score indicating better functioning. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The functioning items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score indicating better functioning. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The functioning items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score indicating better functioning. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The symptom items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score indicating worst symptoms. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The symptom items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score indicating worst symptoms. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The symptom items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score indicating worst symptoms. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The dyspnoea scale items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score the scale indicating worst symptoms. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The insomnia scale items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score on the scale indicating worst symptoms. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The appetite loss scale items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score on the scale indicating worst symptoms. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The constipation scale items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score on the scale indicating worst symptoms. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The diarrhoea scale items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score on the scale indicating worst symptoms. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess five aspects of participant functioning (physical, emotional, role, cognitive, and social), three symptom scales (fatigue, nausea and vomiting, and pain), GHS and QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties) with a recall period of the previous week. Scale scores can be obtained for the multi-item scales. The financial difficulties scale items are scored on a 4-point scale (1=Not at All to 4=Very Much). Scores are linearly transformed on a scale of 0 to 100, with a high score on the scale indicating worst functioning. EORTC QLQ-C30 data was scored according to the EORTC scoring manual v3.0. In the event of incomplete data, if more than 50% of the constituent items were completed, a pro-rated score was computed. For subscales with less than 50% of the items completed, the subscale was considered as missing.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

EORTC QLQ-C30 consists of 30 questions that assess participant functioning (physical, emotional, role, cognitive, and social), symptom scales (fatigue, nausea and vomiting, pain), global health/ QoL, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties). Change in HRQoL was assessed using participant responses to questions regarding Global Health Status (Q29: GHS; "How would you rate your overall health during the past week?") and QoL (Q30: QoL; "How would you rate your overall quality of life during the past week?") and were scored on a 7-point scale (1= Very poor to 7=Excellent). Using linear transformation, raw scores are standardized. Scores range from 0-100. A higher score indicates a better outcome.

FAS included all randomized participants. Overall number analyzed is the number of participants with data available for analysis. Number analyzed is the number of participants with data available for analysis at specified timepoint.

Participants were assessed for clinical benefit at every tumour assessment visit, which was conducted as per the local standard of care. Percentages have been rounded off.

FAS included all randomized participants. Ongoing clinical benefit at Cycle 16 was summarized for overall participants, as due to switches in mode of administration up to Cycle 16, it would be impossible to isolate the effect of the randomization arms. Hence, per-arm data can't be presented.

An AE is untoward medical occurrence in a participant administered a pharmaceutical product and regardless of causal relationship with this treatment. An AE can therefore be any unfavorable and unintended sign (including an abnormal laboratory finding), symptom/disease temporally associated with use of investigational product, whether or not considered related to the investigational product. Percentages have been rounded off.

Safety evaluable population included all participants who received at least one dose of study treatment.

An AE is untoward medical occurrence in participant administered a pharmaceutical product and regardless of causal relationship with this treatment. An AE can therefore be any unfavorable and unintended sign (including an abnormal laboratory finding), symptom/disease temporally associated with use of investigational product, whether or not considered related to investigational product. The safety of switching from atezolizumab SC to atezolizumab IV and from atezolizumab IV to atezolizumab SC is being assessed in this outcome measure. Percentages have been rounded off.

Safety evaluable population included all participants who received at least one dose of study treatment. In this analysis, participants were grouped by study arm and treatment period during the Crossover Period.